                                                              1794

 1    UNITED STATES DISTRICT COURT
      EASTERN DISTRICT OF NEW YORK
 2    ------------------------------x
      THE CITY OF NEW YORK, et al.,
 3    
                         Plaintiffs,          88 Civ. 3474 (JMcL)
 4    
                 v.                           
 5    
      UNITED STATES DEPARTMENT OF COMMERCE,
 6    et al., 
      
 7                       Defendants.
      ---------------------------------x
 8    CITY OF ATLANTA, et al.,
                     
 9                       Plaintiffs,
      
10                v.                          92 Civ. 1566 (JMcL)
                                              
11    MOSBACHER, et al.,
      
12                       Defendants.
      ---------------------------------x
13    FLORIDA HOUSE OF REPRESENTATIVES, 
      et al.
14    
                         Plaintiffs,
15                v.                          92 Civ. 2037 (JMcL)
      
16    BARBARA H. FRANKLIN, Secretary of
      Commerce,
17    
                         Defendant.
18    --------------------------------x
      
      
19    
                                              May 22, 1992
20                                            9:30 a.m.
      
21    
      
22    Before:
      
23               HON. JOSEPH M. McLAUGHLIN,
      
24                                            Circuit Judge
      
25
                                                              1795

 1               THE COURT:  I believe we are ready for 

 2    cross-examination.  Mr. Cohen.  

 3    PETER BOUNPANE      resumed
                                 

 4    CROSS-EXAMINATION 

 5    BY MR. COHEN:

 6         Q.    Good morning, Mr. Bounpane.

 7         A.    Good morning, Mr. Cohen.

 8         Q.    Your title is assistant director of decennial 

 9    censuses, is that correct?

10         A.    Yes, that is correct.

11         Q.    It is correct, isn't it, that your prime 

12    responsibilities have to do with coordinating the external 

13    activities relating to the census?

14         A.    In the latter half of the eighties, yes, that is 

15    correct.

16         Q.    That has been since about 1987?

17         A.    Somewhere around there, yes.

18         Q.    That means dealing with mayors, dealing with 

19    Congress, making presentations at groups who have an 

20    interest in the census, and dealing with the media, isn't 

21    that correct?

22         A.    That's correct.

23         Q.    In 1989 you were given additional responsibility, 

24    which is to help out with a lot of the lawsuits concerning 

25    the adjustment case, correct?
                                                              1796

 1         A.    I'm not sure of the exact timing, but somewhere 

 2    around then, yes.

 3         Q.    That responsibility involves giving advice to the 

 4    legal staff, correct?

 5         A.    Yes, it does.

 6         Q.    You have performed that duty with respect to this 

 7    lawsuit, correct?

 8         A.    Yes.

 9         Q.    You have performed that duty with respect to the 

10    suit by the City of Detroit seeking an adjustment of the 

11    census, correct?

12         A.    To a much lesser degree, but yes.

13         Q.    You have performed that duty with respect to the 

14    suit under the Freedom of Information Act in which the 

15    Florida House of Representatives has sought the adjusted 

16    census status, is that correct?

17         A.    That's correct.

18         Q.    You have performed that duty with respect to the 

19    suit by the California State Assembly in which they are 

20    seeking the adjusted census status, correct?

21         A.    Yes.

22         Q.    You have performed that duty with respect to the 

23    suit by the California Senate seeking the adjusted census 

24    status?

25         A.    Yes.
                                                              1797

 1         Q.    You have performed that duty with respect to the 

 2    suit by Massachusetts challenging the bureau's methods for 

 3    counting overseas residents, is that correct?  

 4         A.    Very little with regard to the suit from 

 5    Massachusetts.

 6         Q.    But the answer is yes, isn't it?

 7         A.    Yes.

 8         Q.    You were assigned those duties by Charles Jones, 

 9    correct?

10         A.    In conjunction with the deputy director of the 

11    census bureau.

12         Q.    The answer is yes?

13         A.    Yes.

14         Q.    Mr. Jones is your supervisor, correct?

15         A.    Yes, he is.

16         Q.    His title is associate director for decennial 

17    censuses, correct?

18         A.    Yes.

19         Q.    You and Mr. Jones were both members of the 

20    Undercount Steering Committee, isn't that correct?

21         A.    Yes.

22         Q.    You and Mr. Jones were the only two members of 

23    the Undercount Steering Committee to dissent from the 

24    technical assessment report, correct?

25         A.    Yes.
                                                              1798

 1         Q.    All other members recommended that the adjusted 

 2    counts are more accurate than the unadjusted counts, is that 

 3    correct?

 4         A.    Yes, on that.

 5         Q.    In about February of this year, you had a 

 6    conversation with Mr. Jones in which you and he talked about 

 7    which of the two of you would be a witness at this trial to 

 8    talk about how the census is taken, isn't that correct?

 9         A.    Yes.

10         Q.    You and Mr. Jones agreed that you would be the 

11    witness because you had primarily been the person who would 

12    do the more public aspects of the census, isn't that 

13    correct?

14         A.    Yes.

15         Q.    By that, you meant dealing with mayors, dealing 

16    with Congress, making presentations at groups who have an 

17    interest in the census, dealing with the media, correct?

18         A.    To the extent that that is involved in that, yes.

19         Q.    It is a fact, isn't it, that you appeared on TV 

20    with Secretary Mosbacher when he announced his decision on 

21    July 15th against the adjustment?

22         A.    At the very end of the press conference, yes.

23         Q.    Mr. Bounpane, do you remember that yesterday you 

24    testified about a computer edit problem that the bureau 

25    found in the coding of erroneous enumerations found in the 
                                                              1799

 1    1990 PES?

 2         A.    Yes.

 3         Q.    As a result of that error, the bureau discovered 

 4    that persons who were counted in the census shouldn't have 

 5    been counted, isn't that correct?

 6         A.    Basically, yes.  But it had already been 

 7    identified in the clerical operations of the 

 8    post-enumeration survey.  The error was that that 

 9    designation was not carried forward into the computer 

10    itself.

11         Q.    You told us that you believed that the correction 

12    of that error reduced the differential undercount, didn't 

13    it?

14         A.    Yes.

15         Q.    You didn't tell us how much that reduction was, 

16    did you?

17         A.    No.

18         Q.    In fact, the reduction in the differential 

19    undercount was about .28 percentage point, isn't that is 

20    correct?

21         A.    I'm not sure about that.  If I saw the sheet 

22    again, I would be happy to calculate it.

23         Q.    Mr. Bounpane, if you look to your right, I think 

24    you will find the binder that defendants' counsel yesterday, 

25    and if you would you look to Exhibit 21 in that.  I ask you, 
                                                              1800

 1    Mr. Bounpane, to turn to Table 8 of Exhibit 21.  I think it 

 2    is the last page.

 3         A.    Yes.

 4         Q.    That is the table you referred us to yesterday, 

 5    isn't that correct?

 6         A.    Yes, it is.

 7         Q.    Looking under the column for the PES, the 

 8    differential between blacks and nonblacks is 3.13, correct?

 9         A.    Subtracting 1.7 from 4.8, correct.  Am I doing it 

10    right?

11         Q.    4.82 minus 1.69, are those the figures?

12         A.    Yes.

13         Q.    That comes out to 3.13, correct?

14         A.    Is it okay if I write, your Honor?  

15               THE COURT:  Sure.  Can't do it in your head?  

16               THE WITNESS:  You know, statisticians cannot add.  

17    And this one in particular.  

18               THE COURT:  They sure can talk, though.

19         A.    Okay.  That is 3.13.

20         Q.    The next column under January 1992 shows the 

21    differential undercount as 2.85 percentage points, correct?  

22    That is 4.2 minus 1.33.

23         A.    2.85 is correct.

24         Q.    So the reduction in the differential undercount 

25    between the July 1991 and the January 1992 figures is about 
                                                              1801

 1    .28 percentage points, isn't that right?

 2         A.    That's correct. 

 3         Q.    You recall your deposition on March 25, 1992, 

 4    don't you?

 5         A.    I remember taking the deposition, yes.

 6         Q.    You remember that Moulton and I were at that 

 7    deposition?

 8         A.    Yes.

 9         Q.    You testified that the differential undercount as 

10    measured by the PES, after accounting for the computer edit 

11    problem, was left substantially intact, don't you?

12         A.    I don't recall that precisely.  If you want to 

13    reference me to a page, I will be happy to read it. 

14         Q.    Mr. Bounpane, do you remember these questions and 

15    your answers?

16              "Q.    Given this reduction of .28, would you say 

17    that the differential undercount has been left substantially 

18    intact after accounting for the computer error?

19              "A.    As measured by the PES, whatever flaws the 

20    PES has, I think that is correct.  It is still basically the 

21    same differential."  

22               Do you remember those questions and answers?

23         A.    If that is what the deposition transcript says, 

24    yes, that is what I said.

25         Q.    I will show it to you.
                                                              1802

 1         A.    That's okay.  I trust you.

 2         Q.    You do remember?

 3         A.    I didn't remember precisely.

 4         Q.    Does that refresh your recollection?

 5         A.    Okay.

 6         Q.    Yes?

 7         A.    Yes.  

 8               THE COURT:  His question is, now that you hear 

 9    it, do you remember saying it?

10         A.    Yes.

11         Q.    You didn't say that yesterday, did you?

12         A.    I answered the question that was asked yesterday.

13         Q.    You weren't asked whether the differential 

14    undercount was left substantially intact, is that correct?

15         A.    That's right.

16         Q.    Do you remember testifying yesterday about the 

17    outlier cluster review done by the census bureau?

18         A.    Yes, I could.

19         Q.    You testified at your deposition, didn't you, 

20    that that review involved taking the blocks that impacted 

21    the most on the dual-system estimator and looking to see if 

22    you can identify the bias in those blocks and then removing 

23    that bias, isn't that right?

24         A.    That was one of the goals of that study, yes.

25         Q.    That study is not generalizable to any block 
                                                              1803

 1    cluster outside the 104 that were studied, isn't that 

 2    correct?

 3         A.    No, it is not.

 4         Q.    Any correction to the counts that result from 

 5    that study would only affect the states that have one or 

 6    more of those blocks, isn't that correct?

 7         A.    Yes, that's correct. 

 8         Q.    Mr. Bounpane, you testified that the outlier 

 9    cluster review study was designed to find out if the blocks 

10    that are the subject of the study contributed in some way to 

11    the estimated .7 percentage points of bias in the undercount 

12    rate as measured by the PES, is that right?

13         A.    Yes.  That was one of the goals of the outlier 

14    cluster review.

15         Q.    It is a fact, isn't it, that that estimate of .7 

16    percent of bias was known to the Undercount Steering 

17    Committee when it made its recommendation?

18         A.    Yes.

19         Q.    In fact, the Secretary notes that estimate of 

20    bias in his decision document, doesn't he?

21         A.    I would think so, yes. 

22         Q.    So the study doesn't locate a new source of bias 

23    above the .7 percent estimate, does it?

24         A.    No.

25         Q.    You didn't say that yesterday, did you?
                                                              1804

 1         A.    I don't think so.

 2         Q.    Let me refer you to Defendants' Exhibit 31 in 

 3    your binder.

 4         A.    Okay.

 5         Q.    You testified about that exhibit yesterday, 

 6    correct?

 7         A.    Yes, I did.

 8         Q.    You referred us to attachment 2 of that exhibit, 

 9    correct?

10         A.    Yes, I did.

11         Q.    Would you turn to that now, please.

12         A.    Okay.

13         Q.    Attachment 2 is a list of the undercount rates 

14    for states after taking into account the computer edit 

15    problem, the select cluster review, and a revised 

16    post-stratification, isn't that correct?

17         A.    Yes, it is.

18         Q.    You directed us to the fourth column, the last 

19    line, correct?

20         A.    Yes, the column labeled U.

21         Q.    That gives a figure for the revised net 

22    undercount nationally, isn't that is correct?

23         A.    Yes, it does.

24         Q.    Mr. Bounpane, would you look at all the figures 

25    in column 4 above the one that you referred us to yesterday.
                                                              1805

 1         A.    Okay.  I see those figures.

 2         Q.    Those figures in that column show differential 

 3    undercount across the states, don't they?

 4         A.    They show undercount rates by state.

 5         Q.    Those undercount rates are differential, isn't 

 6    that correct?

 7         A.    They differ by state, yes.

 8         Q.    It shows higher undercount to some states than to 

 9    others?

10         A.    That is correct.

11         Q.    You didn't mention that yesterday, did you?

12         A.    No.

13         Q.    Isn't it also true that in that table, New Mexico 

14    has the highest estimated undercount rate of any of the 50 

15    states?

16         A.    I am just scanning here.  I don't see any others 

17    over 3, so the answer to that would be yes.

18         Q.    That is the same as the July 1990 estimates, 

19    isn't it?

20         A.    New Mexico had the highest set of estimates, yes.

21         Q.    You didn't tell us about that yesterday, did you, 

22    Mr. Bounpane?

23         A.    No.

24         Q.    Can you go back two pages in Exhibit 31, please, 

25    to page 1 of attachment 2.
                                                              1806

 1         A.    Okay.  

 2               THE COURT:  Go back to page 1 of attachment 2?  

 3               MR. COHEN:  Of Exhibit 31, your Honor.  

 4               THE COURT:  That is where we are.  You said 

 5    attachment 2.  

 6               MR. COHEN:  I'm sorry.  Page 1 of attachment 1.  

 7               THE WITNESS:  That's bad.  I understood him.

 8         Q.    That chart shows the net undercounts for 

 9    different demographic groups based on the same estimates as 

10    the table we just looked at, isn't that correct?

11         A.    Yes, it does.

12         Q.    Mr. Bounpane, would you please look at the part 

13    of that table that shows the estimated undercount rate for 

14    non-Hispanic, white, and other owners in urban areas of 

15    250,000 or more in the northeast.

16         A.    Okay.  I have that number.

17         Q.    What is that figure?

18         A.    That is an estimated 2.13 percent overcount.

19         Q.    The negative sign indicates that it is an 

20    overcount, isn't that correct?

21         A.    Yes, it does.

22         Q.    Would you please look at the estimated net 

23    undercount for black nonowners in urban areas of 250,000 or 

24    more in the Northeast.

25         A.    Black nonowners, 250,000, in the Northeast?
                                                              1807

 1         Q.    Correct?

 2         A.    Yes.

 3         Q.    What does that figure show?

 4         A.    That shows an estimated 6.72 percent undercount.

 5         Q.    Black nonowners.  Still on the first page.  

 6         A.    I'm sorry.  It shows an estimated 8.37 percent 

 7    undercount.

 8         Q.    If you take those two categories, the 

 9    differential between them is about 10.5 points, correct?

10         A.    Assuming that arithmetic is correct, it looks 

11    about right, yes.

12         Q.    It is a pretty substantial differential, wouldn't 

13    you agree?

14         A.    Yes.

15         Q.    Would you look at the next page, please.  That 

16    shows an estimated 6.72 percent net undercount for nonblack 

17    Hispanic nonowners in the Northeast?  Isn't that correct?

18         A.    Wait a minute.  I'll get it right this time.  

19    Nonblack Hispanic nonowners?

20         Q.    Nonowners.

21         A.    Nonblack Hispanic nonownser.

22         Q.    In the the Northeast?

23         A.    In the Northeast, urban areas over 250,000?

24         Q.    Correct.

25         A.    That is an estimated undercount of 6.72 percent.
                                                              1808

 1         Q.    Isn't it true that this table shows estimates 

 2    consistently reflecting a differential undercount between 

 3    minorities and nonminorities?

 4         A.    Yes.

 5         Q.    You didn't show us that chart yesterday, did you, 

 6    Mr. Bounpane?

 7         A.    No.

 8         Q.    The CAPE committee is continuing to deliberate on 

 9    whether to adjust intercensal estimates, correct?

10         A.    Yes.

11         Q.    Intercensal statements are not used for 

12    reapportionment or redistricting, is that correct?

13         A.    That's correct.

14         Q.    You told us yesterday that CAPE is expected to 

15    vote on that issue in July, is that correct?

16         A.    Approximately, yes.

17         Q.    So CAPE hasn't determined whether to correct the 

18    intercensal estimates, is that correct?

19         A.    That is correct.

20         Q.    And it hasn't decided against correcting the 

21    intercensal estimates?

22         A.    That's correct.

23         Q.    You don't know now how that vote will come out, 

24    do you?

25         A.    I don't. 
                                                              1809

 1         Q.    You would say, wouldn't you, that the CAPE has 

 2    reviewed a number of research results over the last year in 

 3    considering this issue?

 4         A.    I didn't understand that question.  Would you say 

 5    it again?

 6         Q.    You would say, wouldn't you, that the CAPE has 

 7    reviewed a number of research results over the last year in 

 8    considering the issue?

 9         A.    Oh, yes.

10         Q.    And if it decides to adjust the intercensal 

11    estimates, it will do so on the basis of the PES data 

12    collected in the 1990 PES, is that correct?

13         A.    That's correct.

14         Q.    Nothing so far has convinced the CAPE to toss 

15    that data out, isn't that correct?

16         A.    I'm sorry?

17         Q.    Nothing so far has convinced CAPE to toss that 

18    data out, isn't that correct?

19         A.    That's correct.

20         Q.    Mr. Bounpane, you know Professor Leo Breiman, 

21    isn't that correct?

22         A.    Just recently met him.

23         Q.    You know that he was retained by the justice 

24    department as an expert in this litigation in about April of 

25    1990, correct?
                                                              1810

 1         A.    That I didn't know.

 2         Q.    You know that he has been retained by the justice 

 3    department?

 4         A.    Yes.

 5         Q.    Mr. Bounpane, you have read reports produced for 

 6    this litigation by Professor Breiman on nonsampling error in 

 7    the PES, isn't that correct?

 8         A.    I read a draft report, yes.

 9         Q.    In fact, you first reviewed the draft report last 

10    winter, isn't that correct?

11         A.    Approximately.  I think earlier this year.  I 

12    don't know if that is winter or not.

13         Q.    Other CAPE members have reviewed that draft 

14    report, correct?  

15               MR. SUBAR:  Objection.  Speculation.  

16               THE COURT:  If you know.

17         A.    I think some other CAPE members have reviewed 

18    that report.

19         Q.    You also know Professor David Freedman, isn't 

20    that correct?

21         A.    Yes, I do.

22         Q.    You also know that he has been retained by the 

23    justice department to be an expert in this litigation?

24         A.    Yes.

25         Q.    You know that he has been retained since May of 
                                                              1811

 1    1989, correct?

 2         A.    I don't know the date.

 3         Q.    Do you know that he was retained before the 

 4    hearing that was scheduled to go forward on the preliminary 

 5    injunction motion in this litigation, correct?

 6         A.    Sorry.  I don't know the date of that hearing 

 7    either. 

 8         Q.    But you do know it was before the date of that 

 9    hearing, correct?

10         A.    I do not know when he was retained.

11         Q.    You have read drafts of Mr. Freedman's reports 

12    too, haven't you?

13         A.    Yes, I have.

14         Q.    You read them last winter for the first time, 

15    correct?

16         A.    Early this year, that's all I can say.

17         Q.    Other CAPE members have read those reports, 

18    correct?

19         A.    I think some other CAPE members have read those 

20    reports, yes.

21         Q.    You also know Professor Ken Wachter, don't you?

22         A.    Yes.

23         Q.    You know that he is retained by the justice 

24    department to be an expert in this litigation?

25         A.    Yes.
                                                              1812

 1         Q.    And that he was retained last August?

 2         A.    I do not know when he was retained.

 3         Q.    You have had available to you, haven't you, 

 4    Professor Wachter's report that he submitted to the 

 5    Secretary in July of 1991?

 6         A.    Yes.

 7         Q.    You also have read drafts of the reports he has 

 8    prepared for this litigation, isn't that correct?

 9         A.    No.

10         Q.    Those reports have been available to other CAPE 

11    members?  

12               MR. SUBAR:  Objection.  That calls for 

13    speculation.

14         A.    I don't know that.  

15               THE COURT:  He said he didn't know.

16         A.    I don't know.

17         Q.    Yesterday, Mr. Bounpane, you testified, didn't 

18    you, that the job of the census bureau in regard to the 

19    adjustment decision is to make a technical assessment of the 

20    accuracy of the counts and for others to make policy 

21    decisions?  Isn't that correct?

22         A.    Yes.

23         Q.    Has that always been your view?

24         A.    Yes.

25         Q.    In 1987 you were a member of the Undercount 
                                                              1813

 1    Steering Committee of the bureau, correct?

 2         A.    Yes.

 3         Q.    There came a time in 1987 when the Undercount 

 4    Steering Committee was called upon to recommend to the 

 5    director of the bureau whether to plan for an adjustment of 

 6    the 1990 census, isn't that true?

 7         A.    Yes.

 8         Q.    The census bureau made a decision to proceed with 

 9    the dual strategy in 1987, correct?

10         A.    Yes.

11         Q.    That dual strategy meant that the bureau should 

12    continue to take the census and the PES, and at the 

13    conclusion of the census look at the results of the PES, and 

14    make a decision then whether or not to adjust the census 

15    based on the accuracy of the counts, isn't that correct?

16         A.    Yes.

17         Q.    The commerce department overruled the census 

18    bureau's decision, didn't it?

19         A.    Yes.

20         Q.    As a member of the Undercount Steering Committee 

21    in 1987, you recommended against an adjustment, isn't that 

22    correct?

23         A.    Not exactly.  I recommended that if I had to make 

24    a decision on the information I had in 1987, I could not say 

25    that we would adjust in 1990.
                                                              1814

 1         Q.    Mr. Bounpane, I left another binder next to you.  

 2    Perhaps you can look through it for Plaintiffs' Exhibit 730.  

 3               MR. COHEN:  Your clerk has your binder, your 

 4    Honor.

 5         Q.    Mr. Bounpane, do you recognize the document that 

 6    has been marked as Plaintiffs' Exhibit 730?

 7         A.    It looks like a printout of an electronic mail 

 8    message that I had sent.

 9         Q.    You sent that electronic mail message to C. 

10    Kincannon, is that correct?

11         A.    Yes.

12         Q.    Mr. Kincannon is the deputy director of the 

13    bureau, correct?

14         A.    That's correct.

15         Q.    He was on April 30, 1987, correct?

16         A.    Yes.

17         Q.    That is when you sent that memo, isn't it, 

18    correct?

19         A.    It is dated April 30, 1987, yes.  

20               MR. COHEN:  Your Honor, I move that Plaintiffs' 

21    Exhibit 730 be admitted into evidence.  

22               MR. SUBAR:  No objection.  

23               THE COURT:  Are 730 is admitted. 

24               (Plaintiffs' Exhibit 730 for identification was 

25    received in evidence).
                                                              1815

 1         Q.    Could I ask you, Mr. Bounpane, to look down at 

 2    the paragraph marked 3 in Plaintiffs' Exhibit 730, and read 

 3    what it says.

 4         A.    Yes.  Do you want me to read it out loud.

 5         Q.    Yes.

 6         A.    "For various reasons (timing, geographic level 

 7    possible, and remaining PES problems) it looks like we 

 8    cannot adjust census by 12-31."

 9         Q.    What does the next paragraph say?

10         A.    "If we can't adjust by 12-31, I do not think we 

11    should adjust by 4-1-91.  To have redistricting and 

12    reapportionment counts be different would be bad."

13         Q.    What does the next paragraph say?

14         A.    I did says, "I recommend our May decision be that 

15    we do not adjust the census."

16         Q.    So it is true, isn't it, that you did not come to 

17    a conclusion in 1987 that it was technically infeasible to 

18    correct the census in 1990?

19         A.    It is correct that I did not come to a conclusion 

20    that it was technically infeasible.

21         Q.    According to this document, your reason for 

22    recommending against an adjustment in 1987 was because you 

23    didn't want a corrected set of redistricting counts if there 

24    were no corrected set of reapportionment counts, isn't that 

25    correct?
                                                              1816

 1         A.    No, I don't think so.

 2         Q.    Would you read paragraph 4 again, Mr. Bounpane.

 3         A.    Yes.  "If we can't adjust by 12-31, I do not 

 4    think we should adjust by 4-1-91.  To have redistricting and 

 5    reapportionment counts be different would be bad."

 6         Q.    That is not a technical decision, is it?

 7         A.    No.

 8         Q.    In 1987 you assumed that it would be the director 

 9    of the census bureau and not the commerce department that 

10    would decide whether to proceed with the plans for an 

11    adjustment, isn't that correct?

12         A.    I don't know that.

13         Q.    Mr. Bounpane, do you remember giving a deposition 

14    in 1989 on this case?

15         A.    I did give a deposition in 1989.

16         Q.    Do you remember the following questions and 

17    answers?  

18               MR. SUBAR:  Can we have a page?

19         Q.    Page 111.

20              "Q.    Did you expect that Mr. Keane would make 

21    the decision whether the 1990 census would be adjusted?

22              "A.    Never really thought about it very much 

23    except to say that the director had made it in 1980.  I 

24    guess I sort of assumed that would have also been the case 

25    in 1990."  I'm sorry.  I added two words there.  "That would 
                                                              1817

 1    have also been in 1990."  

 2               Do you remember giving that testimony?

 3         A.    I don't recall that specifically, no.

 4         Q.    So it doesn't refresh your recollection?  

 5               MR. SUBAR:  Your Honor, perhaps, if the witness 

 6    doesn't recall, he can be shown a copy of the deposition 

 7    transcript.  

 8               MR. COHEN:  We glad to.  

 9               MR. SUBAR:  If Mr. Cohen could please make sure 

10    to make the whole page available to Mr. Bounpane so that he 

11    can read the whole answer, which Mr. Cohen did not read.

12         Q.    Have you had a chance to read that testimony, Mr. 

13    Bounpane?

14         A.    I looked at a couple of pages.  Obviously, if 

15    this is what the transcript said, that is what I said.  But 

16    it was within the context of a set of questions discussing 

17    was it legal for the director of the census bureau to make 

18    the decision.

19         Q.    Quite apart from whether it was legal, you 

20    assumed that the director would make the decision in 1990, 

21    didn't you?

22         A.    I said I hadn't thought about it and I made that 

23    assumption.  That is what I said in 1989.

24         Q.    Your expectation was based on the fact that the 

25    director in 1980 had made the adjustment decision, isn't 
                                                              1818

 1    that correct?

 2         A.    I don't know what my expectation was.  I can't 

 3    recall back then.  I did notice when reading this, I noticed 

 4    a distinction between how the process was handled in 1980 

 5    relative to 1990.

 6         Q.    But, in fact, the 1987 decision that there would 

 7    be no adjustment of the 1990 census was announced by Under 

 8    Secretary Robert Ortner, isn't that correct?

 9         A.    That's correct.

10         Q.    By 1987 you had worked for the census bureau for 

11    about 22 years, correct?

12         A.    About right, yes.

13         Q.    In that entire time it is true, isn't it, that 

14    except for Dr. Ortner's announcement, you are unable to 

15    recall any other major policy decision with respect to the 

16    decennial census having been announced by an individual from 

17    the Department of Commerce, isn't that true?

18         A.    Having been announced by the Department of 

19    Commerce?  I can't recall any others, you're right.

20         Q.    It is true, isn't it, that after the commerce 

21    department's decision against adjustment, the bureau stopped 

22    planning for an adjustment until the stipulation and order 

23    of 1989?

24         A.    Stopped planning for an adjustment.  It did not 

25    stop planning for the PES.
                                                              1819

 1         Q.    The bureau stopped developing standards for an 

 2    adjustment until the stipulation was entered, isn't that 

 3    correct?

 4         A.    That's correct.

 5         Q.    The only work that the bureau was conducting with 

 6    respect to the PES between October 30, 1987, and July 17, 

 7    1989, was related to its use as an evaluation tool, isn't 

 8    that correct?

 9         A.    That's correct.

10         Q.    I think you testified yesterday that minorities 

11    are disproportionately missed in the census, isn't that 

12    correct?

13         A.    That is my expectation, yes.

14         Q.    It is also your belief, isn't it?

15         A.    Yes.

16         Q.    In fact, you don't have any doubt that there was 

17    differential undercount in the 1990 census, isn't that 

18    correct?

19         A.    That's what I think.

20         Q.    And that the differential undercount was 

21    differential by race, correct?

22         A.    Yes, that is what I think.

23         Q.    The census bureau has historical evidence of a 

24    differential undercount of minorities, isn't that correct?

25         A.    Yes.
                                                              1820

 1         Q.    It had that evidence well before 1990, isn't that 

 2    correct?

 3         A.    Yes.

 4         Q.    In fact, the census bureau made special efforts 

 5    for the 1990 census to encourage minorities to return census 

 6    forms, isn't that true?

 7         A.    Yes.

 8         Q.    That is because the bureau had an expectation 

 9    based on its past experience that minorities would be 

10    differentially undercounted, correct?

11         A.    Yes.

12         Q.    Those efforts included efforts to arrange for the 

13    services of minority advertising agencies to appeal more 

14    directly to minority audiences, correct?

15         A.    Yes.

16         Q.    It also included as part of the census education 

17    program the use of census community awareness specialists, 

18    especially in the areas with high proportions of minorities, 

19    isn't that correct?

20         A.    Yes.

21         Q.    The bureau also designed a new Headstart effort 

22    for 1990 to reach low-income populations, correct?

23         A.    In some sense, yes.

24         Q.    The bureau held meetings with leaders of black 

25    churches to obtain their cooperation for 1990, correct?
                                                              1821

 1         A.    Yes.

 2         Q.    The bureau established an undercount behavioral 

 3    research group to learn more about the reasons for decennial 

 4    census undercount among minorities, correct?

 5         A.    Yes.

 6         Q.    All of these efforts were made because minorities 

 7    had been differentially undercounted in past censuses, 

 8    correct?

 9         A.    Yes.

10         Q.    You expected that these programs would ameliorate 

11    the differential undercount, didn't you?

12         A.    We certainly had that as our goal.

13         Q.    That is what your expectation was, too, wasn't 

14    it, Mr. Bounpane?

15         A.    Yes.

16         Q.    The bureau added other new programs for 1990, 

17    correct?  You told us about some of them yet?

18         A.    Yes.

19         Q.    They included the 800 line, correct?

20         A.    Yes.

21         Q.    Parolee/probationer check program, correct?

22         A.    Yes.

23         Q.    Indigenous hiring, correct?  

24         A.    That wasn't new per se, but yes.

25         Q.    These programs cost millions of dollars, correct?
                                                              1822

 1         A.    Yes.

 2         Q.    You urged that they be implemented as part of 

 3    this census effort, didn't you?

 4         A.    Yes.

 5         Q.    You expected that those problems would ameliorate 

 6    the differential undercount, too, didn't you?

 7         A.    Yes.

 8         Q.    The bureau had an extensive promotional effort 

 9    for the 1990 census, isn't that correct?

10         A.    Yes.

11         Q.    The purpose of that promotion campaign was to 

12    achieve a high mail return rate for 1990, isn't that 

13    correct?

14         A.    That was certainly one of its goals.

15         Q.    In fact, wasn't that the most important goal of 

16    the campaign?

17         A.    I'm not sure about that.  I think probably 

18    equally important was the fact to try to encourage people to 

19    make themselves known to the census.

20         Q.    In 1980 the bureau achieved a 83 percent mail 

21    return rate for persons in occupied housing, isn't that 

22    correct?

23         A.    Yes, that is correct.

24         Q.    It is also true, isn't it, that the bureau 

25    achieved a mail response rate of 75 percent in 1980?
                                                              1823

 1         A.    Approximately, yes.

 2         Q.    For 1990 the bureau anticipated a mail response 

 3    rate of about 70 percent, right?

 4         A.    That's correct.

 5         Q.    That is the mail response rate it budgeted for, 

 6    isn't that correct?

 7         A.    Yes.

 8         Q.    You were a participant in the deliberations which 

 9    led to the budgeting for 70 percent mail return rate, 

10    correct?

11         A.    Yes.

12         Q.    The mail response rate for the nation as a whole 

13    was only 63 percent, is that correct?

14         A.    Yes, about that.

15         Q.    You are familiar with type 1 district offices, 

16    aren't you?

17         A.    Yes.

18         Q.    A type 1 district office is generally located in 

19    a central city, correct?

20         A.    Yes.

21         Q.    Isn't it true that the mail response rate for 

22    type 1 district offices was only 60 percent in 1990?

23         A.    I don't know the precise number.  It was lower 

24    than the national average.

25         Q.    Isn't it also true that the mail response rate 
                                                              1824

 1    for long forms in type 1 district offices was only 53 

 2    percent?

 3         A.    Again, I don't know the precise number, but it 

 4    was lower than the national average.  Both of those also 

 5    occurred in 1980.

 6         Q.    It is also true, isn't it, that when the mail 

 7    response rate is lower, the undercount rate will be higher?

 8         A.    People have said they have done correlations 

 9    between mail response rate and undercount and found those 

10    two related.

11         Q.    You agree with that conclusion, don't you?

12         A.    I think that is a reasonable thing to happen, 

13    yes.

14         Q.    If a census form in a mail out-mail back area is 

15    not returned, the bureau tends to enumerate the household 

16    with a personal visit in a nonresponse follow-up made to the 

17    census, is that correct?

18         A.    That's correct.

19         Q.    It is true, isn't it, that as to the extent that 

20    enumerations are obtained through interviews rather than 

21    self-response, you get more error in the responses?

22         A.    I would say generally, yes.  

23               (Continued on next page) 

24    

25    
                                                              1825

 1         Q.    You agree, don't you, that nonresponse follow-up 

 2    is harder to conduct in central cities than elsewhere?

 3         A.    I would think generally yes.

 4         Q.    In 1990, the Census Bureau experienced 

 5    difficulties with nonresponse follow-up in minority 

 6    neighborhoods in central cities, correct?

 7         A.    I'm sure that happened in some central cities, 

 8    yes.

 9         Q.    It's also true, isn't it, that it is harder to 

10    hire and maintain a staff of enumerators in Type 1 district 

11    offices than in Type 2 or Type 3 district offices?

12         A.    Generally yes, although it was less of a problem 

13    in 1990 than it was in 1980.

14         Q.    It was generally true, correct?
                                                

15         A.    Generally true, yes.

16         Q.    Isn't it true that after three visits and three 

17    phone calls to a household, the Bureau enumeration rules 

18    provide for enumeration by last resort procedures?

19         A.    I'm not sure of the precise numbers, but after a 

20    specified number of attempts, yes, that's what happens.

21         Q.    And last resort, the enumerator can get 

22    information from a neighbor or superintendent, isn't that 

23    correct?

24         A.    Yes, that's correct.

25         Q.    That information is less reliable than 
                                                              1826

 1    information from a member of the household, isn't that 

 2    correct?

 3         A.    I would think in general it would be in some 

 4    regards, yes.

 5         Q.    That's the Bureau's knowledge of reliability of 

 6    last resort information, isn't that correct?

 7         A.    I would think so, yes.

 8         Q.    If an enumerator cannot obtain information from 

 9    last resort procedures, he or she may use closeout 

10    procedures, correct?

11         A.    Yes, that's correct.

12         Q.    And in closeout, the enumerator is permitted to 

13    obtain fewer than all of the characteristics of members of 

14    the household, correct?

15         A.    Yes.

16         Q.    Closeout provides information that is even less 

17    reliable than last resort, isn't it?

18         A.    Generally, yes.

19         Q.    If the closeout procedures do not yield an 

20    enumeration, the enumerator is permitted to submit a nondata 

21    define enumeration, isn't that correct?

22         A.    I'm not sure the enumerator does that.  If they 

23    cannot get any information at all as to the number of people 

24    there, even as basic a thing as that, they can turn in the 

25    questionnaire, in effect, blank and that is really a nondata 
                                                              1827

 1    defined person.

 2         Q.    So it is a person who is enumerated without any 

 3    characteristics, correct?

 4         A.    Yes, that's correct.

 5         Q.    And nondata define enumerations are even less 

 6    reliable than closeout enumerations?

 7         A.    Yes.

 8         Q.    Isn't it true in 1990, a greater proportion of 

 9    minorities in the census than nonminorities were enumerated 

10    through last resort closeout than nondata define procedures?

11         A.    I don't know the precise numbers.  I would expect 

12    that might be true.

13         Q.    Isn't it also true that a greater proportion of 

14    the enumerations in Type 1 district offices were through 

15    last resort closeout and nondata define procedures than in 

16    Type 2 or Type 3 offices?

17         A.    Again, I don't know the precise numbers.  I would 

18    expect that it is probably correct.

19         Q.    Type 2 offices are generally suburban offices, 

20    correct?

21         A.    Type 2 is generally suburban, yes.

22         Q.    Type 3 are generally rural offices, isn't that 

23    correct?

24         A.    Generally, yes.

25         Q.    It's true that nonresponse follow-up was 
                                                              1828

 1    scheduled to end in early June 1990, is that correct?

 2         A.    I don't know the precise date, but about that 

 3    sounds right.

 4         Q.    It's also true that some district offices didn't 

 5    complete nonresponse follow-ups until July 1990?

 6         A.    I would expect that was the case, yes.

 7         Q.    Do you know that that is the case?

 8         A.    Yes, I'm sure some were still being done in July.

 9         Q.    In fact, the last office close nonresponse 

10    follow-up is on July 28, correct?

11         A.    I don't know the precise date.  

12               MR. COHEN:  Your Honor, are you having trouble 

13    hearing me as well?  

14               THE COURT:  Yes.  

15         Q.    The end of July was about seven weeks after the 

16    scheduled end of nonresponse follow-up, isn't that correct?

17         A.    Approximately, yes.

18         Q.    It's true, isn't it, that half of all the Type 1 

19    offices didn't complete nonresponse follow-up until July?

20         A.    Again, I do not know the precise number.  That 

21    sounds like a reasonable thing.  That's what happened in 

22    1980 as well.

23         Q.    But 75 percent of all Type 2 offices completed 

24    nonresponse follow-up before July, isn't that correct?

25         A.    I don't know the precise number.  It seems like a 
                                                              1829

 1    reasonable percentage.

 2         Q.    And 85 percent of all Type 3 offices completed 

 3    nonresponse follow-up in June, isn't that correct?

 4         A.    Again, I do not know the precise number.

 5         Q.    You don't consider the differential undercount to 

 6    be one of the major flaws of the 1990 decennial census, do 

 7    you?

 8         A.    Well, I think I do, yes.

 9         Q.    Mr. Bounpane, do you remember the following 

10    questions and answers in your deposition of March 25, 1992.  

11               MR. SUBAR:  Give me the page, please?  

12               MR. COHEN:  Page 30.

13         Q.    "Q.    What, in your view, are the major flaws, 

14    if any, in the census-taking procedures as used in the 1990 

15    decennial census?

16               "A.    I would at least like to preface it by 

17    saying that I think the census was well done.  No census can 

18    be perfect so obviously there might have been some 

19    difficulties, and the ones I would have been concerned about 

20    would have been when someone asked for a questionnaire on 

21    the phone it was not immediately mailed, perhaps not enough 

22    flexibility was allowed to alter the census procedure area 

23    by area.  They are some and more perhaps had to be better.  

24    If I had to list the major ones, those are the ones I would 

25    have been concerned about." 
                                                              1830

 1               Do you remember that answer?

 2         A.    Yes.

 3         Q.    You didn't mention differential undercount, did 

 4    you, Mr. Bounpane?

 5         A.    No.  The question was what census processes I 

 6    thought might have been flawed and that's what I was trying 

 7    to answer.

 8         Q.    So you do agree that differential undercount was 

 9    one of the major flaws of the 1990 census?

10         A.    As an outcome, yes.

11         Q.    You were one of the two in the dissent on the 

12    USC, correct?

13         A.    Yes, I was.

14         Q.    And the position of the minority, the dissenters, 

15    was that it did not have enough information to conclude that 

16    the adjusted counts are more accurate than the unadjusted 

17    counts, isn't that correct?

18         A.    I think that's a fair summary, yes.

19         Q.    You did conclude, didn't you, that for some of 

20    the evaluation post-strata, the adjusted counts are more 

21    accurate than the census?

22         A.    Yes.  We both felt that there were certain 

23    post-strata where that was true.

24         Q.    In fact, it's true, isn't it, that you believed 

25    that using the PES data to adjust for post-strata with large 
                                                              1831

 1    measured undercounts would be acceptable?

 2         A.    No, I don't think we concluded that.

 3         Q.    Mr. Bounpane, may I ask you to turn to the binder 

 4    that I left on your stand this morning and turn to 

 5    Plaintiff's Exhibit 54. 

 6               Would you look at the first page of the 

 7    undercount steering committee's recommendation, please?

 8         A.    Okay.

 9               Do you want me to read that?

10         Q.    In the third paragraph, the next to the last 

11    sentence, would you take a look at that, please.

12               (Pause)

13         A.    Yes, I've read that.

14               Do you want me to read that?

15         Q.    Doesn't that say that the minority believed that 

16    an alternative adjustment using the PES data to adjust for 

17    post-strata with large measured undercounts might be 

18    acceptable to them?

19         A.    Yes, that's what it says, might be acceptable, 

20    because it's not simply because of the fact it is large, it 

21    is large with small measured error associated with it, and 

22    in that instance then you would say the information is 

23    reliable enough to use because even if the error doubled or 

24    something like that you might still feel you have a very 

25    reliable estimate.
                                                              1832

 1         Q.    The post-strata with the large measured 

 2    undercounts were generally minority post-strata, isn't that 

 3    correct?

 4         A.    Generally, yes.

 5         Q.    And you also supported using the adjusted counts 

 6    for intercensal estimates, isn't that correct?

 7         A.    We thought that was a reasonable consideration, 

 8    yes.

 9         Q.    Robert Tortora was a member of the USC, correct?

10         A.    Yes.

11         Q.    He is the chief of the statistical research 

12    division, correct?

13         A.    Yes.

14         Q.    That is the position Kirk Wolter held, correct?

15         A.    Correct.

16         Q.    Mr. Tortora has a Ph.D. in probabilities and 

17    statistics, correct?

18         A.    He has a Ph.D.  I'm not sure in what field.

19         Q.    Mr. Robert Groves was a member of the USC, 

20    correct?

21         A.    Yes.

22         Q.    He was the associate director for statistical 

23    design, methodology and standards, correct?

24         A.    Correct.

25         Q.    That is the position Barbara Bailar held, 
                                                              1833

 1    correct?

 2         A.    Yes.

 3         Q.    Mr. Groves has a Ph.D., correct?

 4         A.    Yes.

 5         Q.    John Thompson was a member of the USC, correct?

 6         A.    Yes.

 7         Q.    Mr. Thompson has a graduate degree in 

 8    mathematics, correct?

 9         A.    He has a graduate degree.  I'm not sure of the 

10    topic, again.

11         Q.    William Butz was a member of the undercount 

12    steering committee, correct?

13         A.    Yes.

14         Q.    Mr. Butz is associate director for demographic 

15    programs, isn't that correct?

16         A.    Yes.

17         Q.    That's a different division than the one you and 

18    Mr. Jones worked for, isn't that correct?

19         A.    It's a different line of authority within the 

20    Census Bureau, yes.

21         Q.    Paula Schneider was a member of the USC, correct?

22         A.    Correct.

23         Q.    She is the chief of the population division, 

24    correct?

25         A.    Yes.
                                                              1834

 1         Q.    That's not part of the division in which you 

 2    work, is it?

 3         A.    No.

 4         Q.    Ms. Schneider was the chair of the USC, correct?

 5         A.    Yes.

 6         Q.    Susan Miskura was a member of the USC, correct?

 7         A.    Yes.

 8         Q.    She is the chief of the year 2000 research and 

 9    development staff, correct?

10         A.    She is currently, yes.

11         Q.    Mr. Jones doesn't have any graduate degrees, does 

12    he?

13         A.    Not that I know of.

14         Q.    Mr. Bounpane, you don't have any graduate 

15    degrees, isn't that true?

16         A.    That's correct.

17         Q.    In the last ten years, you haven't published any 

18    articles on dual system estimation, have you?

19         A.    No.

20         Q.    Nor have you published any articles in the last 

21    decade on census adjustment, isn't that correct?

22         A.    I don't know if it's the last decade or so.  I 

23    have done some in the past, but not very recently, correct.

24         Q.    And you have never published an article in 

25    referee journals, have you?
                                                              1835

 1         A.    Early in my career I think I did, but not 

 2    recently.

 3         Q.    Not in the last 20 years?

 4         A.    I'm not sure about the exact number of years.

 5         Q.    And you are not a Fellow of the American 

 6    Statistical Association, are you?

 7         A.    No, I'm not.

 8         Q.    The USC report was submitted on or about June 21, 

 9    correct?

10         A.    Yes, that's correct.

11         Q.    And thereafter, on or about June 27, an addendum 

12    to the report was submitted?

13         A.    Yes, it was.

14         Q.    And on July 8 you attended a meeting with 

15    Secretary Mosbacher, correct?

16         A.    Correct.

17         Q.    That was a meeting that included several persons 

18    from the Census Bureau to discuss the issue of the accuracy 

19    of the census, correct?

20         A.    Yes.

21         Q.    And that was the only such meeting with the 

22    Secretary you are aware of between June 21 and July 15, 

23    1991, isn't that correct?

24         A.    Between Census Bureau staff and the Secretary.  

25    That's the only one I'm aware of, yes.
                                                              1836

 1         Q.    That meeting lasted about two hours, correct?

 2         A.    Approximately.

 3         Q.    The Secretary left the meeting early, isn't that 

 4    correct?

 5         A.    About 15 minutes early, yes.

 6         Q.    You didn't have a sufficient opportunity to 

 7    express your views at that meeting, did you?

 8         A.    No, I don't think I did.

 9         Q.    Mr. Bounpane, I ask you to look at Plaintiff's 

10    Exhibit 673.  You testified about that yesterday.

11         A.    Okay.

12         A.    I have it.

13         Q.    You recall that memo, don't you?

14         A.    Yes.

15         Q.    That is the memo you wrote to the Secretary on 

16    July 10, isn't that correct?

17         A.    That's correct.

18         Q.    And you wrote that memo because you didn't 

19    believe you had a sufficient opportunity to express your 

20    views at the July 8 meeting, isn't that correct?

21         A.    Correct.

22         Q.    After you submitted that memo, it's true, isn't 

23    it, that Deputy Under Secretary Mark Plant asked you to go 

24    to the Commerce Department's offices to act as a resource 

25    person to help write the Secretary's decision document, 
                                                              1837

 1    isn't that correct?

 2         A.    To help write a decision document, yes.

 3         Q.    And you did that during the weekend preceding 

 4    Monday, July 15, 1991, when the Secretary announced his 

 5    decision, correct?

 6         A.    Yes.

 7         Q.    And as a resource person, you were asked to 

 8    identify any errors in the decision document, correct?  

 9               THE COURT:  In the?  

10               MR. COHEN:  In the decision document.

11         A.    Yes.  That was my role over the weekend.

12         Q.    I want to refer you back to PX 673, second page, 

13    which is Bates stamp numbered 3221, specifically, to the 

14    third paragraph.

15               Would you read that paragraph aloud for us, 

16    please, Mr. Bounpane?

17         A.    "It is also expected that the number of people 

18    added to the dual system estimate from the fourth cell 

19    should be small and that the dual system estimate of total 

20    population should be 'lower than truth.'  That's because 

21    it's not expected that the dual system estimate can fully 

22    reflect people missed in both the census and the PES.  In 

23    past censuses, that has been the case.  For 1990, however, 

24    the data is not consistent with past experience.  Almost 5 

25    million people were added to the dual system estimate of 
                                                              1838

 1    total population from the fourth cell and the dual system 

 2    estimate of total population exceeded the demographic 

 3    analysis of total population - a very unexpected finding.  

 4    Taken together, these findings indicate there may be 

 5    problems in the dual system estimate that are not fully 

 6    understood."

 7         Q.    Thank you.

 8               That part of your memo expressed one of your 

 9    concerned about potential error in the dual system 

10    estimator, didn't it?

11         A.    Yes.

12         Q.    You considered provision of that information to 

13    the Secretary a provision of important information, didn't 

14    you?

15         A.    Yes, one of many.

16         Q.    You were very concerned that the Secretary have 

17    that information available to him, weren't you?

18         A.    Yes.

19         Q.    The number 5 million is a number you calculated 

20    from a computer printout, isn't that correct?

21         A.    Yes.

22         Q.    You calculated the number of persons in the 

23    fourth cell that you reported by yourself, didn't you?

24         A.    Yes.

25         Q.    Mr. Bounpane, the number of persons in the fourth 
                                                              1839

 1    cell of the dual system estimator is not 5 million, is it?

 2         A.    No, it's not.

 3         Q.    You calculated the number wrong, didn't you?

 4         A.    Yes, I did.

 5         Q.    In fact, as best you know at the moment, the 

 6    number of persons estimated to be in the fourth cell is only 

 7    about 1.5 million, correct?

 8         A.    Somewhere around that, yes.

 9         Q.    And that is not a number that gives you any 

10    concern about the error in the dual system estimate, is it?

11         A.    That's far closer in my range of expectation, 

12    yes.

13         Q.    Would you please go back to Plaintiff's Exhibit 9 

14    in your binder, which is the Secretary's decision, and I 

15    will ask you specifically to turn to page 2-20.

16               (Pause)

17               Do you have page 2-20, Mr. Bounpane?

18         A.    Oh, 2-20?  I'm sorry.

19               Okay, I have 2-20.

20         Q.    Would you look at the first full paragraph on 

21    page 2-20.

22         A.    The one that begins, "One also"?

23         Q.    That's correct?

24         A.    Okay.

25         Q.    You may read that to yourself, Mr. Bounpane.
                                                              1840

 1               (Pause)

 2         A.    Okay.

 3         Q.    Have you read that paragraph?

 4         A.    Yes.

 5         Q.    Would you agree, Mr. Bounpane, that the languages 

 6    in the first full paragraph on page 2-20 of the Secretary's 

 7    report is virtually identical to the paragraph in your July 

 8    10 memo that we just read that you explained?

 9         A.    Yes, it's very similar.

10         Q.    Let me refer you back to your July 10 memo, 

11    please, Plaintiff's Exhibit 673.

12         A.    Okay.

13         Q.    And specifically to the second page.

14         A.    Okay, I have it.

15         Q.    Could you read aloud, please, the last five 

16    sentences of the second paragraph beginning with, 

17    "Unfortunately, we have no direct data"?

18         A.    I'm sorry, I can't find where you are, Mr. Cohen.

19               The first page of the attachment or the second 

20    page of the attachment?

21         Q.    The second page of the attachment, I believe it's 

22    the first page of the discussion to the Secretary.

23         A.    The second page --

24         Q.    The last five sentences of the second paragraph 

25    in the --
                                                              1841

 1         A.    "Unfortunately," all right.

 2         Q.    Read it aloud, please.  

 3               MR. SUBAR:  I'm sorry, your Honor, the witness 

 4    apparently has it, I don't know if the court does, I don't.

 5               (Pause)

 6         Q.    The first page of the attachment, the second 

 7    page.  

 8               MR. SUBAR:  Thank you.

 9         A.    Do you want me to read that now?

10         Q.    Please.

11         A.    "Unfortunately, we have no direct data to verify 

12    if the assumptions in estimating the fourth cell are met.  

13    One piece of data indicates there may be a problem we don't 

14    fully understand.  The traditional wisdom is that males are 

15    generally more subject to correlation bias.  It is a 

16    reasonable assumption (supported by past data) that males 

17    are more likely to be missed in both the census and the PES.  

18    But in 1990, about one half of the people added to the dual 

19    system estimate of total population from the fourth cell are 

20    women."

21         Q.    You made the calculation about the proportion of 

22    males in the fourth cell of the dual system estimator at the 

23    same time that you calculated the total number of persons in 

24    the fourth cell, isn't that correct?

25         A.    Yes.
                                                              1842

 1         Q.    That was also a finding you were very concerned 

 2    to bring to the attention of the Secretary, isn't that 

 3    correct?

 4         A.    I was concerned, yes.

 5         Q.    Isn't it true, Mr. Bounpane, that you also 

 6    miscalculated the proportion of males and females estimated 

 7    by the fourth cell?

 8         A.    Well, to the extent I miscalculated the 5 

 9    million, yes, but it's my understanding that even on the new 

10    number the split is still almost 50/50.

11         Q.    It's true at your deposition of March 27 that you 

12    didn't know what that new number was?

13         A.    That's correct, I didn't know it at that time.

14         Q.    Mr. Bounpane, census procedures do not remain 

15    static from one census to the next, do they?

16         A.    No.

17         Q.    In fact, over the past 200 years, the census has 

18    constantly updated its procedures to adopt new more 

19    effective methodologies, isn't that correct?

20         A.    Yes.

21         Q.    The Census Bureau makes extensive use of 

22    computers in conducting the census, doesn't it?

23         A.    Yes.

24         Q.    And apart from adjustment based on the PES, the 

25    Bureau uses other statistical processes when compiling the 
                                                              1843

 1    counts, doesn't it?

 2         A.    Yes.

 3         Q.    Imputation is a statistical process, isn't it?

 4         A.    Yes.

 5         Q.    The Bureau has used imputations in several 

 6    censuses, correct?

 7         A.    Yes.

 8         Q.    In fact, imputation was challenged in the lawsuit 

 9    in 1980 brought by the State of Indiana, correct?

10         A.    One form of imputation was challenged by the 

11    State of Indiana, yes.

12         Q.    And you submitted an affidavit in defense of that 

13    lawsuit, correct?

14         A.    Actually, I don't recall, but perhaps I did.

15         Q.    And the Census Bureau prevailed in that lawsuit, 

16    isn't that correct?

17         A.    That's my understanding, yes.

18         Q.    Sampling was used in the 1970 census, wasn't it?

19         A.    Yes, it was.

20         Q.    And it added in the range of a million people to 

21    the count, isn't that correct?

22         A.    I don't know the precise number, but something 

23    like that sounds right.

24         Q.    Sampling is a statistical technique, isn't that 

25    correct?
                                                              1844

 1         A.    Yes.

 2         Q.    In fact, the PES is a sample survey, isn't that 

 3    correct?

 4         A.    Yes.

 5         Q.    It is fair to say, isn't it, that the 200 year 

 6    tradition of census-taking includes the use of sampling and 

 7    other statistical techniques, isn't that correct?

 8         A.    Well, at least recent ones do.  I'm not sure 

 9    about some of the earlier once.

10         Q.    Mr. Bounpane, I would like you to turn to the 

11    first page of the Secretary's decision, that is PX 9.

12         A.    Is that the --

13         Q.    I'm sorry, the first page of the Secretary 

14    statement.

15         A.    The summary statement?

16         Q.    The summary statement.

17         A.    Okay.

18         Q.    It is actually page 1-1, I believe, of the 

19    Secretary's report.  It is behind the page that starts 

20    Section 1.

21               You see the first sentence in the second 

22    paragraph where it reads, "If we change the counts by a 

23    computerized statistical process, we abandon a 200 year 

24    tradition of how we actually count people." 

25               Do you see that, Mr. Bounpane?
                                                              1845

 1         A.    Yes, I do.

 2         Q.    The Secretary is saying that by adjusting, we 

 3    would abandon a 200 year tradition, isn't that correct?  

 4               MR. SUBAR:  Objection.  It calls for speculation.  

 5               THE COURT:  Sustained.

 6         Q.    Mr. Bounpane, would you look at page 2-13 of the 

 7    Secretary's report, please.

 8               At the bottom of the page is footnote 18.  Do you 

 9    see that?

10         A.    Yes, I do.

11         Q.    You see where it says, "Nevertheless, this was at 

12    least the second best census ever conducted"?

13         A.    Yes, I see that.

14         Q.    If we are to consider only censuses since 1940, 

15    you agree with that statement, don't you?

16         A.    Well, I think what I agree, if you use the 

17    standard of net measured undercount by demographic analysis, 

18    then since 1940, 1990 was at least the second best census 

19    ever conducted.

20         Q.    And you don't know any other standard except 

21    using the net measured undercount under demographic analysis 

22    by which one could reach that conclusion, isn't that true?

23         A.    There are other standards.  I think at my 

24    deposition, for example, I mentioned to you at least in my 

25    personal observations of the three I have been involved in, 
                                                              1846

 1    I make some judgments about it as well.

 2         Q.    Let me ask you if you remember this question and 

 3    this answer at your deposition:

 4               "Q.    Are there any other bases upon which you 

 5    make that conclusion beside measurement under the 

 6    demographic analysis?

 7               "A.    Well, I guess other people could use other 

 8    standards.  I don't know what they could be." 

 9               Do you remember that question and that answer?

10         A.    I guess if that's what I said, that's what I 

11    said.

12         Q.    It's true, isn't it, that the differential 

13    undercount between blacks/nonblacks estimated by demographic 

14    analysis is higher for 1990 than for any other census for 

15    which we have a measurement?

16         A.    On the demographic analysis, yes.

17         Q.    In reaching your conclusion about whether the 

18    adjusted counts are more accurate than the unadjusted 

19    counts, did you consider the reports of the population 

20    division of the Bureau?

21         A.    I'm sorry, I'm not sure I know what reports are 

22    being referred to.

23         Q.    Did you consider the D projects?

24         A.    I'm sorry.  The demographic analysis and the 

25    studies that were associated with them?
                                                              1847

 1         Q.    Yes.

 2         A.    Yes.

 3         Q.    So you are aware, are you not, that the summary 

 4    of the D studies says that despite the inability of the 

 5    demographic method to estimate undercount levels with 

 6    precision, the demographic coverage estimates allows us to 

 7    make statements about differences in patterns of coverage, 

 8    isn't that correct?

 9         A.    That seems like a reasonable statement, yes.

10         Q.    You agree, don't you, that demographic analysis 

11    is not appropriately used to estimate undercount levels with 

12    precision?

13         A.    Yes.

14         Q.    And you agree, don't you, that the demographic 

15    analysis is appropriately used to make statements about 

16    differences in patterns of coverage, isn't that true?

17         A.    Differences in patterns of coverage over time?

18         Q.    Yes.

19         A.    Yes, that's one of its purposes.

20         Q.    One of those patterns is that the net undercount 

21    is greater for males than for females, correct?

22         A.    Generally true.

23         Q.    The demographic analysis estimates in 1990 show 

24    greater net undercount for males than for females, don't 

25    they?
                                                              1848

 1         A.    Yes.

 2         Q.    The PES estimates in 1990 show greater net 

 3    undercount for males than for females, isn't that correct?

 4         A.    Yes.

 5         Q.    Another pattern is that net undercount is greater 

 6    for blacks than for nonblacks, isn't that correct?

 7         A.    Yes.

 8         Q.    The DA estimates in 1990 show greater net 

 9    undercount for blacks than for nonblacks, isn't that 

10    correct?

11         A.    Yes.

12         Q.    The PES estimates of 1990 show greater net 

13    undercount for blacks than for nonblacks, correct?

14         A.    Yes.

15         Q.    Mr. Bounpane, I would like you to refer you back 

16    to the bottom of page 2-9 of the Secretary's report.  The 

17    last sentence reads, "Many argue that the DA estimates 

18    broadly corroborate differential undercounts implied by 

19    PES-adjusted counts, however, like the minority on the 

20    undercount steering committee, I find there are some 

21    important and puzzling differences." 

22               Do you see that, Mr. Bounpane?

23         A.    Yes.

24         Q.    That statement accurately represents the view of 

25    the USC minority, doesn't it?
                                                              1849

 1         A.    I think so.

 2               You mean the last piece of it?

 3         Q.    That's right.  There are some important and 

 4    puzzling --

 5         A.    There are some important and puzzling 

 6    differences, yes.

 7         Q.    The difference between the DA and the PES 

 8    estimates of net undercount was a very important reason for 

 9    your conclusion not to recommend in favor of adjustment, 

10    isn't that correct?

11         A.    I think it was an important reason, yes.  

12               THE COURT:  Are you anywhere near finished, Mr. 

13    Cohen?  

14               MR. COHEN:  No, I have a bit more to go, your 

15    Honor.  

16               THE COURT:  Let's take a break.

17               We will resume at 11:00 o'clock.

18               (Recess) 

19               THE COURT:  Mr. Cohen.  

20               MR. COHEN:  Thank you, your Honor.  

21    BY MR. COHEN:

22         Q.    Mr. Bounpane, we were talking about differences 

23    at the national level between the demographic analysis 

24    estimate of net undercount and the PES estimate of net 

25    undercount.
                                                              1850

 1               Do you recall that?

 2         A.    Okay.

 3         Q.    The difference at the national level was about 

 4    two/tenths of a percentage point, isn't that correct?

 5         A.    The PES was 2.08 and the demographic analysis was 

 6    about 1.8, yes, so two to three/tenths of a percent.

 7         Q.    The DA estimates for 1990 show greater net 

 8    undercount for black males than the PES estimates show, 

 9    isn't that correct?

10         A.    Yes.

11         Q.    That isn't a puzzling result, is it?

12         A.    No, I don't think so.

13         Q.    In fact, wouldn't you say that the undercoverage 

14    of black males in the PES in comparison to the DA was a 

15    result that everybody fully expected in 1990?

16         A.    Generally you would expect that, yes.

17         Q.    Let me refer you to page 2-10 of the Secretary's 

18    report.

19         A.    I have it.

20         Q.    The Secretary says there in the first sentence of 

21    the second full paragraph, does he not, "The third 

22    disturbing comparison between the PES and DA undercount 

23    rates is that all groups of black males (except those aged 

24    10 to 19) are substantially undercovered by the PES relative 

25    to the DA." 
                                                              1851

 1               Isn't that what he says?

 2         A.    That's what he says.

 3         Q.    Mr. Bounpane, it's true, isn't it, that the 

 4    undercount steering committee considered variation among 

 5    alternative ways of computing match status when there was 

 6    missing data in the PES?

 7         A.    Yes.

 8         Q.    And the undercount steering committee concluded, 

 9    did you it not, that the variation among alternative ways of 

10    imputing is relatively small, tended to generate the same 

11    answer and was not a problem, isn't that correct?

12         A.    For the level of imputations we got, yes, that's 

13    correct.

14         Q.    In fact, the undercount steering committee 

15    reached unanimous agreement on that point, isn't that 

16    correct?

17         A.    Yes.

18         Q.    And the undercount steering committee was aware 

19    of what the imputation rates in the P and E samples were 

20    when it reached unanimous agreement, isn't that correct?

21         A.    Yes.

22         Q.    If the estimated number of gross omissions was on 

23    the order of 14 million, you wouldn't find the imputation 

24    rate for the P-sample of about 1.9 percent high, isn't that 

25    correct?
                                                              1852

 1         A.    I need to understand just a little bit.

 2               An imputation rate of 1.5 percent translates to 

 3    about how many people and then compare that to 14 million.

 4         Q.    About 3 million.

 5         A.    All right.  I don't consider 3 million relative 

 6    to 14 million high.

 7         Q.    Mr. Bounpane, let me turn your attention to page 

 8    2-16 of the Secretary's report.

 9               About two-thirds down the page, it states, 

10    "The imputation rates for the two parts of the PES called 

11    the P and E samples, were high, 1.7 percent and 2.1 percent, 

12    respectively, which is equivalent to 3,900,000 and 5,025,000 

13    individuals in the census and weighted up to the national 

14    population estimate." 

15               Do you see that?

16         A.    Yes.

17         Q.    You agree, don't you, that the Secretary has not 

18    compared the imputation rates to the levels of gross 

19    omissions in that passage, don't you?

20         A.    Well, in reading the next sentence, it doesn't, 

21    he compared it to the net undercount apparently.

22         Q.    Mr. Bounpane, before July 15, 1991, you weren't 

23    aware of any data demonstrating that the error rate in the 

24    matching as estimated by the Bureau's evaluation studies was 

25    too low, isn't that correct?
                                                              1853

 1         A.    No, I wasn't aware of any data that showed that.

 2         Q.    I would like to direct your attention for a 

 3    moment to page 2-17 of the Secretary's decision.

 4               The paragraph under the heading "matching error" 

 5    in the last sentence of that paragraph reads:

 6               "There is evidence, though anecdotal, that is 

 7    suggestive of the fact that variance due to matching error 

 8    is conservatively estimated in the total error model." 

 9               Do you see that sentence?

10         A.    Yes.

11         Q.    That is Professor Wachter's anecdotal evidence 

12    the Secretary is referring in the report?

13         A.    I assume so.  Let me read further. 

14               (Pause)

15               Yes, apparently so.

16         Q.    You were never asked by the Commerce Department 

17    or the Secretary to review Professor Wachter's data?

18         A.    No.

19         Q.    Nor did you review his methodology for his 

20    anecdotal study of matching error, isn't that correct?

21         A.    I didn't review his study.

22         Q.    And you don't know of anybody else at the Bureau 

23    who was asked to review that study before July 15, 1991, do 

24    you?

25         A.    No, I don't.
                                                              1854

 1         Q.    As used with respect to the PES in the census, 

 2    the term "correlation bias" generally refers to those 

 3    disproportionately missed by both the census and the PES, 

 4    isn't that correct?

 5         A.    Generally, yes.

 6         Q.    Mr. Bounpane, you are confident, are you not, 

 7    that males are more subject to correlation bias than 

 8    females?

 9         A.    Generally, yes.

10         Q.    And you base your confidence on evidence that in 

11    general undercount in surveys tends to be concentrated in 

12    males rather than in females, isn't that correct?

13         A.    In censuses, yes.

14         Q.    And that same body of knowledge gives you 

15    confidence that those missed in both the census and the PES 

16    are disproportionately minority, correct?

17         A.    I would expect that, yes.

18         Q.    And that same body of knowledge leads you to 

19    conclude that those missed by both the census and the PES 

20    are disproportionately poor, correct?

21         A.    To the extent that more of the poor are minority 

22    than in the population at large, yes.

23         Q.    You agree with that statement, correct?

24         A.    I think so, yes.

25         Q.    And that same body of knowledge leads you to 
                                                              1855

 1    conclude that those missed by both the census and the PES 

 2    are disporportionaly renters as opposed to owners, isn't 

 3    that correct?

 4         A.    I'm sorry, would you say the question again?

 5         Q.    Sure.

 6               And that same body of knowledge leads you to 

 7    conclude that those missed by both the census and the PES 

 8    are disporportionaly renters as opposed to owners, isn't 

 9    that correct?

10         A.    That I'm not sure about.  To the extent that 

11    minorities and poor are higher level renters than the 

12    population at large, then I would say yes.

13         Q.    Mr. Bounpane, do you remember the following 

14    question and answer at your deposition on March 25, on page 

15    193.

16               "Q.    Does that same body of knowledge and 

17    reasoning lead you to conclude that those missed by both the 

18    census and the PES are disproportionately renters as opposed 

19    to owners?

20               "A.    I would think so, yes." 

21               Do you remember that question and that answer?

22         A.    I guess so, and I think that's similar to what I 

23    was just trying to say.

24         Q.    Your memory is now refreshed on that?

25         A.    Yes.
                                                              1856

 1         Q.    So it is fair to say that you have the same 

 2    amount of confidence that those missed by the PES and the 

 3    census are disproportionately male, minority and renters as 

 4    opposed to female, nonminority and owners, correct?

 5         A.    I would expect that, yes.

 6         Q.    Mr. Bounpane, let me refer you to the Secretary's 

 7    report at page 2-11.

 8         A.    Okay.

 9         Q.    In the middle of the page the decision says, does 

10    it not, "Special advisory member panel member Wachter 

11    estimates that the number of people missed by both the 

12    census and the PES may be as high as half a million.  We do 

13    not know where these people are." 

14               You see that, correct?

15         A.    Yes.

16         Q.    Part of the body of knowledge that you have 

17    reached your conclusions on is participant observer studies, 

18    isn't it?

19         A.    Yes.

20         Q.    Participant observer studies are also known as 

21    ethnographic studies?

22         A.    Yes.

23         Q.    And those studies are among the body of knowledge 

24    that bears on the question of where the unreached, both the 

25    census and the PES, live, correct?
                                                              1857

 1         A.    Well, I think it bears on where the unreached 

 2    live.

 3         Q.    Those are unreached by both the census and the 

 4    PES as well, correct?

 5         A.    Well, it's hard to separate that out, I think.

 6         Q.    So you agree?

 7         A.    I agree to what?

 8         Q.    You agree that it also bears on the question 

 9    where those unreached by both surveys are missed, correct?

10         A.    I said it's hard to learn that from them, yes.

11         Q.    And at your deposition you testified, didn't you, 

12    that you would be surprised if those studies were not 

13    contained in the administrative record of the Secretary's 

14    decision?  

15               MR. SUBAR:  Objection; outside the scope.  

16               THE COURT:  Overruled.

17         Q.    Do you recall that?

18         A.    I recall something like that, yes.

19         Q.    You recall the following questions and answers at 

20    your deposition of March 25 on page 196:

21               "Q.    As an example, what study is in the 

22    administrative record that you are referring to?"  

23               MR. SUBAR:  Objection.  Your Honor, it appears as 

24    if Mr. Cohen is going to try to use this deposition 

25    transcript to impeach Mr. Bounpane's testimony, but Mr. 
                                                              1858

 1    Bounpane just testified in a fashion consistent with what he 

 2    said he believes he testified to at his deposition.  

 3               THE COURT:  I will permit it.

 4         Q.    "A.    I would think that coverage from prior 

 5    censuses, and that is definitely in there.

 6               "Q.    Anything else?

 7               "A.    I would think some of these participant 

 8    observer studies that were done, I think they are in there.  

 9    That would shed light on this particular issue.

10               "Q.    You mean ethnographic studies?

11               "A.    Yes, that's the word I used for them.

12               "Q.    Is it your belief that those are in the 

13    administrative record?

14               "A.    I think so.

15               "Q.    You think that they should be in the 

16    administrative record? 

17               "Mr. Subar:  Objection to form in that it calls 

18    for a legal conclusion.

19               "Q.    Would you expect them to be in the 

20    administrative record?

21               "A.    I thought that they would be, yes. 

22               "Q.    Would you be surprised if they were not?

23               "A.    Well, only to the extent I thought they 

24    were in there and if they weren't I would be surprised." 

25               Do you remember those questions and those 
                                                              1859

 1    answers?

 2         A.    Yes.

 3         Q.    Were you in court the other day, Mr. Bounpane, 

 4    when it was learned that the ethnographic studies were not 

 5    in the administrative record?

 6         A.    I'm not sure I was in the court when that 

 7    specifically was discussed.

 8         Q.    That matter has come to your attention?

 9         A.    I now know they are not in the administrative 

10    record.

11         Q.    Mr. Bounpane, it's true that looking at a figure 

12    for net undercount by itself doesn't tell anything about 

13    differential undercount, isn't that correct?

14         A.    I'm sorry, looking at net undercount by itself 

15    does not --

16         Q.    Does not tell us anything about differential 

17    undercount?

18         A.    Well, if you break the net undercount by black or 

19    nonblack, that would tell you the difference of the 

20    undercount.

21         Q.    If you had one figure --

22         A.    Net undercount for the total population?  Yes, it 

23    does not tell you that.

24         Q.    And you agree, do you not, that the predominant 

25    reason that the PES was designed was to get good estimates 
                                                              1860

 1    of undercount by minority groups, correct?

 2         A.    Yes, that was the goal of the PES.

 3         Q.    And the purpose of getting those good estimates 

 4    was to be in a position to correct for differential 

 5    undercount, isn't that correct?

 6         A.    If the error structure in the PES was low enough 

 7    to permit it, yes.

 8         Q.    You agree, don't you, that nobody really cares 

 9    about net undercount if it is distributed proportionately?

10         A.    Generally people say that, yes.

11         Q.    And you agree with it, isn't that correct?

12         A.    In the sense that if the undercount is the same 

13    error, it doesn't effect relative shares, yes, I agree with 

14    that.

15         Q.    Mr. Bounpane, let me refer you to page 2-23 of 

16    the Secretary's decision.

17               Do you see where it says in the first paragraph, 

18    "On the one hand, the errors introduced by measured flaws in 

19    the PES process seem small.  On the other hand, the model 

20    does show that the PES is biased toward overestimating the 

21    undercount and that a biased corrected estimate of the 

22    undercount would be about 1.4 percent rather than the 

23    production estimate of 2.1 percent.  This means about a 

24    third of the net undercount adjustment in the DSE comes from 

25    bias in the PES." 
                                                              1861

 1               Do you see that?

 2         A.    Yes.

 3         Q.    The 2.1 percent figure that is recited in that 

 4    passage refers to net undercount, does it not?

 5         A.    Yes, it does.

 6         Q.    And that figure doesn't tell us anything about 

 7    differential undercount, does it?

 8         A.    No.

 9         Q.    Mr. Bounpane, let me refer you to page 2-25 of 

10    the Secretary's decision.

11               You see where the it says in the second full 

12    paragraph, "I conclude that the Constitutional and legal 

13    purposes for the census must take precedence, and accuracy 

14    should be designed predominantly in terms of getting the 

15    proportional distribution of the population right among 

16    geographical and political units." 

17               Do you see that sentence, Mr. Bounpane?

18         A.    Yes.

19         Q.    You agree with the Secretary, do you not, that 

20    distributive accuracy is of greater importance than numeric 

21    accuracy?

22         A.    In general, yes.

23         Q.    That was your concern when you deliberated on 

24    whether the adjusted counts are more accurate than the 

25    unadjusted counts, isn't that correct?
                                                              1862

 1         A.    I think so, yes.

 2         Q.    Some census programs relay on local 

 3    participation, correct?

 4         A.    Census programs --  you mean in terms of taking 

 5    the decennial census or the Census Bureau?

 6         Q.    In taking the decennial census, the Census Bureau 

 7    seeks the active participation of local governmental units, 

 8    correct?

 9         A.    That's correct.

10         Q.    Post-census local review is an example of such a 

11    program, correct?

12         A.    Yes.

13         Q.    Post-census local review jurisdictions are given 

14    preliminary census counts to compare with local data for 

15    discrepancies, correct?

16         A.    Correct.

17         Q.    Isn't it true that the effectiveness of 

18    post-census local review depends on the degree of 

19    participation of the local jurisdiction?

20         A.    In some ways, yes.

21         Q.    The feedback from local jurisdictions varied in 

22    post-census local review, isn't that correct?

23         A.    Feedback varied, is that the question?  Yes.

24         Q.    Certain jurisdictions participated more 

25    aggresively than others, isn't that correct?
                                                              1863

 1         A.    Correct.

 2         Q.    Some jurisdictions didn't participate at all, 

 3    isn't that correct?

 4         A.    Yes.

 5         Q.    Local review could increase the count of some 

 6    areas of a state and leave others unchanged, isn't that 

 7    correct?

 8         A.    Yes.

 9         Q.    And that certainly happened in 1990, didn't it?

10         A.    I would expect so, yes.

11         Q.    And certain local areas put more resources in 

12    advertising and outreach than others did, isn't that 

13    correct?

14         A.    Yes.

15         Q.    It is your position, isn't it, that advertising 

16    and outreach helped to enhance the count of a jurisdiction 

17    that uses them aggressively?

18         A.    In general, yes.

19         Q.    The Census Bureau encourages areas to participate 

20    in post-census local review, isn't that correct?

21         A.    Yes.

22         Q.    And it encourages local jurisdictions to conduct 

23    outreach and advertising, correct?

24         A.    Yes.

25         Q.    The Bureau never discourages a jurisdiction from 
                                                              1864

 1    participating in a program because neighboring jurisdictions 

 2    are not participating, isn't that correct?

 3         A.    That's correct.

 4         Q.    And the Bureau doesn't expect all jurisdictions 

 5    to participate, isn't that correct?

 6         A.    We hope they did, but it doesn't expect it, no.

 7         Q.    And your experience is that not all local 

 8    jurisdictions do participate, isn't that correct?

 9         A.    That's correct.

10         Q.    And the Bureau expects that of those 

11    jurisdictions that do participate, there have been varying 

12    degrees of participation, correct?

13         A.    In general, yes.

14         Q.    And the Bureau expects that programs for local 

15    participation will be more successful in some areas than in 

16    others, correct?

17         A.    I would expect so, yes.

18         Q.    Isn't it a fact that these programs add different 

19    proportions of persons to the count in different 

20    jurisdictions?

21         A.    It's certainly a possibility.  I don't know of 

22    any data that show that.

23         Q.    Isn't that your expectation?

24         A.    It is, except that most of these come relatively 

25    late in the census process, so we are talking about very 
                                                              1865

 1    small numbers of people being added through them.

 2         Q.    We discussed earlier that generally minority 

 3    areas have lower mail response areas than nonminority areas.  

 4    You remember that?

 5         A.    Yes.

 6         Q.    And nonresponse follow-up is more difficult in 

 7    minority and low income neighborhoods than in other 

 8    neighborhoods, correct?

 9         A.    Yes.

10         Q.    The Bureau never tries to reduce the 

11    effectiveness of mailout/mailback in nonminority areas in 

12    order to bring down the effectiveness of the procedure to 

13    the level seen in minority areas, does it?

14         A.    There were some double negatives in that.  Can we 

15    go over the question again and make sure I answer it right?

16         Q.    Sure.

17               The Bureau never tries to reduce the 

18    effectiveness of mailout/mailback in nonminority areas in 

19    order to bring down the effectiveness of the procedures to 

20    the level seen in minority areas, does it?

21         A.    No, it doesn't do that.

22         Q.    And the Bureau never tries to reduce the 

23    effectiveness of nonresponse follow-up so that it is at the 

24    same level in nonminority areas as in minority areas, 

25    correct?
                                                              1866

 1         A.    No, it doesn't do that.

 2         Q.    And the Bureau does not conduct any evaluation of 

 3    address list compilation procedures to determine their 

 4    effect on the distributive accuracy of the counts, correct?

 5         A.    Only broadly, for example, Type 1 areas as 

 6    opposed to Type 2 areas, but no more than that.

 7         Q.    In fact, the Bureau does not conduct any test to 

 8    determine the effect of any Census Bureau procedure on 

 9    distributive accuracy, isn't that correct?

10         A.    I'm not sure I can say any.  Certainly some 

11    things might get analyzed by racial background, but in 

12    general no.

13         Q.    In general, isn't it correct that the only test 

14    for distributional accuracy that the Bureau applies to any 

15    of these programs is whether they are achieving a more 

16    complete count?

17         A.    In general, yes.

18         Q.    Mr. Bounpane, it is your view, is it not, that 

19    loss function analysis is a proper statistical test of which 

20    set of counts is likelier to be closer to the truth?  

21               MR. SUBAR:  Objection; outside the scope.  

22               THE COURT:  Overruled.

23               If you are able to, you may answer that.

24         A.    It's my view that loss function analysis is one 

25    tool to help you answer the question of whether or not 
                                                              1867

 1    adjusted counts are more accurate than census counts, yes.

 2         Q.    And you reviewed the Bureau's loss function 

 3    analyses during your deliberations as a member of the USC, 

 4    didn't you?

 5         A.    Yes.

 6         Q.    The Bureau's loss function analysis for states 

 7    provided an aggregated sum of losses across all states, 

 8    isn't that correct?

 9         A.    Yes, it did.

10         Q.    And there was more aggregate loss under the 

11    census than under the adjusted counts in terms of 

12    distributional accuracy in the analyses you saw, isn't that 

13    correct?

14         A.    You mean in total population?  Is that what you 

15    mean by aggregate?

16         Q.    The aggregate loss that was reported in the loss 

17    function test.

18         A.    Yes.  Under the understanding that aggregate loss 

19    means that for total population the loss for the census was 

20    worse than the loss for adjustment, yes, in the ones we saw.

21         Q.    The loss function analyses also reported for any 

22    given state whether there was a greater likelihood of loss 

23    in accuracy under the census or under the adjustment, 

24    correct?

25         A.    Yes.
                                                              1868

 1         Q.    In considering the loss function analysis, it was 

 2    more important to you how many states were likely to be 

 3    improved or disimproved than what the aggregate loss was, 

 4    isn't that correct?

 5         A.    Yes.

 6         Q.    And you agree, don't you, that in assessing the 

 7    loss function in that manner, you are weighting a small loss 

 8    in accuracy in one state equally with a large gain in 

 9    accuracy in another state, correct?

10         A.    Under certain loss functions, yes.

11         Q.    Under the loss functions that you looked at 

12    that's true, isn't it?

13         A.    Under some of the loss functions we looked at 

14    that was true, yes.

15         Q.    And it is your view that weighting each state 

16    equally, no matter how large the improvement or 

17    disimprovement, is the appropriate analysis, correct?

18         A.    Yes, under the condition that loss function 

19    analysis is just one tool for coming to this conclusion of 

20    whether or not adjusted counts are more accurate than census 

21    counts.

22         Q.    So the answer is yes?

23         A.    Yes.

24         Q.    You also didn't consider the relative 

25    probabilities that states would be improved or disimproved, 
                                                              1869

 1    isn't that correct?

 2         A.    No, I don't think we considered that.

 3         Q.    You didn't consider them, did you?

 4         A.    No, I didn't if the group didn't.

 5         Q.    In making your assessment, you were looking to 

 6    see that 80 to 90 percent of all states were improved under 

 7    adjustment, correct?

 8         A.    I said that that was my personal opinion about 

 9    how many I would have expected to be better before I would 

10    be confident that an adjustment should go forward.

11         Q.    And that was weighting each state equally no 

12    matter how large the gain or loss, is that correct?

13         A.    No, not necessarily, because some of those loss 

14    function analyses do take account of the size of the state.

15         Q.    And that was about the same guideline, that 80 to 

16    90 percent, that you looked for in improvement for cities, 

17    isn't that correct?

18         A.    I said it would be about the same, yes.

19         Q.    And you applied the same ratio to counties, 

20    correct?

21         A.    Generally, yes.

22         Q.    And that percentage that you applied was your 

23    understanding of the requirement of Guideline One that the 

24    census shall be considered the most accurate count of the 

25    population of the United States at the national, state and 
                                                              1870

 1    local level, unless an adjusted count is shown to be more 

 2    accurate, isn't that correct?

 3         A.    That was my personal view on that, yes.

 4         Q.    And that's how you conducted your deliberations 

 5    on the USC, isn't that correct?

 6         A.    Yes.

 7         Q.    You worked with the USC majority, correct?

 8         A.    I'm sorry?

 9         Q.    You had deliberations with the majority of the 

10    USC, correct?

11         A.    Oh, yes.

12         Q.    You heard the members of the majority express 

13    their views?

14         A.    Yes.

15         Q.    You agree, don't you, that the majority of the 

16    USC did not assume that the adjusted counts were more 

17    accurate unless the unadjusted counts were shown to be 

18    better, don't you?  

19               MR. SUBAR:  Objection, it calls for speculation.  

20               THE COURT:  You may answer it if you can.

21         A.    I'm sorry, can I have the question repeated?

22         Q.    You agree, do you not, that the majority of the 

23    USC did not assume that the adjusted counts were more 

24    accurate unless the unadjusted counts were shown to be 

25    better?
                                                              1871

 1         A.    I'm trying to restate the question.  Did not 

 2    assume that the unadjusted counts were more accurate?

 3         Q.    Unless --  I'm sorry, let me repeat it for you.

 4         A.    Okay.

 5         Q.    You agree, do you not, that the majority of the 

 6    USC did not assume that the adjusted counts were more 

 7    accurate unless the unadjusted counts were shown to be 

 8    better?

 9         A.    I'm sorry, the question is so long I'm having a 

10    hard time with that.

11               I think that the majority came to the conclusion 

12    that they felt on average the adjusted counts were more 

13    accurate.

14         Q.    But they weren't starting with the assumption 

15    that the adjusted counts were more accurate, were they?

16         A.    No, they weren't starting with that assumption.

17         Q.    They weren't seeking to have the superior 

18    accuracy of the unadjusted counts demonstrated before they 

19    rejected the adjusted counts, were they?

20         A.    They were not trying to do what about the 

21    unadjusted counts? 

22               I'm sorry, say the question again.

23         Q.    You agree, do you not, that the majority of the 

24    USC did not assume that the adjusted counts were more 

25    accurate unless the unadjusted counts were shown to be 
                                                              1872

 1    better?

 2         A.    If I understand your question right, yes, I agree 

 3    with that.

 4         Q.    Let me refer you to page 2-29 of the Secretary's 

 5    decision document.

 6               You see in the top line where it says, "Rather 

 7    than using Guideline One's mandate that the actual 

 8    enumeration be deemed more accurate unless the adjusted 

 9    counts are shown convincingly to be more accurate, the 

10    Census Bureau reverse --  they preferred adjusted counts if 

11    the actual enumeration was not proven more accurate?

12         A.    I see that sentence.

13         Q.    That is not your understanding of what the USC 

14    majority did, is it?

15         A.    No.  I told you what I thought we concluded.

16         Q.    Thanks.

17               A moment ago, Mr. Bounpane, you were discussing 

18    an 80 percent to 90 percent improvement guideline that you 

19    applied to states, cities and counties.

20               Do you recall that?

21         A.    Yes.

22         Q.    You didn't apply that same percentage to census 

23    blocks, isn't that correct?

24         A.    No, I didn't.

25         Q.    And that's because you believed that it is 
                                                              1873

 1    impossible to say whether for any given block adjusted or 

 2    unadjusted counts are better, isn't that correct?

 3         A.    Generally, yes.

 4         Q.    The Census Bureau doesn't evaluate any of its 

 5    programs to determine whether they improve or degrade block 

 6    level accuracy, isn't that correct?

 7         A.    I don't know of any others.

 8         Q.    Reapportionate counts are reported before 

 9    December 31 of the census year, correct?

10         A.    Yes.

11         Q.    They are used for apportioning seats in the House 

12    of Representatives among states, correct?

13         A.    Yes.

14         Q.    Reapportionate counts are not reported at the 

15    block level, isn't that correct?

16         A.    Yes.

17         Q.    Mr. Bounpane, let me direct your attention to 

18    page 2-42 of the Secretary's report.

19               About two-thirds of the way into the first 

20    paragraph, you see the sentence that states, "In Wachter's 

21    simulations, the level of the population is improved about 

22    twice as often as it is worsened by an adjustment." 

23               Do you recall we discussed Professor Wachter's 

24    study of the homogeneity assumption at your deposition?

25         A.    Just generally, yes.
                                                              1874

 1         Q.    Now, you agree, don't you, that a simulation test 

 2    in which the level of the population is improved about twice 

 3    as often as it is worsened by an adjustment shows a 

 4    significant improvement, don't you?

 5         A.    Well, it shows an improvement.

 6         Q.    As you understand what the Secretary states about 

 7    Professor Wachter's study, that is a result that would give 

 8    you confidence in the homogeneity assumption, correct?

 9         A.    I'm not sure that --  I'd have to read all of 

10    this to understand what he was saying about the homogeneity 

11    assumption here.

12               And this is on absolute counts, not relative, if 

13    I may read it right.

14         Q.    You may read it.

15               (Pause)

16         Q.    It's only one paragraph.

17         A.    No.  I would like to know how it fits in with 

18    everything else.

19               (Pause)

20               Do you have the question in mind, Mr. Bounpane?

21         A.    Okay.  It looks like this is a study of about ten 

22    blockbusters in which Professor Wachter was trying to learn 

23    something about the homogeneity assumption and when he did 

24    it on absolute shares, he found that twice as often you got 

25    improvement with the adjusted counts.  That's how I read 
                                                              1875

 1    what it says there.  And then he got a different result when 

 2    he did it on proportionate shares.

 3    

 4    

 5    

 6               (Continued on the next page) 

 7    

 8    

 9    

10    

11    

12    

13    

14    

15    

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1876

 1         Q.    You agree, though, that with respect to the 

 2    numerical counts, that that result shows a significant 

 3    improvement, don't you?

 4         A.    Yes.  I would think so, yes.
                                             

 5         Q.    Look two sentences later, where it states, "On 

 6    average, 59 percent of the office proportions are better."

 7         A.    Yes.

 8         Q.    You agree, don't you, that results showing 

 9    improvement in proportions on average, 59 percent of the 

10    offices, tends to support the homogeneity assumption?

11         A.    Would tend to support, yes.

12         Q.    The Secretary states in the next sentence, "In 7 

13    percent of the simulation trials, the majority of the 

14    districts are made worse."  Do you see that? 

15         A.    Yes.

16         Q.    You consider a simulation test in which 93 

17    percent of the majority of shares were improved to be a 

18    positive sign, tending to support the holding of the 

19    homogeneity assumption, don't you?

20         A.    If you consider better than 50 a support, yes.  I 

21    assume that is all it is referring to, how many times was it 

22    under 50 percent improvement and how many times was it over.  

23    It doesn't say how much over or anything like that.

24         Q.    The homogeneity assumption doesn't have to hold 

25    exactly in order for their to be improvement from 
                                                              1877

 1    adjustment, isn't that correct?

 2         A.    Not exactly, no.

 3         Q.    You don't consider the findings of Professor 

 4    Wachter's study, at least as reported by the Secretary in 

 5    this paragraph, to support a conclusion that local 

 6    heterogeneity is a serious problem for adjusting the 1990 

 7    census, do you?

 8         A.    I conclude from this the same as I conclude from 

 9    the studies that we looked at, which is there is just not 

10    enough information to know whether the homogeneity 

11    assumption holds sufficiently to get more accurate counts.

12         Q.    So the answer is you don't conclude from 

13    Professor Wachter's study that local heterogeneity is a 

14    problem, isn't that correct?

15         A.    That local heterogeneity is a problem?

16         Q.    I don't know how to answer that at this time 

17    without understanding this better.

18         Q.    Mr. Bounpane, do you remember the following 

19    question and answer at your deposition on March 25th through 

20    the 27th?

21              "Q.    I take it, then, you would not find that 

22    the facts as I just asked you to assume them would support a 

23    conclusion that local heterogeneity is a serious problem for 

24    adjusting in the 1990 census?

25              "A.    By themselves, no."  
                                                              1878

 1               Do you remember that question and answer?

 2         A.    If that is what I said then, that is what I said.

 3         Q.    Turn, please, to page 2-53 of the Secretary's 

 4    report.

 5         A.    Okay.

 6         Q.    Do you see the last two sentences of the first 

 7    full paragraph where it says, "For example, the variance 

 8    presmoothing changes the estimated net undercount in the 

 9    west south central region from 2.95 percent to 2.76 percent.  

10    In the east south central region, it changes from 2.43 

11    percent to 2.68 percent.  Again, these are changes of a very 

12    significant magnitude, given the size of the national head 

13    undercount."  

14               Do you see those sentences, Mr. Bounpane?

15         A.    Yes.

16         Q.    You agree, don't you, that the change reported in 

17    the west south central region from 2.95 percent to 2.76 

18    percent of net undercount is not very significant?

19         A.    I would not call that very significant relative 

20    to the national undercount, no.

21         Q.    You also disagree that the change for the east 

22    south central region is very significant, don't you?

23         A.    Pardon?  I'm sorry.

24         Q.    You also disagree that the change for the east 

25    south central region is very significant, don't you?
                                                              1879

 1         A.    The difference there starts to be more.  Whether 

 2    we are approaching the very significant magnitude, I don't 

 3    know yet.  It's about a tenth of the estimated undercount.

 4         Q.    Mr. Bounpane, you don't know of any evidence that 

 5    shows that state and local government participation in 

 6    future census efforts will be reduced by an adjustment of 

 7    the 1990 census, do you?

 8         A.    No, I don't know of any evidence.

 9         Q.    And you don't know of any evidence that shows 

10    that residents will be less likely to participate in future 

11    censuses if there is an adjustment in 1990, isn't that 

12    correct?

13         A.    I don't know of any such evidence.

14         Q.    Mr. Bounpane, what is the total error model?

15         A.    The total error model is an attempt to take each 

16    of the component errors in the dual-system estimate PES 

17    process, put them all together into a total called the total 

18    error model.

19         Q.    The P study, P-16, embodies that total error 

20    model, is that correct?

21         A.    I'm not sure I know the study by its number, but 

22    there is a study that does that, yes.

23         Q.    Is Mary Mulry the author of the study on the 

24    total error model?

25         A.    Yes, I think so.
                                                              1880

 1         Q.    How long has Mary Mulry worked on the total error 

 2    model?

 3         A.    Many years.  I don't know how many, but a long 

 4    time.

 5         Q.    Do you consider her knowledgeable about the total 

 6    error model?  

 7               MR. SUBAR:  Objection.  Outside the scope.  

 8               THE COURT:  I will permit it.

 9         A.    Yes, I do.  

10               MR. COHEN:  At this time I have no further 

11    questions, your Honor.  

12               THE COURT:  Any redirect?  

13               MR. SUBAR:  Yes, your Honor.  

14    REDIRECT EXAMINATION 
                           

15    BY MR. SUBAR:

16         Q.    Mr. Bounpane, was the census a perfect 

17    enumeration of the United States population in 1990?

18         A.    No, I don't think so.

19         Q.    Does that mean that the PES is necessarily the 

20    way to fix the differential undercount for all 

21    jurisdictions?

22         A.    No.

23         Q.    Is an evaluation poststratum a piece of 

24    geography, Mr. Bounpane?

25         A.    No, it isn't.
                                                              1881

 1         Q.    I would like to refer you to Defendants' Exhibit 

 2    31, attachment 1.

 3         A.    Yes.

 4         Q.    Do you recall Mr. Cohen asked you some questions 

 5    about that document?

 6         A.    Yes.

 7         Q.    In particular, he asked you some questions about 

 8    some numbers in the column under Northeast.  Do you recall 

 9    that?

10         A.    Yes.

11         Q.    Is the Northeast a state?

12         A.    No.

13         Q.    Mr. Bounpane, do you agree with every point, 

14    every jot and tittle in the Secretary's decision document?

15         A.    No.

16         Q.    Do you believe that he reached a reasonable 

17    conclusion in that document?

18         A.    Yes.

19         Q.    Do you recall that Mr. Cohen asked you about 

20    sampling that was used in the 1970 census?

21         A.    Yes.

22         Q.    What sort of sampling was used in that census?

23         A.    Briefly, at the conclusion of the census, it 

24    appeared that there was a problem with the count, so a 

25    sample of units designated vacant in the census was 
                                                              1882

 1    reinterviewed.  From that reinterview a conversion rate to 

 2    occupied was determined.  That was weighted up from the 

 3    sample and applied to the nation as a whole.  There was a 

 4    similar kind of activity reported on that check with the 

 5    post office done after the census was over which had the 

 6    acronym PPOC.

 7         Q.    Mr. Bounpane, I am going to read to you a passage 

 8    from Judge Sprizzo's decision in Cuomo v. Baldridge, at page 

 9    1107 of 674 F.Supp.  I would like to ask you whether you 

10    agree with it.  

11               MR. COHEN:  I object to asking this witness, who 

12    is not an attorney, to comment on a legal opinion, your 

13    Honor.  I don't see the purpose of it.  

14               THE COURT:  It may not be legal.  

15               MR. COHEN:  If that is the way your Honor is 

16    leaning, then I have no objection.  

17               THE COURT:  Let me hear the sentence.

18         Q.    The sentence is as follows.  "In the first place, 

19    none of those adjustments in 1970 were even remotely similar 

20    to the types of wholesale adjustments presently suggested by 

21    plaintiffs."  

22               Do you agree with that?

23         A.    Yes.  In general I think the kinds of things that 

24    were done in 1970 are not very similar to the kind of 

25    adjustment we are talking about in this case.
                                                              1883

 1         Q.    Mr. Cohen asked you some questions about 

 2    ethnographic studies.  Do you know whether the Secretary of 

 3    Commerce had the ethnographic studies available to him 

 4    before he made his decision?

 5         A.    No, I don't.

 6         Q.    Do you know whether Barbara Bryant had them 

 7    available to her before she made her recommendation to the 

 8    Secretary?

 9         A.    No, I don't.

10         Q.    Did the Undercount Steering Committee have the 

11    final studies available to it in June?

12         A.    If we had any, it was one or two available to us.  

13    Most of them came after our deliberations. 

14         Q.    Mr. Bounpane, does the bureau provide block-level 

15    data to states for any particular purpose?

16         A.    Yes.

17         Q.    For what purpose?

18         A.    So that states can draw within state districts, 

19    whether they be Congressional districts, state legislative 

20    districts, city wards, etc.

21         Q.    To calculate the state level apportionment 

22    counts, does the census bureau aggregate block level census 

23    counts for all blocks in each state?

24         A.    Yes, it does.

25         Q.    What was the dual strategy that the census bureau 
                                                              1884

 1    was contemplating back in 1987?

 2         A.    Basically, that refers to our goal of doing the 

 3    census and also continuing the research on the PES and 

 4    adjustment, examining the results after the census was over, 

 5    and then, based on examination of those results, making a 

 6    determination about whether or not to adjust the census.

 7         Q.    Could you take a looking, please, at Plaintiffs' 

 8    Exhibit 730:  Do you have it?

 9         A.    Yes, I do.

10         Q.    Mr. Cohen asked you some questions about 

11    paragraph 4 of that exhibit.  What does that paragraph mean?

12         A.    This was my response to a question about should 

13    the answer be yes or no, we should try to adjust by 12/31 

14    and trying to make that decision in April of 1987.  So it 

15    was my opinion about what would happen if the census bureau 

16    said yes, we will adjust and yes, we will adjust by this 

17    date.

18         Q.    Mr. Bounpane, there was a .7 percent bias with 

19    regard to the July 1991 PES numbers, is that correct?

20         A.    That's correct.

21         Q.    Was that a significant number, in your view?

22         A.    Yes, it was.

23         Q.    Why was that?

24         A.    I consider a level of error of one-third of the 

25    number we are trying to estimate significant.  And, more 
                                                              1885

 1    importantly, we had no real evidence of how that error 

 2    distributed to geographic locations.

 3         Q.    Mr. Bounpane, could you take a look, please, at 

 4    Plaintiffs' Exhibit 673, which is also Defendants' Exhibit 

 5    76, which is also part of the Defendants' Exhibit 1.

 6         A.    I have it.

 7               MR. SUBAR:  At least two of which are in 

 8    evidence, your Honor.

 9         Q.    Mr. Cohen asked you some questions about the 

10    material on the second page of the document, which is the 

11    first page of the attachment to the cover memo.  Do you 

12    recall that?

13         A.    Yes.

14         Q.    Mr. Cohen asked you whether you calculated the 

15    number 5 million that appears in the third paragraph of that 

16    page.  Do you recall that?

17         A.    Yes.

18         Q.    He pointed out, as you testified yesterday, I 

19    believe, that that number is erroneous, correct?

20         A.    Yes.

21         Q.    When did you first find out that there was a 

22    problem with that number?

23         A.    Four, five, six months ago, something like that.

24         Q.    Was it after the July 1991 decision?

25         A.    Yes, it was.
                                                              1886

 1         Q.    Did you make any efforts before if July 1991 

 2    decision to present this number and the calculations that 

 3    went into that number and your thoughts with regard to that 

 4    number to your colleagues on the Undercount Steering 

 5    Committee?

 6         A.    Yes, I did.

 7         Q.    Could you tell the Court about that, please.

 8         A.    I made efforts to present a lot of the things 

 9    that are in this particular memo that I wrote to the 

10    Secretary at our deliberations of the Undercount Steering 

11    Committee.  I really would like to have discussed these with 

12    my colleagues a lot more fully before writing this.  In my 

13    opinion, unfortunately, at several times during the 

14    deliberations it seemed like I was cut off, and I didn't 

15    have a chance to fully discuss this with my colleagues.

16         Q.    In particular, were you cut off in that way at or 

17    near the time of the final Undercount Steering Committee 

18    vote with regard to its recommendation?

19         A.    Yes.  In particular, the Friday before the 

20    weekend that the report was written, so the last decision 

21    date for our committee, was the date that I had hoped to be 

22    able to present all of this and didn't get a chance to do 

23    that.

24         Q.    Under what circumstances did you not get a chance 

25    to do that?
                                                              1887

 1         A.    I tried.  I wrote some things on the blackboard, 

 2    my list of concerns here, this number.  I asked the chair if 

 3    she would please recognize me to have that discussed.  I 

 4    think the committee was just fatigued or something, because 

 5    I just did not get a chance to discuss any of this with 

 6    them.

 7         Q.    Mr. Bounpane, referring your attention to the 

 8    second paragraph on the same page.

 9         A.    Yes.

10         Q.    Mr. Cohen asked you some questions about the 

11    sentences reading, "One piece of data indicates there may be 

12    a problem we don't fully understand.  Traditional wisdom is 

13    that males are generally more subject to correlation bias.  

14    It is a reasonable assumption, supported by past data, that 

15    males are more likely to be missed in both the census and 

16    the PES, but in 1990 about one half of the people added to 

17    the DSE estimate of total population for the fourth cell are 

18    women."

19               When Mr. Cohen asked you about that, you made 

20    reference to some information that you have now with regard 

21    to whether that split between males and females is still 

22    50-50.  Could you tell the Court about that evidence? 

23               MR. COHEN:  Objection.  

24               THE COURT:  Overruled.

25         A.    There were two reasons why I made a mistake in 
                                                              1888

 1    coming up with the 5 million.  One was I simply made a 

 2    computational error.  But the other is, unbeknownst to me, 

 3    the computer printout that I had listed a set of numbers 

 4    which were only the positive values in the fourth cell.  I 

 5    did not realize that there were also a whole number of 

 6    negative numbers in the fourth cell to offset that.  

 7               You may remember that two-by-two table there 

 8    yesterday.  Obviously, it is impossible to have negative 

 9    people in the fourth cell.  You can't have negative people 

10    not counted in the census and not in the PES.  They arise 

11    because of sort of the fragile nature of this dual-system 

12    estimate.  By estimating the total and doing the 

13    subtraction, you sometimes arrive at negative numbers there.  

14               I didn't know that at the time that I added these 

15    up.

16         Q.    Did you prepare that computer printout that you 

17    just mentioned?

18         A.    No.

19         Q.    Who did?

20         A.    Mary Mulry.

21         Q.    Early on in your cross-examination, Mr. Cohen 

22    asked you whether you had mentioned several aspects of the 

23    PES in your testimony yesterday.  You indicated that several 

24    things that Mr. Cohen pointed out this morning you had not 

25    mentioned yesterday.  
                                                              1889

 1               In addition to the aspects of the PES that you 

 2    mentioned yesterday and the aspects of the PES that Mr. 

 3    Cohen mentioned this morning, are there any other aspects of 

 4    the PES that you haven't told the Court about?

 5         A.    Yes, surely.

 6         Q.    Many, few?

 7         A.    Many.

 8         Q.    Was Paula Schneider on the Undercount Steering 

 9    Committee?

10         A.    Yes.

11         Q.    Was she the chair of that committee?

12         A.    Yes, she was.

13         Q.    Does she have any graduate degree?

14         A.    I don't think so.

15         Q.    Was Susan Miskura on the Undercount Steering 

16    Committee?

17         A.    Yes.

18         Q.    Does she have any graduate degree?

19         A.    Not that I know of.

20         Q.    Is Charles Jones a fellow of the American 

21    Statistical Association?

22         A.    Yes, he is.

23         Q.    How long has Charles Jones been involved in the 

24    planning and evaluation of the decennial census?

25         A.    30 years or so.
                                                              1890

 1         Q.    How long have you been involved in the planning 

 2    and evaluation of the decennial census?

 3         A.    27 years.

 4         Q.    Do you recall that at the beginning of your 

 5    cross-examination Mr. Cohen asked you whether Mr. Jones is 

 6    your boss?

 7         A.    Yes.

 8         Q.    Did Mr. Jones tell you how to vote on the 

 9    Undercount Steering Committee?

10         A.    No.

11         Q.    Have you ever disagreed with Mr. Jones with 

12    regard to census matters?

13         A.    Yes.

14         Q.    Many occasions or few?

15         A.    Many.  I talked about a few of those this morning 

16    on reading back my deposition, that I thought some basic 

17    choices in how to take the 1990 census I disagreed with Mr. 

18    Jones on.

19         Q.    Did you make your disagreements known to Mr. 

20    Jones?

21         A.    Surely.

22         Q.    Did I ask you whether Mr. Jones told you how to 

23    vote on the Undercount Steering Committee?

24         A.    Yes, you asked me that.

25         Q.    Your answer was?
                                                              1891

 1         A.    No.

 2         Q.    Why did you vote as you did with regard to your 

 3    recommendation on using the PES numbers?

 4         A.    Personal assessment on my part when looking at 

 5    the information we had available to us and all potential 

 6    error that could be in it, that I was not convinced that 

 7    adjusted counts were going to be more accurate than census 

 8    counts.  

 9               MR. SUBAR:  I have nothing further, your Honor.  

10               MR. COHEN:  One brief question.  

11    RECROSS-EXAMINATION 
                          

12    BY MR. COHEN:

13         Q.    You mentioned on Mr. Subar's examination that you 

14    didn't have complete reports of the ethnographic studies, 

15    isn't that correct?

16         A.    I don't think all the ethnographic studies were 

17    done at the time that the Undercount Steering Committee met 

18    to make its recommendation.

19         Q.    You had drafts of all the studies, isn't that 

20    correct?

21         A.    I think we had drafts of some of them.  I don't 

22    think we had all.

23         Q.    Isn't it true as well that there was a conference 

24    before the Secretary's decision attended by census bureau 

25    personnel at which those ethnographic studies were 
                                                              1892

 1    presented?

 2         A.    I think there was.  I can't remember the exact 

 3    date, but I think it was before his decision, yes.

 4         Q.    And that conference was video taped for 

 5    distribution within the census bureau, correct?

 6         A.    It was video taped, yes.

 7         Q.    The video tape was generally available to the 

 8    commerce department as well, isn't that correct?

 9         A.    I would assume if it was made, it was available 

10    to them, yes.  

11               MR. COHEN:  No further questions.  

12    REDIRECT EXAMINATION 
                           

13    BY MR. SUBAR:

14         Q.    Mr. Bounpane, do you know whether the Secretary 

15    reviewed that video tape before he made his decision?

16         A.    No, I don't.  

17               MR. SUBAR:  Nothing further.  

18               THE COURT:  Thank you, Mr. Bounpane.  You may 

19    step down.  

20               (Witness excused) 

21               THE COURT:  Take three or four minutes before you 

22    call next witness.  

23               (Recess) 

24               THE COURT:  Mr. Millet?  

25               MR. MILLET:  Defendants call Dr. Robert Fay.  
                                                              1893

 1    ROBERT FAY, 

 2         called as a witness by the defendants, having

 3         been duly sworn, testified as follows: 

 4               THE COURT:  Would you state your full name and 
                                                                

 5    spell your last name for the record.  

 6               THE WITNESS:  Robert E. Fay, F A Y.  

 7               MR. MILLET:  Your Honor will find Dr. Fay's CV 

 8    under Tab 32.  

 9               THE COURT:  Thank you.  

10    DIRECT EXAMINATION 

11    BY MR. MILLET:

12         Q.    Dr. Fay, where are you employed?

13         A.    I am employed at the U.S. Bureau of the Census.

14         Q.    What is your position at the U.S. Bureau of the 

15    Census?

16         A.    I am the senior mathematical statistician.

17         Q.    Could you please describe to the Court your 

18    duties as a senior mathematical statistician?

19         A.    I report to the associate director for 

20    statistical design methodology and standards in the 

21    director's office.  I am responsible to investigate and 

22    contribute to the program of mathematical statistics over 

23    the entire range of the census bureau.

24         Q.    Could you summarize for the Court your 

25    educational background, please.
                                                              1894

 1         A.    Yes.  I received a Bachelor's in mathematics and 

 2    a Master's and Ph.D and statistics all from the University 

 3    of Chicago.

 4         Q.    Could you summarize for the Court your employment 

 5    history at the time that you became senior mathematical 

 6    statistician at the census bureau?

 7         A.    I first worked at the census bureau after 

 8    graduate school starting in 1974 and switched divisions in 

 9    1978 from the statistical research division to the 

10    statistical methods division, and then was promoted to the 

11    senior mathematical statistician in 1990.

12         Q.    Dr. Fay, are you a member of any professional 

13    societies?

14         A.    Yes, I am.

15         Q.    Could you please describe them to the Court?

16         A.    My most important affiliation is with the 

17    American Statistical Association, but my CV also lists my 

18    attachment to the American Association for the Advancement 

19    of Science, the Institute of Mathematical Statistics, the 

20    Population Association, the Washington Statistical Society, 

21    the Biometric Society, and the International Association of 

22    Survey Statisticians.

23         Q.    Dr. Fay, you referred to your CV.  Are you 

24    referring now to Defendants' Exhibit 32?

25         A.    Yes, I am.
                                                              1895

 1         Q.    Dr. Fay, are you a fellow of the American 

 2    Statistical Association?

 3         A.    Yes, I am.

 4         Q.    When did you become a fellow of the ASA?

 5         A.    In 1987.

 6         Q.    Have you served as an editor of any professional 

 7    publications?

 8         A.    I have.

 9         Q.    Could you please describe them to the Court?

10         A.    I served as an associate editor for approximately 

11    three years in the theory and methods section of the Journal 

12    of the American Statistical Association.  I am also 

13    currently standing as an associate editor of the same 

14    journal for a special section of papers being prepared for 

15    the 1990 census undercount.  I additionally have just been 

16    asked in this area to be an associate editor on another 

17    journal by the editor, in survey methodology.

18         Q.    Dr. Fay, have you published any articles on 

19    issues relevant to the decennial census?

20         A.    Yes, I have.

21         Q.    Could you just describe the general subject 

22    matters of those articles?

23         A.    They concern the undercount.  I would mention 

24    that my dissertation was on demographic analysis, and its 

25    uncertainty back in 1974.  Additionally, I was one of three 
                                                              1896

 1    authors on the principal report about the evaluation of 

 2    undercoverage in the 1980 census.

 3         Q.    Have you also published any articles concerning 

 4    the topic of smoothing?

 5         A.    Yes, I have.

 6         Q.    Could you describe those in general terms?

 7         A.    They include an article in the Journal of the 

 8    American Statistical Association listed on page 3 that 

 9    appears under "Principal Publications" that described an 

10    application of smoothing type procedures to estimate income 

11    for the 1970 census.  Additionally, I have published several 

12    other articles on the subject.

13         Q.    Have you also had occasion to publish articles on 

14    the topic of variance estimation?

15         A.    Yes, I have.

16         Q.    Could you summarize those for the Court?

17         A.    Think include some papers prepared in the early 

18    eighties and then in the late eighties on the theory of 

19    variance estimation for complex samples, and additionally 

20    some papers describing practical applications including a 

21    paper describing work on my program VPLX.

22         Q.    Could you explain what the VPLX program is to the 

23    Court?

24         A.    The VPLX program is a general program that I am 

25    working on to compute variances from complex samples such as 
                                                              1897

 1    used generally by the census bureau and other particularly 

 2    large-scale social survey organizations.

 3         Q.    Dr. Fay, in addition to your work at the census 

 4    bureau and your publications which you just summarized to 

 5    the Court, do you engage in any other kinds of statistical 

 6    work outside your duties at the census bureau?

 7         A.    I generally don't engage in too much consulting, 

 8    because I am primarily a civil servant.  However, I will 

 9    point to one thing that I have done.  On page 2 I note that 

10    I have consulted every election year since 1978 on the 

11    principal two-year elections for CBS News.  My function 

12    there is among the six to ten people who are making the 

13    recommendations or in fact decisions to tell Dan Rather what 

14    he may announce on the air as called victories for specific 

15    candidates.  

16               MR. MILLET:  Your Honor, at this time defendants 

17    would offer Defendants' Exhibit 32 into evidence.  

18               MR. SHERMAN:  No objection.  

19               THE COURT:  32 is admitted.  

20               (Defendants' Exhibit 32 for identification was 

21    received in evidence)

22               MR. MILLET:  At this time we would also tender 

23    Dr. Fay as an expert in the field of statistics and ask that 

24    he be allowed to express his opinion.  

25               MR. SHERMAN:  We have no objection, your Honor.  
                                                              1898

 1               THE COURT:  He is and may.  

 2               MR. MILLET:  Thank you, your Honor.

 3         Q.    Dr. Fay, in your opinion, is there an undercount 

 4    in the 1990 decennial census?

 5         A.    Yes, there is.

 6         Q.    Dr. Fay, in your opinion, is there a differential 

 7    undercount of minority groups in the 1990 decennial census?

 8         A.    Yes.

 9         Q.    In your work at the census bureau, did you have 

10    any responsibilities concerning the bureau's activities 

11    regarding the post-enumeration survey?

12         A.    Yes, I did.

13         Q.    Could you describe them briefly to the Court.

14         A.    In addition to other responsibilities, I had two 

15    primary functions.  One was I was a member of the 

16    nine-member Undercount Steering Committee in the census 

17    bureau who made a recommendation to the director about the 

18    adjustment decision.  Secondly, I was primarily responsible 

19    for both developing the software and overseeing its 

20    implementation to compute the design-based variances for the 

21    raw adjustment factors, and additionally the variances and 

22    covariances used in all the evaluation studies of the VPLX.

23         Q.    We have heard testimony most recently by Mr. 

24    Bounpane about the functions of the Undercount Steering 

25    Committee, and I am not going to repeat those for the Court. 
                                                              1899

 1               Let me simply ask you, Dr. Fay, did there come a 

 2    time when the Undercount Steering Committee made a 

 3    recommendation to the director of the census bureau on 

 4    adjusting the 1990 census?

 5         A.    Yes, we did.

 6         Q.    What was that recommendation?

 7         A.    The recommendation, by vote of 7 to 2, was to 

 8    recommend adjustment to the director of the 1990 census 

 9    using the production estimates for that purpose.

10         Q.    Did you agree with that recommendation?

11         A.    Yes, I did.

12         Q.    Approximately when was that recommendation, Dr. 

13    Fay?

14         A.    I recall mid June.

15         Q.    Can you explain to the Court why you agreed with 

16    that recommendation to the director to adjust the 1909 

17    sends?

18         A.    Yes.  I tried at the time to balance the evidence 

19    both in favor and against the adjustment and to trade off 

20    the case for and against, to consider the evidence that had 

21    been presented to the Undercount Steering Committee.

22         Q.    At the time of the Undercount Steering 

23    Committee's recommendation to the director, had all of the 

24    evaluations of the PES been completed?

25         A.    Essentially, as designed, yes.
                                                              1900

 1         Q.    In your view, Dr. Fay, did the evaluations that 

 2    were completed address all the outstanding issues in your 

 3    mind?

 4         A.    No, they didn't.

 5         Q.    What were the outstanding issues in your mind 

 6    that remained open at the time of the Undercount Steering 

 7    Committee's recommendation to the director?

 8         A.    The most important one to me at the time was the 

 9    considerable uncertainty in the smoothing itself.  I was 

10    concerned that what we thought to be the level of accuracy 

11    for the smoothing might, in fact, be just too optimistic.  

12    Additionally, I was concerned that we had not done really 

13    enough work to understand several other problems, one 

14    particularly being the effect of the homogeneity assumption 

15    on the adjustment geographically.

16         Q.    Did the Undercount Steering Committee advise the 

17    director of this uncertainty concerning the estimated 

18    variances which you just referred to?

19         A.    Yes, we did.

20         Q.    Could you explain to the Court how you were able 

21    to make a recommendation to the director in the presence of 

22    this uncertainty concerning the variances?

23         A.    I felt that I could express in my vote an overall 

24    judgment while nonetheless reporting the sources of concern 

25    available to me at the time, that way informing the director 
                                                              1901

 1    of the outstanding issues.

 2         Q.    You referred to the process of smoothing, and 

 3    there has been some testimony about that already.  Could you 

 4    just in brief and general terms explain to the Court your 

 5    understanding of the smoothing process?

 6         A.    Just very generally, your Honor, we started with 

 7    sample data that had been computed for 1392 poststrata 

 8    through very traditional survey means.  These estimates, as 

 9    we anticipated, had high sampling variability.  In other 

10    words, our conventional measures, the sampling variability 

11    suggested there was a relatively high level of error in our 

12    estimates.  

13               Following our general plan set in advance, we 

14    applied a smoothing model which took these raw estimates 

15    and, by combining them with statistical theory using the 

16    model and under a certain set of assumptions, could produce 

17    different numbers, smooth numbers that we expected to have 

18    considerably smaller sampling variance even though we might 

19    introduce some level of bias which we hoped to be 

20    acceptable.

21         Q.    Did you also become familiar with the process of 

22    crease moving during your work on the Undercount Steering 

23    Committee?

24         A.    Yes, I did.

25         Q.    Could you explain that, again, in general terms 
                                                              1902

 1    to the Court?

 2         A.    In very general terms, the presmoothing concerned 

 3    the issue of the variances themselves, the variances of the 

 4    raw adjustment factors.  These variances are measures of 

 5    their uncertainty when we estimate each poststratum 

 6    separately without using a model.  

 7               These variances, themselves being estimates, 

 8    tended to be quite variable.  I particularly believe that 

 9    this might have a harmful effect on our smoothing.  We 

10    attempted to do something like a smoothing to the variance 

11    estimates themselves by fitting what is called a linear 

12    regression model, another kind of statistical model.

13         Q.    You just referred to the variances of the raw 

14    adjustment factors, Dr. Fay.  Could those variances in the 

15    raw adjustment factors be measured or estimated in some 

16    fashion?

17         A.    We could produce estimates based on our sample 

18    design and using our sample data.  The theory for this 

19    estimation of uncertainty is relatively well established.  

20    There is, in fact, a complete convention from the 

21    design-base perspective on how to estimate variance for 

22    certain kinds of sample totals, and for more complex 

23    statistics like the raw adjustment factors there are 

24    different methods that tend to give quite similar results.

25         Q.    After this smoothing and presmoothing process, 
                                                              1903

 1    Dr. Fay, did the smoothed adjustment factors still have 

 2    variance?

 3         A.    Yes, they did. 

 4         Q.    Could the variance of the smoothed adjustment 

 5    factors be measured in the same way that you just described 

 6    as for the raw adjustment factors?

 7         A.    Not in the same way.

 8         Q.    Can you describe how the variances of the 

 9    smoothed adjustment factors could be estimated?

10         A.    The variances of the smoothed adjustment factors 

11    was an important part of this process, but it really 

12    required in fact the same statistical model that produced 

13    the smoothing itself.  In other words, we used statistical 

14    assumptions and statistical modeling to come up with 

15    estimates of the variability of these smoothed factors, 

16    which were, in fact, the most important product of this 

17    entire PES.

18         Q.    Dr. Fay, does the variability of the smoothed 

19    adjustment factors that you have just referred to have any 

20    effect on the population estimates derived using those 

21    smoothed adjustment factors?

22         A.    Yes, it does.

23         Q.    Could you explain what that effect is, please?

24         A.    Just as our raw adjustment factors start with a 

25    fairly large degree of sampling variability, we were hoping 
                                                              1904

 1    to obtain smoothed factors with much smaller variability.  

 2    Nonetheless, they would vary over different repetitions of 

 3    the PES sample.  Had we started all over again and drew 

 4    simply a different random sample, we would have gotten 

 5    different results for our smoothed estimates.  

 6               Any use of these smoothed estimates then, in 

 7    fact, would have variance.  For example, any estimate of a 

 8    state's population would have some degree of sampling 

 9    variance.

10         Q.    Dr. Fay, I am going to show you what has been 

11    marked and I believe received as evidence as Plaintiffs' 

12    Exhibit 28.  

13               MR. MILLET:  Your Honor, I want to see if perhaps 

14    we can graphically illustrate what Dr. Fay just had to say.

15         Q.    Dr. Fay, if you could look at the last page of 

16    Exhibit 28.  

17               MR. MILLET:  Your Honor, I am referring to the 

18    bar graph.  I believe it is the bar graph that Professor 

19    Fienberg referred to during his testimony the other day.  

20               Are we all on the same page?

21         Q.    Dr. Fay, looking at the last page of Plaintiffs' 

22    Exhibit 28 and the bar graph, can you explain to the Court 

23    what the effect of this variability that you were just 

24    describing has on the population estimates as shown by this 

25    chart?
                                                              1905

 1         A.    The chart, in fact, shows the estimates as little 

 2    white circles inside these bars.  The little circles 

 3    represent the amount of adjustment that would be made to 

 4    each of those states by the process.  

 5               The width of the bar, from the bottom to the top, 

 6    was dependent on the variances we computed under this model 

 7    for the smoothed factors.  In other words, given our 

 8    assumptions, we assumed that these intervals would be -- let 

 9    me explain what what these intervals attempt to describe.  

10               The point here is still a sample estimate, and it 

11    is still uncertain.  But we hope that the interval that we 

12    construct by this process would cover the true value in 95 

13    percent of the samples.  We can't say in our way of thinking 

14    that we know that the probability is 95 percent that the 

15    true point is within this interval, but in a somewhat 

16    convoluted way we can say that under 95 percent of the 

17    samples we would have drawn, we would have covered the true 

18    value somewhere in that interval.

19         Q.    Dr. Fay, if the variances of the smoothed 

20    adjustment factors were increased, what effect would that 

21    have on the bar graph that we are just looking at here on 

22    Plaintiffs' Exhibit 28?

23         A.    Increases in the variances for the smoothing 

24    would have the effect of moving the width of these bars up 

25    in both directions under the assumption we have made, so 
                                                              1906

 1    that the bars would become wider.

 2         Q.    Dr. Fay, at the time of the Undercount Steering 

 3    Committee's report to the director, did you attempt to 

 4    quantify the variance of the smoothed adjustment factors?

 5         A.    At the time of the decision, having looked at 

 6    several pieces of evidence and expressing my concerns about 

 7    pieces of the uncertainty left out of the variance 

 8    computation, I formed a judgmental interval based only on my 

 9    judgment and not hard or specific calculation, that perhaps 

10    the variances were understated by a factor of 1.7 to 3.0 for 

11    poststratum groups or large aggregates such as states.  That 

12    would have the effect, speaking approximately, of increasing 

13    the width of these bars by factors of perhaps 1.3 to 1.7, 

14    approximately.

15         Q.    Dr. Bounpane just described that process as being 

16    judgmental.  Was there any uncertainty in your estimate of 

17    the variance of the smoothed adjustment factors at the time 

18    you made your recommendation to the Undercount Steering 

19    Committee?

20         A.    At the time I only reported an interval expressed 

21    considerable uncertainty to begin with.  But I realized that 

22    even this interval might either be too high or too low and 

23    had very little direct evidence about whether I was correct.

24         Q.    What, if anything,, Dr. Fay, did you do as a 

25    result of this uncertainty in measuring the variance in 
                                                              1907

 1    smoothed adjustment factors concerning the Undercount 

 2    Steering Committee's report to the director?

 3         A.    At the time of the vote I looked at this concern 

 4    and weighted against the rather considerable evidence still 

 5    in support of the adjustment.  Trying to take this account, 

 6    although I described this process to the fellow members of 

 7    the Undercount Steering Committee, I roughly guessed that we 

 8    would still have a margin of safety to allow us or allow 

 9    myself to make the recommendation.

10         Q.    Did you make any allowance for this uncertainty 

11    in the drafting of the Undercount Steering Committee's 

12    report to the director?

13         A.    We specifically added a technical appendix, I 

14    believe number 6, to the report to describe the evidence and 

15    the concerns.

16         Q.    Dr. Fay, I would like to show you what has been 

17    marked and received as evidence as Plaintiff's Exhibit 54.  

18    It is also part of Defendants' Exhibit 1.  Dr. Fay, does 

19    that contain the Undercount Steering Committee's report to 

20    the director?

21         A.    Yes, I believe it does.

22         Q.    Can you identify for us the portion of that 

23    report that contains the appendix to which you were just 

24    referring?

25         A.    Yes.  The Appendix 6 is the last appendix to the 
                                                              1908

 1    report and begins on Bates stamp 950.

 2         Q.    Following the Undercount Steering Committee's 

 3    report to the director, did you participate in any meetings 

 4    with the Secretary concerning the question of adjusting the 

 5    1990 decennial census?

 6         A.    Yes, I did.

 7         Q.    Was there one meeting or more than one meeting?

 8         A.    I recall only one meeting.

 9         Q.    Why did you attend that meeting?

10         A.    I was asked to by the director.

11         Q.    Do you know why the director asked you to attend 

12    that meeting?

13         A.    The director told me that she chose me to 

14    represent, along with Bob Groves, five remaining members of 

15    the Undercount Steering Committee who had voted in the 

16    affirmative.

17         Q.    Can you explain to the Court what you did at that 

18    meeting?

19         A.    At the meeting I both reported my reasons for 

20    supporting my decision, and I also answered his questions 

21    and questions of his staff about several aspects of the 

22    recommendation.

23         Q.    Did the Secretary address any questions to you 

24    concerning whether reasonable statisticians can differ on 

25    the question of adjusting the census?
                                                              1909

 1         A.    Yes, he did.

 2         Q.    Can you relate to the Court what the Secretary 

 3    said to you and your response?

 4         A.    Sorry.  I told the Secretary that I believed, 

 5    yes, reasonable statisticians could differ on this 

 6    conclusion.

 7         Q.    During this meeting did you discuss the estimated 

 8    variances for the smoothed adjustment factors?

 9         A.    Yes, that subject came up.

10         Q.    Can you explain what you said?

11         A.    I tried to again recount and summarize in a 

12    nontechnical manner the concerns that led to this memorandum 

13    and the way it was measured and allowed that it did soften 

14    considerably the case for adjustment.

15         Q.    Following the Secretary's decision on adjustment 

16    in this case, did you undertake any research concerning the 

17    uncertainty in the final estimates resulting in the smoothed 

18    adjustment factors?

19         A.    Yes, I did.

20         Q.    Why did you undertake that research?

21         A.    I really had three reasons to do so.  First of 

22    all, the appendix, which was primarily written at my 

23    initiative and primarily by me, left a very great veil of 

24    uncertainty over the census bureau's own recommendation.  So 

25    I felt, in fact, some personal responsibility to see if I 
                                                              1910

 1    could clarify this matter further.  

 2               Secondly, I wished to take advantage of, in fact, 

 3    the special section of the Journal of the American 

 4    Statistical Association on the undercount issue.  It was, in 

 5    fact, a professional opportunity for me to explore this 

 6    issue directly and to present my findings to my peers.  

 7               In addition, I suspected now that these issues, 

 8    although probably much less formed, were reasonably 

 9    outstanding issues in at least some other applications of 

10    smoothing.  I was interested in the general implications 

11    beyond the implications it might have for the specific 

12    adjustment question.  

13               MR. SHERMAN:  Your Honor, I would like to object 

14    and we move to strike the last answer.  It goes to post-July 

15    15th work.  It is irrelevant to the issue in the case.  

16               MR. MILLET:  Your Honor, this is a subject motion 

17    in limine that has been denied.  

18               THE COURT:  Overruled.

19         Q.    Dr. Fay, were you requested to undertake the 

20    research that you just described by anyone at the commerce 

21    department?

22         A.    No, I wasn't.

23         Q.    Were you requested to undertake that research by 

24    anyone at the justice department?

25         A.    No.
                                                              1911

 1         Q.    Can you approximate for the Court the amount of 

 2    time that this research has taken so far, at least since the 

 3    Secretary's decision?

 4         A.    Trying to convert the time to a full-time basis, 

 5    perhaps two to three months on this specific topic itself.  

 6    But additionally, the calculations here required general 

 7    advances in, or were facilitated, I should say, by general 

 8    advances that I made in the variance program VPLX, the 

 9    computer acronym I made up.  This may have taken a month or 

10    two prior to the time I started the research.

11         Q.    Can you describe the part of the research that 

12    you have undertaken since the Secretary's decision on the 

13    question of uncertainty in the final estimates of the 

14    smoothed adjustment factors?

15         A.    Very generally, I began with, in fact, the list 

16    of concerns that have been laid out in the technical 

17    appendix.  Except I decided that the point I could -- in 

18    fact, those concerns --

19         Q.    You are referring now to Appendix 6?

20         A.    I'm sorry.  In Appendix 6 the concerns are 

21    somewhat described or are described in pages 4 and 5.  The 

22    appendix lists four concerns, but I decided that the four 

23    for research purposes really were three and collapsed the 

24    fourth into the third.  

25               When I attempted my paper, in fact wrote an 
                                                              1912

 1    abstract for the paper before doing the research, I 

 2    attempted to address each of those to the degree I could 

 3    allowed by the time that the journal required the paper.

 4         Q.    You just referred to the paper, Dr. Fay.  Could 

 5    you take a look at what has been marked as Defendants' 

 6    Exhibit 33.  

 7               MR. MILLET:  Your Honor, that is at Tab 33 of the 

 8    binder.

 9         Q.    Is that a copy of the paper to which you were 

10    just referring?

11         A.    Yes.  This is a manuscript which I submitted to 

12    the special section of the Journal of the American 

13    Statistical Association I referred to earlier.

14         Q.    Approximately when did you prepare Defendants' 

15    Exhibit 33?

16         A.    During the month of January, and dated the draft 

17    January 21, 1992.

18         Q.    At the time you prepared Defendants' Exhibit 33, 

19    did it represent your views as to the proper estimates of 

20    the variances of the smoothed adjustment factors?

21         A.    It did express all of my findings up to the point 

22    that I prepared the paper.

23         Q.    Could you summarize for the Court what your 

24    findings were at that time?  

25               MR. SHERMAN:  Your Honor, we would object to this 
                                                              1913

 1    line of questioning for several reasons.  First, this is, as 

 2    Dr. Fay has received, a draft prepared for a journal, in 

 3    accordance with my other objection also opposed to July 

 4    15th.  So I think it should be irrelevant for that reason.  

 5    But it is not a bureau document.  

 6               THE COURT:  It is his own document.  

 7               MR. SHERMAN:  It is his own document.  It is a 

 8    draft.  

 9               THE COURT:  So what?  He is now testifying to 

10    these findings.  The fact that he refers to the draft as 

11    opposed to a final document -- 

12               MR. MILLET:  In fact, your Honor, I hope Dr. Fay 

13    will testify to the findings that he came to after this 

14    draft.  

15               THE COURT:  The objection is overruled.

16         Q.    I believe the question was, Dr. Fay, what were 

17    your views concerning the proper estimate of the variances 

18    of the smoothed adjustment factors at the time you wrote 

19    Defendants' Exhibit 33?

20         A.    The paper reports evaluations that I made up to 

21    that time of the three assumptions or three problems that I 

22    had identified in the Appendix 6 previously in the 

23    Undercount Steering Committee's report.  That there was a 

24    potential problem was confirmed by the calculations I did in 

25    the paper, although I found that the problems occurred in 
                                                              1914

 1    somewhat different areas than my initial judgment had let 

 2    led me to believe.  

 3               There were three, as I mentioned, three topics.  

 4    The first of them, briefly, concerned the fact that even 

 5    though we had presmoothed variances, we had used what I 

 6    might call raw covariance estimates, that is, directly 

 7    estimating covariances and variances to produce raw 

 8    correlations or measures of association between various raw 

 9    adjustment factors, and actually used these estimates in 

10    deriving the smoothing model.  

11               In effect, we had a very, very large number of 

12    estimates, in the tens of thousands, included in the 

13    calculation.  But we had made no accounting for the effect 

14    of variability of all these sample estimates in the 

15    smoothing.  

16               I did a complex calculation using the same 

17    computer facilities as we had done our initial processing 

18    for the PES and, in effect, drew several samples of the PES 

19    from itself to try to assess, through a process called a 

20    bootstrap, the additional uncertainty introduced by 

21    estimating tens of thousands of correlations and then 

22    treating them as if those estimates were known.  

23               By looking at the new and larger variability that 

24    came out comparing estimates that we derived from each of 

25    those separate calculations done compared to what we had 
                                                              1915

 1    assumed if we treated them all as known, I found a very 

 2    large increase in probability, approximately a factor of 2.

 3         Q.    Dr. Fay, I think perhaps you used what may be the 

 4    most colorful statistical term so far in the case.  Could 

 5    you explain to the Court what you mean by bootstrap?

 6         A.    A bootstrap has been used in other parts of the 

 7    PES and perhaps has appeared before.  

 8               MR. ZIMROTH:  Your Honor, I object.  One of our 

 9    witnesses used the term "jackknife."  

10               MR. MILLET:  That is much too hostile, your 

11    Honor.  

12               THE COURT:  They are not the same thing, are 

13    they?  

14               MR. MILLET:  If you would like, Dr. Fay can talk 

15    about both.  

16               THE COURT:  We are speaking of jackknives and 

17    bootstraps.  

18               THE WITNESS:  Your Honor, I did them both.  The 

19    production variances themselves were computed through a 

20    jackknifing, in fact a stratified jackknifing approach.  The 

21    variances, in effect, were computed of the raw adjustment 

22    factors, and these were critical inputs into our smoothing, 

23    were derived by taking one of the 5,000-plus PES block 

24    clusters and deleting it from the estimates, and thereby, in 

25    effect, seeing how it affected all 1392 raw adjustment 
                                                              1916

 1    factors; then, in turn, leaving one of these individual 

 2    components out, and in fact doing approximately 5,000-plus 

 3    estimates and looking at the variability.  That, in fact, 

 4    was the manner in which we computed our production variances 

 5    that we had used in the July decision.  I already have done 

 6    jackknifing.  

 7               On top of that, to assess the variability of 

 8    these estimates, I applied a bootstrapping procedure.  We 

 9    had also used some bootstrapping in other parts of the PES.  

10    But in order to look at the variability from estimating 

11    these correlations, I created new samples from the PES by, 

12    in effect, drawing with replacement a sample from the PES 

13    blocks, so that in this new bootstrap sample a given PES 

14    block might be omitted or used once or twice or three times 

15    at random, perhaps more.  In fact, indeed, sometimes more.  

16               This technique is another way to assess 

17    variability and allowed me to look at the approximate effect 

18    of the estimation of the covariances on the entire process.

19         Q.    In keeping with our colorful allusions, Dr. Fay, 

20    can I direct your attention to the last paragraph of the 

21    text of your paper in Defendants' Exhibit 33, at page 21.  

22    You make a reference there to gothic architecture.  If you 

23    would, please, I would like you to explain to the Court what 

24    you meant by that paragraph.  

25               MR. SHERMAN:  Your Honor, I would object to this.  
                                                              1917

 1    I don't think Dr. Fay has been qualified as an expert in 

 2    gothic architecture.  

 3               THE COURT:  I am sure so far as it appears on 

 4    statistics, he is.  I will permit it.

 5         Q.    If you would, please, Dr. Fay, explain the last 

 6    paragraph.

 7         A.    Let me summarize the last paragraph, which is in 

 8    fact a large metaphor.  In fact, what I describe is an 

 9    historic process by which architects in the Middle Ages 

10    were, with a new architectural form, reaching higher and 

11    higher limits in what they could do.  In fact, I describe a 

12    specific instance in which they simply built too far, and 

13    the cathedral itself collapsed, at least the roof as 

14    intended collapsed, so that a reduced cathedral was built 

15    thereafter.  

16               THE COURT:  An open-air one.

17         A.    In fact, experimentally, this set the limits of 

18    gothic architecture which other architects followed after 

19    that point.

20         Q.    Did you intend to draw any analogy to the post- 

21    enumeration survey with that paragraph?

22         A.    Actually, I made several, but I think it would be 

23    helpful if I would explain them.  

24               If one, the smoothing model was, in my opinion, 

25    the most complex form of empirical base estimation involving 
                                                              1918

 1    many aspects that had been built up by a number of people to 

 2    try to address a very complex problem.  In fact, all these 

 3    very large pieces of or assumption-prone pieces of work were 

 4    being put together to try to produce an overall effect.  

 5               I actually think, in terms of learning from this 

 6    experience, it was much too complicated.  That is the 

 7    message I want to convey to my peers.  In effect, we are 

 8    looking at a process that took a sample size of 150, 170,000 

 9    households, depending on how you count them in our PES.  

10    Instead of using this data effectively to produce an 

11    estimate, it is almost as if we had thrown half the data 

12    away and then tried to use the remaining half to make an 

13    estimate.  In other words, by making a very bad estimate, we 

14    had done much less than we set out to do.

15         Q.    Dr. Fay, since writing Defendants' Exhibit 33, 

16    have you conducted any further research on this issue?

17         A.    Yes, I have.

18         Q.    Could you please describe that research.

19         A.    I have done primarily two additional 

20    calculations.  One was that one of the sections which 

21    describes the uncertainty due to modeling or choosing 

22    variables to best subset selection, I had only written that 

23    section by referencing the general design of the smoothing 

24    process and doing analytic approximations to say what I 

25    thought the effect might be approximately and to suggest 
                                                              1919

 1    very approximately its effect.  

 2               I have since been successfully able to use the 

 3    same bootstrap samples, and in an integrated way do the 

 4    bootstrapping combining the uncertainty in the covariance 

 5    matrices and the best subsets selection procedure.  Overall, 

 6    I get about the same result I attempted to report in this 

 7    draft, although a bit lower.  The comparison is about 1.5 I 

 8    suggested earlier, now seeing about 1.4 on the average.  But 

 9    given that I was trying to be in the ballpark, I was pleased 

10    by that finding.  

11               The second calculation, given that I had done a 

12    bootstrap, I wanted to at least test the robustness of 

13    sensitivity of my calculation to what I thought was a 

14    significant alteration of the calculation.  I will ask 

15    forgiveness for the jargon that follows.  I did what is 

16    called an infinitesimal bootstrap in which I altered the 

17    bootstrap samples by making them look much more like the 

18    original data, so that I in fact was not altering the 

19    covariation variances by nearly so much as I did initially.  

20    I get very similar findings to my full bootstrap, actually a 

21    bit higher increase in variance, which was a little 

22    different from what I expected.  But, again, very similar.

23         Q.    Again, Dr. Fay, directing your attention back to 

24    Plaintiffs' Exhibit 28, the last page, the bar graph, the 

25    effect of what you are discussing is what on this bar graph?
                                                              1920

 1         A.    Again, what I am doing is not moving any of these 

 2    mid points, but I am asking the question, how wide should 

 3    these bars be?  I am saying a significant increase in their 

 4    uncertainty.

 5         Q.    Are you continuing your research, Dr. Fay?

 6         A.    Yes, I am.

 7         Q.    Could you summarize briefly what research you are 

 8    still conducting?

 9         A.    In fact, the paper as written will see clearly 

10    addressed each problem separately.  But I am moving towards 

11    a combined calculation that will replace this graph with a 

12    different graph with wider bars that have been computed 

13    through just the source of uncertainty I have been able to 

14    look at.

15         Q.    Dr. Fay, based on your present research, do you 

16    still favor the adjustment recommended by the Undercount 

17    Steering Committee to the director?

18         A.    No, I don't.

19         Q.    Why not?

20         A.    On several grounds.  The paper, however, the 

21    findings of the paper I found most convincing.  I believe 

22    now that the adjustment can be justified under only some 

23    narrow grounds rather than the much wider grounds we had 

24    hoped, on a technical basis.  

25               For example, I now believe that the justification 
                                                              1921

 1    might fall on some specific choices of loss functions and 

 2    yet be not recommended under others or certainly not 

 3    significant and fail to find a convincing case.  

 4               Furthermore, I am engaged in further work, partly 

 5    related to this paper but partly independently, to find new 

 6    means of assessing the actual effect of the homogeneity 

 7    assumption on problems of this sort.  This homogeneity 

 8    assumption was not, for example, built into the -- or the 

 9    effect of the homogeneity assumption was not built into the 

10    width of the bars as you see them on the paper.

11         Q.    Based upon the data now available to you, Dr. 

12    Fay, would you favor an adjustment of the 1990 decennial 

13    census?

14         A.    Taking the question more broadly and allowing 

15    alternative ways of adjusting the census, I haven't made up 

16    my mind.  I am continuing to consider it.

17         Q.    Why aren't you sure?

18         A.    I am a member of the CAPE committee now 

19    considering adjustment of the base used to compute 

20    post-censal estimates.  We are looking earnestly at whether 

21    alternative adjustments of the census, avoiding the problems 

22    that I have described in my paper, might produce an 

23    acceptable means of accounting for undercount as we produce 

24    pose censal estimates.

25         Q.    Dr. Fay, at the meeting at which the Undercount 
                                                              1922

 1    Steering Committee cast its vote on the report to the 

 2    director, did you bring anything out of the ordinary with 

 3    you to that meeting?

 4         A.    At a key decision meeting I showed up, in fact, 

 5    in a T-shirt saying "Statistics is Never Having to Say 

 6    You're Certain."  The source of this was the American 

 7    Statistical Association.  I was expressing, and I believed 

 8    that other members of the committee understood, that at the 

 9    time I made the vote there was a considerable weight of 

10    uncertainty on which way I felt I should vote.  

11               MR. MILLET:  Thank you, Dr. Fay.  Your Honor, I 

12    have no further questions.  

13               THE COURT:  We will break for lunch now and 

14    resume cross at two o'clock.  

15               (Luncheon recess) 

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1923

 1               A F T E R N O O N     S E S S I O N.

 2                                      2:00 o'clock p.m.

 3    ROBERT FAY,               resumed. 

 4               THE COURT:  Mr. Sherman.  

 5               MR. SHERMAN:  Good afternoon, your Honor.  

 6    CROSS-EXAMINATION 

 7    BY MR. SHERMAN:

 8         Q.    Good afternoon, Dr. Fay.

 9         A.    Good afternoon, Mr. Sherman.

10         Q.    Dr. Fay, despite your T-shirt, the undercount 
                                                               

11    steering committee took its responsibility very seriously, 

12    didn't it?

13         A.    Yes, we did.

14         Q.    And the committee put a tremendous amount of time 

15    and effort into its decision last June, didn't it?

16         A.    Yes.

17         Q.    The committee took the guidelines very seriously, 

18    didn't it?

19         A.    Yes.

20         Q.    In fact, although the committee's actual mandate 

21    was only to make a decision about the accuracy of the census 

22    versus the PES, the committee translated that charge into 

23    one which included consideration of the guidelines, isn't 

24    that right?

25         A.    Yes, we did.
                                                              1924

 1         Q.    And the committee, therefore, determined that in 

 2    order to find the adjusted counts more accurate, the 

 3    evidence had to show convincingly that adjustment would make 

 4    the counts more accurate, isn't that right?

 5         A.    Yes.

 6         Q.    So the statement that the Bureau, "Preferred 

 7    adjusted counts if the actual enumeration was not proven 

 8    more accurate," is not true, is it?  

 9               MR. MILLET:  May the witness be told the document 

10    that the quote comes from?  

11               THE COURT:  What?  

12               MR. MILLET:  The document where the quote comes 

13    from?

14         Q.    Do you know where that quote comes from, Dr. Fay?

15         A.    I may have heard it this morning as I sat in 

16    testimony, so it may be from the Secretary's decision.

17         Q.    It's not correct, is it?

18         A.    I'm sorry, I would disagree with the statement.

19         Q.    The undercount steering committee didn't start 

20    with a pro-adjustment bias, did it?

21         A.    In my opinion, no.

22         Q.    In coming to its decision, the undercount 

23    steering committee considered whether the PES successfully 

24    controlled the sources of error that had been cited as 

25    problems in the 1980 PEP and other census tests throughout 
                                                              1925

 1    the eighties, isn't that right?

 2         A.    Yes.

 3         Q.    And the seven-two majority found that it had, 

 4    correct?

 5         A.    That it --  excuse me --  that it had --  I need 

 6    to finish your sentence.

 7         Q.    That it had successfully controlled those 

 8    problems.

 9         A.    I'm sorry.  Yes.

10         Q.    The committee considered whether the PES results 

11    reasonably agreed with results of demographic analysis, 

12    didn't it?

13         A.    We considered that.

14         Q.    And the seven-two majority found that it did, 

15    correct?

16         A.    Two a large extent, although there were comments 

17    that go about different --  presented about some 

18    differences.

19         Q.    In fact, the undercount steering committee 

20    considered all the evidence available in rendering its 

21    seven-two decision in favor of adjustment, didn't it?

22         A.    Certainly all the evidence that we could within 

23    the amount of time we had.

24         Q.    In your view, did the committee consider the 

25    views of all its members?
                                                              1926

 1         A.    Yes.

 2         Q.    The undercount steering committee never concluded 

 3    that census counts were more accurate than adjusted counts 

 4    at the block level, did it?

 5         A.    We didn't conclude that.

 6         Q.    At the time of its decision, the undercount 

 7    steering committee knew that research would be continuing?

 8         A.    Yes.

 9         Q.    And the committee recognized at the time of its 

10    decision that with time, adjustment methodologies could be 

11    revised in order to make adjustment counts even better, 

12    correct?

13         A.    Yes, we did.

14         Q.    The seven-two majority didn't find that a reason 

15    not to recommend adjustment, did it?

16         A.    That's right.

17         Q.    And in making its decision, the committee took 

18    account of the uncertainties in the adjustment procedures, 

19    correct?

20         A.    The uncertainties that we were aware of at the 

21    time we took our vote.

22         Q.    You heard Mr. Baunpane's testimony this morning 

23    or his cross-examination?

24         A.    Yes, I did.

25         Q.    Mr. Bounpane was one of the two dissenting votes 
                                                              1927

 1    from the undercount steering committee, correct?

 2         A.    Yes.

 3         Q.    Have you seen a July 10 memorandum written by Mr. 

 4    Bounpane to the Secretary of Commerce?

 5         A.    I believe I know the one you are referring to.  I 

 6    saw one about that date, yes.

 7         Q.    You don't agree with Mr. Bounpane that the 

 8    differences he cited there between demographic analysis and 

 9    the PES weren't expected, do you?

10         A.    I, I took exception, in fact, in writing, to the 

11    use of "unexpected" in that at least many of these 

12    differences had appeared before.

13         Q.    And you don't agree with Mr. Bounpane that 

14    balancing is a significant problem with the 1990 PES, do 

15    you?

16         A.    I believe that we had considered most of the 

17    balancing questions.

18         Q.    It is your view, isn't it, that modest departures 

19    from the synthetic assumptions don't prevent adjustment from 

20    improving upon the census?

21         A.    That would depend upon "modest."  It is a factor 

22    and it concerned us, as I said.

23               This synthetic assumption would not have to hold 

24    perfectly in order for adjustment to be successful.

25         Q.    A few inconsistent results in face validity tests 
                                                              1928

 1    are not unexpected, are they, Dr. Fay?

 2         A.    A few, again, taking a few, is it a handful?  We 

 3    would not expect in a large data set all impressions to be 

 4    consistent with what we expected.

 5         Q.    Face validity tests are a rough check on results, 

 6    correct?

 7         A.    Yes.  

 8               MR. SHERMAN:  May I approach the witness, your 

 9    Honor?  

10               THE COURT:  Yes.

11               (Handing to the witness)

12               (Pause)

13         Q.    Dr. Fay, I have handed you a copy of PX 141, 

14    which I believe has already been admitted into evidence.

15               I just want you to confirm for the court that 

16    this is a memorandum that you wrote to the Secretary of 

17    Commerce.

18         A.    Yes, it is.

19         Q.    You wrote this after the July 8 meeting with the 

20    Secretary, isn't that right?

21         A.    Yes.

22         Q.    Dr. Fay, did you testify this morning or this 

23    afternoon, early this afternoon that at the July 8 meeting 

24    with the Secretary, you told the Secretary of Commerce that 

25    reasonable statisticians could disagree in this decision?
                                                              1929

 1         A.    Yes.

 2         Q.    Dr. Fay, I have handed you a copy of PX 41, which 

 3    has previously been admitted into evidence in this case.

 4               These are notes prepared of the Secretary's 

 5    meeting on July 8.

 6               Have you seen these before?

 7         A.    I believe I have, yes, at least once.

 8         Q.    Do you know who prepared these?

 9         A.    I'm sorry, I forget it.  It may have been 

10    prepared --  I'm sorry, I forget.

11         Q.    Did you see these in any other form other than 

12    the form you see it now?

13         A.    I did see a draft of --  I believe I saw a draft 

14    of this material.

15         Q.    Did you have comments on the draft?

16         A.    Yes.

17         Q.    Dr. Fay, I wonder if you could flip through this 

18    document and tell me if you see anywhere recorded in there 

19    the statement that you testified earlier you made to the 

20    Secretary of Commerce at this meeting on July 8.  

21               MR. MILLET:  Objection, your Honor.  It is a 

22    reading exercise.  The document speaks for itself.  

23               THE COURT:  Is it in there?  

24               MR. SHERMAN:  No.  

25               THE COURT:  All right.  That saved a couple of 
                                                              1930

 1    minutes.

 2         Q.    Dr. Fay, I have handed you a copy of Plaintiff's 

 3    Exhibit 244.

 4               Have you ever seen this document before?

 5         A.    I don't recognize it.

 6         Q.    Let me tell you, sir, that these purport to be 

 7    notes of the July 8 meeting of the Secretary of Commerce as 

 8    well.

 9               I wonder if you would flip through these, or if 

10    the court wishes I can make the same representation --  

11               THE COURT:  I think that is the easiest way to do 

12    it.  

13               MR. SHERMAN:  I will make a representation that 

14    in these notes there is also no reference to the statement 

15    that Dr. Fay testified to earlier.  

16               THE COURT:  Thank you.  

17               MR. SHERMAN:  I would move 244.  

18               MR. MILLET:  No objection.  

19               THE COURT:  244 is admitted.

20               (Plaintiff's Exhibit 244 marked for 

21    identification was received in evidence.) 

22    BY MR. SHERMAN:

23         Q.    Dr. Fay, you testified that you are familiar with 

24    the Bureau's smoothing procedures, correct?

25         A.    Yes.
                                                              1931

 1         Q.    In fact, the Bureau relied upon your expertise in 

 2    developing the smoothing model, didn't it?

 3         A.    Yes, to a degree.

 4         Q.    You testified that you wrote the BPLX program?

 5         A.    Yes.

 6         Q.    And you are familiar with what is known as 

 7    presmoothing, aren't you?

 8         A.    Yes.

 9         Q.    And it is still your position, isn't it, Dr. Fay, 

10    that presmoothing provides a better variance estimate than 

11    not presmoothing?

12         A.    Yes.

13               Let me clarify your question that I believe that 

14    for purposes of the smoothing that it was better to 

15    presmooth than to leave the variances alone.

16         Q.    And the smoothing and presmoothing operations 

17    were almost completely prespecified, weren't they?

18         A.    Close to completely.

19         Q.    You did testify this morning that presmoothing 

20    proceeded from a relatively well-established theory, 

21    correct?

22         A.    Yes.

23         Q.    And that it was almost a convention from a 

24    designed based perspective, is that correct?

25         A.    I'm sorry, that's --  I don't believe --  are you 
                                                              1932

 1    asking me if I said that this morning?

 2         Q.    Did you testify to that this morning?

 3         A.    In effect, I don't recall that statement, and I'm 

 4    speaking of presmoothing in the abstract before it becomes 

 5    part of a smoothing procedure.

 6         Q.    You don't think that presmoothing destroys the 

 7    connection between adjustment factors and the variances, do 

 8    you?

 9         A.    It, it could alter them, and there appears to be 

10    a wide open area for future research.

11         Q.    You still do believe that presmoothing was the 

12    right procedure to use if you were going to smooth, correct?

13         A.    I --  it was better than doing nothing.  I don't, 

14    in fact, think it is the right procedure any more.

15         Q.    Is it still your position that presmoothing 

16    provides a better variance estimate than not presmoothing?

17         A.    For use in the smoothing, yes.

18         Q.    Is smoothing an indispensable part of any 

19    adjustment procedure?

20         A.    No.

21         Q.    You don't think that in order to adjust the 

22    census, statistical smoothing models must be used?

23         A.    I'm sorry, in any --  in a very general context, 

24    I don't feel that that is a necessary requirement.

25         Q.    Whether smoothing is necessary or not depends on 
                                                              1933

 1    the sample size, correct?

 2         A.    Yes.

 3         Q.    You are not sure whether smoothing would have 

 4    been necessary with the sample size of 300,000 households, 

 5    are you?

 6         A.    It would have, in fact, depended on the design.  

 7    In fact, to answer your question, we are now trying to 

 8    adjust the census for postcensal estimations not using 

 9    smoothing and using the same sample size we had, but with a 

10    different design, so certainly with a different design and 

11    300,000, that fact I expect would be doable.  I'm not sure 

12    with the 1,392 design whether it was doable or not.

13         Q.    So you are not sure?

14         A.    I'm not sure.

15         Q.    You testified about the amount of variance in the 

16    PES, correct?

17         A.    Yes.

18         Q.    And the PES sample size of 300,000 households 

19    would have at least cut that variance in half, is that 

20    correct?

21         A.    Would have cut the sampling variance in half.

22         Q.    Dr. Fay, did the Bureau have any experience prior 

23    to 1990 with controlling sampling error?

24         A.    Yes.

25         Q.    And, in fact, the Bureau's decisions about 
                                                              1934

 1    smoothing were based at least in part on those experiences, 

 2    weren't they?

 3         A.    The decisions on smoothing followed from the size 

 4    of the variances of the raw factors.

 5         Q.    And were they also based in part on the Bureau's 

 6    previous experience in testing of smoothing?

 7         A.    It was a previous theoretical experience with 

 8    smoothing as a general abstract quantity.

 9         Q.    Is it your testimony that the Bureau had never 

10    tested smoothing before?

11         A.    No, it's not.

12         Q.    In fact, they had tested it in the 1988 dress 

13    rehearsal?

14         A.    Yes.

15         Q.    And in the 1986 TARO, correct?

16         A.    I recall that less well.  I'm not certain today.

17         Q.    Are you aware of what model the Bureau used to 

18    smooth the 1986 TARO?

19         A.    No.  Those, those results, if they do exist, are 

20    really prior to my very active involvement in this program.

21         Q.    Dr. Fay, is the model used by the Bureau in 1990 

22    the same as the Ericksen/Kadane model proposed for 

23    adjustment in 1980?

24         A.    It does have some general similarities to that 

25    model.
                                                              1935

 1         Q.    There are important differences between the 1980 

 2    model and the 1990 model, correct?

 3         A.    Yes, there are.

 4         Q.    In your view, Dr. Fay, should those differences 

 5    be obvious to any reasonably competent statistician?

 6         A.    Yes.

 7         Q.    If someone didn't understand those differences, 

 8    they would not understand the 1990 smoothing procedures very 

 9    well, would they?

10         A.    I'm sorry, your question is just if a 

11    statistician could not study the 1980 smoothing and the 

12    current one and read both, both documentations, that then 

13    they might not understand the 1990 very well? 

14               I'm sorry, is that your question?

15         Q.    Actually, I think you answered my question in the 

16    previous answer.  Thank you.

17               In your view, Dr. Fay, did the Bureau conduct a 

18    PES as a stage prop for the views advocated by Drs. Ericksen 

19    and Tukey?

20         A.    I, I, I have no idea of the context.

21               We, we --  I've told you before, we worked very 

22    seriously on this project and tried to use our own best 

23    professional judgment.

24         Q.    The Bureau attempted to measure bias in the PES, 

25    correct?
                                                              1936

 1         A.    Yes.

 2         Q.    The Bureau didn't propose that the biases should 

 3    be subtracted from the PES in doing the 1990 adjustment, did 

 4    it?

 5         A.    In fact, that wasn't technically feasible to do 

 6    that within the adjustment itself in the timetable.

 7         Q.    Dr. Fay, I would like to call your attention to 

 8    the chart that is on the easel over here.  It is Defendant's 

 9    Exhibit 68.

10               If it would be helpful, I can give you a version.

11         A.    Yes, please.

12               (Handing to the witness)

13               (Pause)

14         Q.    Have you had a chance to review it?

15         A.    Briefly, yes.

16         Q.    You have seen this before, haven't you, Dr. Fay?

17         A.    I believe at my deposition.

18         Q.    And calling your attention to the equations three 

19    lines from the bottom of the page, you have seen that 

20    before, haven't you?

21         A.    Because there is so much notation, I'm sorry, I'm 

22    taking my time.

23         Q.    That's all right.

24         A.    This wasn't among the primary documents I worked 

25    with before.
                                                              1937

 1               Just to set the context, I'm looking to see that 

 2    the sample Y is described in terms of its sampling variance, 

 3    okay, and we are also considering an error term in the 

 4    regression v, which appears to be proportional to the same 

 5    covariance matrix.

 6         Q.    I promised the reporter I would try to get 

 7    through this without actually doing the notations, so let's 

 8    see if we can.

 9               This is not, drawing your attention to what you 

10    just described, the assumption there is not the assumption 

11    used by the Bureau in the 1990 smoothing, is it?

12         A.    That's correct.

13         Q.    This was an alternative proposed by Dr. David 

14    Freedman, wasn't it?

15         A.    I think now that I understand what it is, yes, 

16    that's my understanding.

17         Q.    And you understood the alternative that Dr. 

18    Freedman proposes, don't you?

19         A.    I now do, yes.

20         Q.    The practical effect of this alternative would be 

21    to pull all the raw adjustment factors the same distance 

22    toward the regression line, correct?

23         A.    Yes, same proportional distance.

24         Q.    So it would tend to pull estimates with small 

25    variances too much toward the regression line, wouldn't it?
                                                              1938

 1         A.    Well, if the model is correct, that's what it's 

 2    supposed to do.

 3         Q.    And it would tend to pull estimates with large 

 4    variances too little toward the regression line, correct?

 5         A.    Well, under this model, that's, in fact --  I'm 

 6    sorry, the --  when you say too little, that's apparently 

 7    relative to some other standard.  Is that how I should take 

 8    that?

 9         Q.    In your view, Dr. Fay, is that a preferable 

10    alternative to the assumption used by the Bureau in the 1990 

11    smoothing?

12         A.    No, it's not preferable.

13         Q.    Did Dr. Freedman ever contact the Bureau to 

14    propose this alternative when the Bureau was considering how 

15    to do it's smoothing?  

16               MR. MILLET:  Objection, it calls for speculation.  

17    I also object to the characterization that it is being 

18    proposed by Professor Freedman.  There is no evidence of 

19    that.  

20               MR. SHERMAN:  He has testified to it.  

21               MR. MILLET:  Not in this proceeding, nor 

22    has he --  

23               THE COURT:  I will let it stand.

24               You may answer.

25         Q.    Do you remember my question, Dr. Fay? 
                                                              1939

 1               The question is, did Dr. Freedman ever contact 

 2    the Bureau and propose this alternative when the Bureau was 

 3    contemplating the smoothing operation?

 4         A.    He did not.

 5         Q.    Has Dr. Freedman ever spoken to you about this 

 6    alternative?

 7         A.    I pause, because we have discussed some of his 

 8    work.

 9               I don't recollect discussions of this specific 

10    model, so the best of my recollection, no.

11         Q.    He does know what your position at the Bureau is, 

12    doesn't he?

13         A.    I believe so.  

14               MR. MILLET:  Objection.

15         Q.    Dr. Fay, the Bureau conducted a range of loss 

16    functions to test the accuracy of the census against 

17    adjustment, correct?

18         A.    Yes.

19         Q.    Did that range include numeric accuracy loss 

20    function?

21         A.    My recollection is it did.

22         Q.    And to be sure the judge understands, by numeric 

23    accuracy, we are talking about loss functions that measure 

24    whether the census or adjustment is closer to the estimated 

25    truth in terms of population totals rather than population 
                                                              1940

 1    shares, correct?

 2         A.    Yes.

 3         Q.    The results of those numeric accuracy loss 

 4    functions were presented to the undercount steering 

 5    committee?

 6         A.    My recollection is, yes, they were.

 7         Q.    They were very one-sided in favor of adjustment, 

 8    weren't they?

 9         A.    The numeric ones were, yes.

10         Q.    Do you know if the numeric accuracy loss 

11    functions were included in the administrative record?  

12               MR. MILLET:  Objection now, your Honor.  It is 

13    calling for speculation on the part of the witness and we 

14    are getting well beyond the part of the direct.  

15               THE COURT:  If he is that familiar with the 

16    administrative record, I'm sure he can answer.

17               I will let it stand.

18         A.    I can say I don't know.  

19               THE COURT:  Okay, that's enough.

20         Q.    Counting up pluses and minuses is not the only 

21    way to interpret the Bureau's loss function analysis for 

22    states, is it?

23         A.    No.

24         Q.    In fact, you personally would have preferred to 

25    report the Bureau's best estimates of how many states would 
                                                              1941

 1    be made more accurate kdistributively and how many made less 

 2    accurate, isn't that right?

 3         A.    That is, I believe I recall saying on deposition 

 4    that I would have been interested in knowing how many more 

 5    states or an estimate of how many states were improved.

 6         Q.    Do you recall saying that you would have 

 7    preferred to report that?

 8         A.    Yes.

 9         Q.    You testified earlier that you now believe that 

10    the adjustment decision may fall upon a specific choice of 

11    loss functions, correct?

12         A.    Yes.

13         Q.    Last July when the Secretary made his decision, 

14    that wasn't true, was it?  

15               MR. MILLET:  Objection.  That he testified to?  

16    The question is not clear.  

17               MR. SHERMAN:  I am asking whether what he 

18    testified to was true last July, that is, that adjustment 

19    decision would fall on a specific choice of loss functions.

20         A.    I'm sorry, the evidence that was considered by 

21    the undercount steering committee included different types 

22    of loss functions and was generally consistent between those 

23    presented.

24         Q.    And when you say consistent, they all favored 

25    adjustment, didn't they?
                                                              1942

 1         A.    Yes, as presented to the committee.

 2         Q.    Dr. Fay, you attempted to do some hypothesis 

 3    tests last June at Dr. Darby's request, correct?

 4         A.    Yes.

 5         Q.    Dr. Darby asked for those hypothesis tests after 

 6    he had seen results from the Bureau's loss function 

 7    analysis, correct?  

 8               MR. MILLET:  Objection.  That calls for 

 9    speculation.  

10               THE COURT:  If you know.

11         A.    The timing would have been such that I would --  

12    I believe I recollect that he did review loss function 

13    analysis and that he spoke to us and that that was, in fact, 

14    the basis for his criticism of the absence of the hypothesis 

15    test.

16         Q.    You attempted to test the hypothesis that 

17    measured undercounts from different states were actually 

18    different from the national average undercount, correct?

19         A.    Yes, I did.

20         Q.    Was that what Dr. Darby requested?

21         A.    It's as close I could do to what he requested in 

22    the time.

23         Q.    Did you test the hypothesis that the observed 

24    results of the Bureau's loss function analysis in favor of 

25    adjustment was due to random error?
                                                              1943

 1         A.    I obtained a set of mixed results, in fact, and 

 2    it seemed to depend somewhat on the assumptions.

 3               In fact, I then felt at the time that had further 

 4    weakened the case for adjustment.

 5         Q.    What I am asking you, Dr. Fay, is whether you 

 6    tested the hypothesis that the observed results of the loss 

 7    function analysis was due to random error?

 8         A.    I attempted to do that, I'm sorry, yes.

 9         Q.    Dr. Fay, you testified about your uncertainty at 

10    the time of the undercount steering committee's decision 

11    last June, correct?

12         A.    Yes.

13         Q.    And you testified that you don't today favor the 

14    adjustment recommended to the Secretary, correct?

15         A.    Yes.

16         Q.    And that is because you are now more uncertain 

17    than you were last June, correct?

18         A.    I would say it's more correct to say that I'm 

19    more certain that the evidence weighs against the 

20    adjustment.

21         Q.    When you say the evidence weighs against, you are 

22    reconsidering your vote in the context of what you take to 

23    be your obligation under the guidelines, correct?

24         A.    Well, not in addition to what I think is actually 

25    reasonable of and, in fact, my technical obligations 
                                                              1944

 1    generally even in the absence of the guideline.

 2         Q.    But that would include a presumption in favor of 

 3    the census, correct?

 4         A.    Yes.

 5         Q.    You didn't testify on direct, did you, Dr. Fay, 

 6    that you now believe the census count is more accurate than 

 7    the adjustment count?

 8         A.    No, I did not.

 9         Q.    Even with your uncertainty, when you voted for 

10    adjustment last June, you were somewhere around 95 percent 

11    confident that adjustment was more accurate than the census, 

12    correct?

13         A.    Actually, what I --  I believe at the time when I 

14    was looking at incorrect results, my uncertainty would have 

15    been about that level, I suppose.

16         Q.    Didn't you testify at your deposition that you 

17    would want a 95 percent degree of confidence in order to 

18    vote for adjustment?

19         A.    I believe that's right.

20         Q.    Dr. Fay, how many times have you met or spoken 

21    with lawyers from the Justice or Commerce Department since 

22    your vote last June?  

23               MR. MILLET:  Objection.  What is the relevance of 

24    that? 

25               I hope he talked to me before this proceeding.  
                                                              1945

 1               THE COURT:  I will sustain that objection.

 2         Q.    Dr. Fay, did you work with the lawyers on this 

 3    case prior to your vote last June?  

 4               THE COURT:  Did he work with the lawyers?  What 

 5    does that mean?

 6         Q.    Did you provide any advice to the lawyers in this 

 7    case?  

 8               MR. MILLET:  Objection and I will interpose the 

 9    attorney-client objection as well on that question.  

10               MR. SHERMAN:  I am not asking for the nature of 

11    the advice.  

12               THE COURT:  I understand the distinction.

13               Your objection is overruled.

14         A.    I'm sorry, to any of my recollection, I didn't 

15    know any of the attorneys from the Justice Department 

16    working on the case --  

17               THE COURT:  Before? 

18               THE WITNESS:  I'm sorry, before the vote in June.

19         Q.    In 1988, Dr. Fay, after this lawsuit was filed, 

20    did you provide advice to lawyers --

21         A.    Oh, excuse me, yes.

22         Q.    You recall now?

23         A.    That was in a --  

24               MR. MILLET:  Objection to the extent he is asking 

25    for the nature of that advice.  
                                                              1946

 1               MR. SHERMAN:  I didn't ask.  

 2               THE COURT:  He didn't ask.  

 3               MR. MILLET:  I believe Dr. Fay was about to state 

 4    what he stated.  

 5               THE COURT:  Well, don't do that.

 6         Q.    At that time, Dr. Fay, you were helping defend 

 7    the Commerce Department's decision to take the adjustment 

 8    decision away from the Bureau, correct?

 9         A.    Can I just set the context of when that was that 

10    I believe in 1988 there was litigation leading up to the 

11    stipulation agreement under which we worked.  

12               MR. MILLET:  Your Honor, we are getting into the 

13    substance of communications with counsel at this point.  

14               THE COURT:  Not really.  He is just narrating 

15    when he spoke to counsel.

16         A.    Forgive me.  At that point I was involved with 

17    counsel.  I forgot that was all one case.  I'm sorry.

18         Q.    At that time you were involved with counsel even 

19    though you had mixed feelings about the decision, isn't that 

20    right?

21         A.    Yes.

22         Q.    Even though you weren't sure what the 

23    consequences of the decision would be in terms of whether 

24    the question of adjustment would be addressed objectively, 

25    correct?
                                                              1947

 1         A.    Just repeat the question once again, please?  

 2               THE COURT:  I don't understand it, either.

 3               MR. SHERMAN:  All right.  Let me rephrase it, 

 4    your Honor.

 5         Q.    Do you remember testifying at your deposition 

 6    with regard to the decision of the Commerce Department to 

 7    take the adjustment decision away from the Bureau, you 

 8    weren't sure of the consequences of that decision in terms 

 9    of whether the question of adjustment would be addressed 

10    objectively?  

11               MR. MILLET:  I am going to object to going into 

12    the transcript.  All the witness asked is for the question 

13    to be restated, not for counsel to go into deposition 

14    transcript pages, and if we are going to do that I think in 

15    fairness to Dr. Fay he ought to get a copy to look at what 

16    his testimony was.  

17               MR. SHERMAN:  I am asking him whether he 

18    testified to that.  

19               MR. MILLET:  There is no foundation for asking 

20    that.  

21               THE COURT:  I will permit it.

22               Do you understand the question, Dr. Fay? 

23               THE WITNESS:  Do I remember testifying in my 

24    deposition that I wasn't sure of the consequences of --

25         Q.    In terms of whether the adjustment would be 
                                                              1948

 1    addressed objectively.

 2         A.    If there is a transcript I could consult, I will 

 3    check that.

 4         Q.    When Commerce took the decision from the Census 

 5    Bureau, did you have concerns about whether the adjustment 

 6    decision would be addressed objectively?

 7         A.    I'm sorry.  This is about 1987?

 8         Q.    Whenever the decision was taken from the Census 

 9    Bureau, did it create for you concerns about whether the 

10    adjustment decisions would be addressed objectively?

11         A.    Yes, I, I would have had some concerns at that 

12    time.

13         Q.    Did you work with the government lawyers on the 

14    lawsuit over adjustment in 1980, Dr. Fay?

15         A.    Yes.

16         Q.    You testified in support of the government's 

17    decision not to adjust in 1980, correct?

18         A.    In the, in the court in November 1980, I, I 

19    testified to the inability to guarantee the success of an 

20    adjustment.

21         Q.    Were you opposed to adjustment in 1980?

22         A.    I hadn't even seen the numbers.  I had mixed 

23    feelings about it.

24         Q.    In your deposition, you did say it was your 

25    opinion in the period 1983 to 84 that adjustment might be 
                                                              1949

 1    possible, correct?

 2         A.    Yes.

 3         Q.    Let me just say I have the chronology correct.

 4               In 1980 you testified for the government against 

 5    adjustment, although you hadn't seen the data.

 6               In 1983 and 84, you thought adjustment might be 

 7    possible.

 8               In 1988 you worked with the government against 

 9    this lawsuit.

10               In 1991 you voted for adjustment and in 1992 you 

11    are testifying for the government against adjustment.

12               Dr. Fay, either you change your mind a lot or 

13    once the government makes a decision, you support it.  

14               MR. MILLET:  Objection, your Honor.

15         Q.    Which is it?  

16               MR. MILLET:  That is argumentative.  

17               THE COURT:  You are badgering him now.

18               Sustained. 

19         Q.    Do I have the chronology right, Dr. Fay?  

20               MR. MILLET:  Objection.  I think he misstated the 

21    testimony.  

22               THE COURT:  The question is, did he state the 

23    chronology correctly? 

24         A.    Mr. Sherman, I believe the dates are correct.  

25               THE COURT:  What you did on those dates has not 
                                                              1950

 1    been accurately described? 

 2               THE WITNESS:  I would like to redescribe it.  

 3    That's right.  

 4               THE COURT:  Just leave it that way.  

 5               MR. SHERMAN:  All right.

 6         Q.    Dr. Fay, you testified earlier about your draft 

 7    paper, correct?

 8         A.    Yes.

 9         Q.    You have gotten comments from other statisticians 

10    on your draft, haven't you?

11         A.    Yes.

12         Q.    Some of the comments have been quite critical of 

13    your analysis, haven't they?

14         A.    Really, I'd only say perhaps one, although what 

15    was unusual is the comments seemed most critical of a very 

16    standard design base for estimating variance and that's what 

17    took me most by surprise.

18         Q.    Who was that comment from?

19         A.    That's from know Cressie, University of --  I'm 

20    sorry.

21         Q.    Did you also receive comments from Donald Rubin 

22    and Alan Zaslausky of Harvard University?

23         A.    Yes.

24         Q.    Didn't they suggest that another method or 

25    analysis would result in lower variances than those you 
                                                              1951

 1    calculated?

 2         A.    Yes.  What they suggested is that had a different 

 3    procedure, a Bases procedure applied to the PES using 

 4    different estimates, we might have had lower variances.

 5    

 6    

 7    

 8    

 9               (Continued on the next page) 

10    

11    

12    

13    

14    

15    

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1952

 1         Q.    Dr. Fay, the estimated variance numbers that you 

 2    gave us earlier, were those at the poststratum group level?

 3         A.    I was reporting general reports at the 
                                                        

 4    poststratum group level on state.

 5         Q.    And state as well?

 6         A.    And state as well.

 7         Q.    Have you changed your estimate of the state level 

 8    variance since I took your deposition?

 9         A.    Yes.  I have refined it a bit.

10         Q.    The ranges of uncertainty that you testified 

11    about, Dr. Fay, they are not calculated ranges, are they?

12         A.    The ranges are still judgmental.

13         Q.    The only calculated variances are your point 

14    estimates, correct?

15         A.    Yes.

16         Q.    Your analysis does not use the actual variance 

17    calculated by the bureau for the production smoothing, does 

18    it?

19         A.    It uses something very similar but slightly 

20    different, yes.

21         Q.    You attempt to reconfigure the variance, correct?

22         A.    Yes, in order to do a valid analysis.

23         Q.    So your analysis depends on whether the 

24    reconfigured variance is analogous to the original estimated 

25    variance, correct?
                                                              1953

 1         A.    Yes.

 2         Q.    With apologies to the Court, bootstrapping and 

 3    jackknifing are not exactly the same, are they?

 4         A.    They both involve replication, but they are 

 5    different specific techniques.

 6         Q.    One of the subjects of your draft paper is the 

 7    overfitting of variables, correct?

 8         A.    Yes.

 9         Q.    Is there any other evidence available about 

10    possible overfitting of variables?

11         A.    That's a very broad question.  There is research 

12    in the area generally.  I have heard other scholars discuss 

13    it.  There were frequent findings that the smoothing 

14    estimates often appeared very close to the to the raw 

15    factors.  It's a form of evidence.

16         Q.    Is there any other study that you know of about 

17    possible overfitting of variables with regard to the 1990 

18    PES?

19         A.    In fact, I recall that work was done by David 

20    Hoaglin, trying to look at alternative approaches and to the 

21    specific model selection procedure used or to determine a 

22    better one.

23         Q.    Did you take that evidence into account when you 

24    did your January 21st draft?

25         A.    It didn't seem very directly related to the 
                                                              1954

 1    question I was considering, so I didn't study it much 

 2    further than having heard their presentation -- I'm sorry.  

 3    David Hoaglin's presentation.

 4         Q.    Their study did have to do with overfitting of 

 5    variables, right?

 6         A.    That's right, yes.

 7         Q.    One of the things you studied in your paper was 

 8    overfitting of variables, correct?

 9         A.    But I specifically studied, in fact, the effects 

10    of overfitting of variables on the measures of uncertainty.  

11    I can't recall that their results had -- that they had any 

12    results on that.

13         Q.    You never bothered to read the Hoaglin report, 

14    did you?

15         A.    No.

16         Q.    That was no?

17         A.    Yes, that was no.

18         Q.    Dr. Fay, your January 21st paper does not 

19    conclude that the raw adjustment factor should not have been 

20    smoothed, does it?

21         A.    No.

22         Q.    Your draft paper does not conclude that the 

23    possible increased variance makes the adjusted numbers less 

24    accurate than the census, does it?

25         A.    That's right.
                                                              1955

 1         Q.    I would like to refer you again to the bar graph 

 2    which is part of PX-28.  Do you have it?

 3         A.    Yes.

 4         Q.    Dr. Fay, the points on this graph don't change, 

 5    do they?

 6         A.    I'm sorry?  Yes.

 7         Q.    As a result of your research, the points on this 

 8    graph will not change, will they?

 9         A.    The focus of my research is not on the points on 

10    this graph.

11         Q.    The focus of your research is just on the 

12    uncertainty around the points, correct?

13         A.    Yes.  

14               MR. SHERMAN:  We have nothing else at this time, 

15    your Honor.  

16               THE COURT:  Any redirect?  

17               MR. MILLET:  Yes, your Honor, if I may.  

18    REDIRECT EXAMINATION 
                           

19    BY MR. MILLET:

20         Q.    Dr. Fay, Mr. Sherman asked you a moment ago about 

21    Dr. Hoaglin's study in overfitting variables.  How did you 

22    become familiar with Dr. Hoaglin's study on that issue?

23         A.    I did skim an earlier report of his analyzing the 

24    regression variables themselves.  Secondly, I attended a 

25    seminar at which he presented and summarized his results.
                                                              1956

 1         Q.    Mr. Sherman also asked you whether your original 

 2    opinion on adjusting the 1990 decennial census as a member 

 3    of the Undercount Steering Committee was within a 95 percent 

 4    confidence interval, and you referred to relying on some 

 5    incorrect results, I believe is how you termed it.  Could 

 6    you explain to the Court what those incorrect results were.

 7         A.    When the Undercount Steering Committee did take 

 8    its vote, we were working, in fact, with substantially wrong 

 9    information in the loss function analysis.  What had 

10    happened is a mistake had been made in setting up the 

11    analysis.  That mistake made the results much more favorable 

12    towards adjustment than they should have been.

13         Q.    Could you explain that mistake in general terms 

14    to the Court?

15         A.    In fact, the loss function analysis in a general 

16    way was constructing a target for purposes of comparison and 

17    then comparing the census to that target and also comparing 

18    how different the target and a hypothetical PES sample 

19    estimate would be, seeing whether one was closer than the 

20    other.  

21               The regrettable feature of the way this analysis 

22    was set up was that the census was being charged for errors 

23    in constructing the target, which was in fact based on the 

24    PES production smoothing, although indirectly, but that 

25    target contained as much error as already in the PES.  In 
                                                              1957

 1    the original loss function analysis, that was being charged 

 2    against the census.  

 3               Once the loss function analysis was reviewed both 

 4    by Dr. Darby, who saw the error, and then the analysis, the 

 5    formal paper about the analysis under peer review -- the 

 6    error was reported, as I understand it, by a referee -- the 

 7    two pieces of information coming together about the same 

 8    time, he realized that even after we had taken the 7 to 2 

 9    vote, we needed to redo our analysis because it was 

10    incorrect.

11         Q.    So Under Secretary Darby was one of the people 

12    who pointed out the error in the census bureau's analysis?

13         A.    Yes.

14         Q.    Under Secretary Darby was in the commerce 

15    department, correct?

16         A.    Yes.

17         Q.    Mr. Sherman asked you about Dr. Darby's request 

18    for hypothesis testing.  Dr. Fay, as you sit here today, is 

19    it your understanding that null hypothesis testing is 

20    consistent with guideline 1?

21         A.    Yes, it is.

22         Q.    Dr. Fay, taking a look at what I will call the 

23    Freedman alternative, Defendants' Exhibit 68, the V sub 3 

24    term -- that is the term that varies from the smoothing 

25    model, isn't that correct?
                                                              1958

 1         A.    Yes, that's right.

 2         Q.    Dr. Fay, you indicated to Mr. Sherman that you 

 3    did not find that alternative to be preferable.  Is that an 

 4    alternative that you would consider to be worthy of further 

 5    study?

 6         A.    Yes, I think that it is an interesting 

 7    alternative in that it shows something else that might 

 8    happen.  In fact, there are some natural arguments to make 

 9    that if, for example, children tend to live with their 

10    mothers and that we see correlations occur between our raw 

11    adjustment factors for mothers and the sampling errors of 

12    how we think mothers are missed and how we think children 

13    are missed, then maybe that, in fact, belongs in the model 

14    term as well.  If the model fails to predict the omissions 

15    of, say, mothers, it may fail to predict in the same 

16    direction the degree to which we missed their children.  So 

17    I think it is an interesting alternative.

18         Q.    Dr. Fay, would you take a look at Plaintiffs' 

19    Exhibit 41, which are the notes of the meeting of July 8th.  

20    I would like you to take a look at the first page, the 

21    second paragraph, second sentence, which I will quote.  

22    "He," meaning the Secretary, "stated that he has an open 

23    mind on the issues and that it is clear that reasonable 

24    persons can disagree on the question of adjustment."  

25               Dr. Fay, does that indicate to you that there was 
                                                              1959

 1    a discussion with the Secretary concerning the question of 

 2    whether reasonable persons can disagree on the question of 

 3    adjustment?  

 4               MR. SHERMAN:  Objection, your Honor.  

 5               THE COURT:  Sustained.  

 6               MR. MILLET:  I have no further questions, your 

 7    Honor.  

 8    RECROSS-EXAMINATION 
                          

 9    BY MR. SHERMAN:

10         Q.    Dr. Fay, after the correction to the loss 

11    function that Mr. Millet had asked you about, you had a 

12    chance to reconsider your vote on the question of adjustment 

13    last June, didn't you?

14         A.    I had some potential to do that.

15         Q.    In fact, the entire Undercount Steering Committee 

16    had a chance to reconsider its vote, didn't it?

17         A.    Yes.

18         Q.    Did anyone change their vote?

19         A.    No.  The vote was left.  

20               MR. SHERMAN:  Thank you very much.  

21               THE COURT:  Thank you very much, Dr. Fay.  

22               (Witness excused) 

23               THE COURT:  We will take five minutes.  

24               (Recess)

25               MR. GLASS:  Your Honor, we call Mr. Paul Meier. 
                                                              1960

 1    PAUL MEIER, 

 2         called as a witness by the defendants, having

 3         been duly sworn, testified as follows: 

 4               THE COURT:  Would you state your full name and 

 5    spell your last name for the record.  

 6               THE WITNESS:  My name is Paul Meier spelled 

 7    M E I E R.  

 8    DIRECT EXAMINATION 
                         

 9    BY MR. GLASS:

10         Q.    Mr. Meier, let me start by referring you and the 

11    Court to Defendants' Exhibit 34.  That is in the book in 

12    front of you there.  Let me ask you if Defendants' Exhibit 

13    34 is a copy of your curriculum vitae.

14         A.    It is.

15         Q.    As is indicated on page 1 of your CV, you are a 

16    professor of statistics at the University of Chicago, are 

17    you not?

18         A.    I am.

19         Q.    Do you hold other appointments at the University 

20    of Chicago?

21         A.    I hold several.  I think Chicago has a bit more 

22    of a habit of assigning several titles.  I am professor of 

23    statistics in the department of pharmacological and 

24    physiological sciences in the department of medicine, and I 

25    am, I believe it is, lecturer in law.
                                                              1961

 1         Q.    As an aside, Mr. Meier, what is the preferred 

 2    mode of address among the faculty members at the University 

 3    of Chicago?

 4         A.    There is an old tradition at Chicago.  When it 

 5    was founded, all people had doctorates, so it became the 

 6    habit to use "Doctor" for MDs and "Mister" for everybody 

 7    else.  If it is all the same to you, "Mister" is fine.

 8         Q.    Thank you.  So the record is clear, you do hold 

 9    advanced degrees, do you not, including a doctorate in 

10    mathematics?

11         A.    Of course, yes.

12         Q.    Your advanced degrees are described on page 1 of 

13    your CV?

14         A.    That's right.

15         Q.    Your CV lists at page 5 a number of honors that 

16    you have received.  Have you been the recipient of any other 

17    major honors since the date of your CV?

18         A.    Yes.  I was recently elected to the Institute of 

19    Medicine, which is an affiliate of the National Academy and 

20    should not be confused.  Is not like election to the 

21    academy, but it is nonetheless an honor of some substance.

22         Q.    At pages 2 through 5 of your curriculum vitae you 

23    list 44 different consultant activities and special 

24    appointments that you have held.  I am wondering if you 

25    could give the Court a thumbnail sketch of the types of 
                                                              1962

 1    matters in which you have served as a consultant.

 2         A.    I have never counted them.  I thank you for that.  

 3               THE COURT:  I wouldn't rely on it.  

 4               THE WITNESS:  Then you certainly wouldn't rely on 

 5    me.  

 6               MR. GLASS:  We can always smooth the answer, your 

 7    Honor.

 8         A.    In any event, my consulting really began soon 

 9    after I got my Ph.D and was at the Johns Hopkins University.  

10    I was asked to consult on statistical problems in the 

11    pharmaceutical industry.  That involved both problems in 

12    design and analysis of laboratory studies and was really my 

13    first introduction to the similar questions, design and 

14    analysis of clinical trials, which has since become a major 

15    concern of mine.  

16               I also came to consult in matters having to do 

17    with scientific sampling that arose out of legal problems 

18    and have done quite a bit of that.  

19               I have consulted in the form of participating in 

20    study sections, that is to say, review committees of the 

21    National Institutes of Health.  Those study sections review 

22    proposals for grants and score those proposals in order of 

23    priority.  

24               More recently, I have done quite a bit of 

25    consulting with the Food and Drug Administration in 
                                                              1963

 1    connection with their problems in evaluating new drug 

 2    applications to see what ought to be approved for marketing.

 3         Q.    In connection with your activities as a 

 4    consultant, have you had occasion to appear as an expert 

 5    witness?

 6         A.    Yes, on a fair number of occasions.

 7         Q.    In what types of matters have you appeared as an 

 8    expert?

 9         A.    My first experience as expert witness was in the 

10    sampling domain shortly after I got to Chicago.  I guess 

11    because someone had told me to teach the sampling course.  

12    That was a new one for me, but those people didn't know.  

13    They called me up and asked if I would consult on sampling 

14    in connection with railroad rate cases, at that time all 

15    held before a hearing examiner of the Interstate Commerce 

16    Commission.  Finance docket 31503, which I will never 

17    forget, was my first experience in sampling, and that went 

18    on for three and a half years before it was all over.  

19               I was thereafter called upon in many cases of 

20    similar sort, where railroad traffic had to be sampled and 

21    analysis of the results presented.  

22               Later I became an expert witness in a real court 

23    of law.  One of those had to do with statistical evidence of 

24    fraud in chemical analyses presented in a contract dispute.  

25    Another had to do with possible bias in drawing the district 
                                                              1964

 1    lines for voting in the City of Chicago.  

 2               Then, I think because my friend Fred Mosteller 

 3    was kind enough to suggest my name, I got drawn into 

 4    statistical testimony in equal employment opportunity cases, 

 5    and I have done quite a bit of testifying as an expert, 

 6    sometimes for plaintiffs, more often for defendants, in that 

 7    domain.

 8         Q.    Have you ever appeared as an expert witness 

 9    against the federal government?

10         A.    Indeed, I have.  Most, not all, but most of the 

11    defendants' cases in the discrimination area were with, of 

12    course, the EEOC on the other side.

13         Q.    Your CV lists at pages 5 through 10, and again I 

14    have taken the time to reenumerate all of this, 71 articles 

15    that you have written, another 7 reports and documents, and 

16    a number of works in progress.  Could you describe for the 

17    Court generally the types of areas in which you have 

18    written?

19         A.    There is the work that I began early on 

20    statistical estimation theory, particularly problems having 

21    to do with incomplete observations, and I have written 

22    several papers in that domain.  There is a great deal of 

23    applied work dealing with clinical trials, the area in which 

24    I have a very special interest.  Then there is a fair amount 

25    written -- well, there is some written on probabilistic 
                                                              1965

 1    models in biology, physiology, and a fair amount written on 

 2    the problems of statistical testimony in law.

 3         Q.    Thank you.  

 4               MR. GLASS:  Your Honor, we would ask for the 

 5    admission of Mr. Meier's CV in Defendant's Exhibit 34.  

 6               MR. SOLOMON:  No objection.  

 7               THE COURT:  It is admitted. 

 8               (Defendant's Exhibit 34 for identification was 

 9    received in evidence)

10               MR. GLASS:  And we would offer Mr. Meier as an 

11    expert in statistics.  

12               MR. SOLOMON:  But for the areas where he has 

13    already disclaimed expertise, which are in the deposition, 

14    we don't have an objection.  If we get into those areas, I 

15    will object at the time.  Is that satisfactory?  

16               THE COURT:  Is it all right by you, Mr. Glass?  

17               MR. GLASS:  I don't recall any areas where 

18    expertise was disclaimed, but maybe we can take that as it 

19    comes.  

20               THE COURT:  With that understanding, Mr. Meier is 

21    qualified and may express his opinions.  

22               MR. GLASS:  Thank you, your Honor.

23         Q.    Mr. Meier, is this case your first involvement 

24    with the census?

25         A.    No, it is not.  I was first involved with the 
                                                              1966

 1    census very early in my career in the 1950s, when I was 

 2    asked by the Institute of Mathematical Statistics to be 

 3    their representative on the Census Advisory Committee.  That 

 4    must have had to do, I guess, with the 1960, planning for 

 5    the 1960 census.  

 6               Our domain of activity there was largely fussing 

 7    with, A, the propriety of certain kinds of questions -- as I 

 8    recall, that may have been the first time a religion 

 9    question was either put on or at least discussed -- and 

10    small area statistics and what kind of thing you should do 

11    about them.  I can't say we were crowned with any great 

12    success in that, but that was my first involvement with the 

13    census.  

14               In the controversy about the 1980 census and the 

15    problem of adjustment, I was involved, but always 

16    informally.  I was invited to read everything.  And it may 

17    have been, I won't swear to this, but I believe Barbara 

18    Bailar hoped that I might be willing to testify on behalf of 

19    the census bureau that adjustment was not appropriate.  That 

20    was too big a bite for me at the time, and I figured there 

21    were others quite qualified to do it and I decided I 

22    wouldn't do that.  I read quite a bit of the material that 

23    she sent to me, but I did not further participate in it.  

24               Now, of course, through the agency of my friend 

25    David Freedman, my attention has been drawn to the same 
                                                              1967

 1    problem, adjustment in the 1990 census, and this time I have 

 2    been induced to agree to testify.

 3         Q.    Based on your experience with the census and your 

 4    expertise in statistics, Mr. Meier, you have certain views, 

 5    do you not, on what a census should be?

 6         A.    Yes, I do.  Let me comment on that 

 7    systematically.  I take note that the census historically 

 8    had a role as a head count -- 

 9               MR. SOLOMON:  Pardon me, your Honor.  I see the 

10    witness is reading from something.  

11               THE WITNESS:  Right.  You are welcome to this.  

12               MR. SOLOMON:  I have no objection.  Is that what 

13    you were reading from at your deposition?  

14               THE WITNESS:  No, it is not.  It is essentially 

15    the same stuff, but I rewrote it.  

16               MR. GLASS:  Would you like to take a look.  

17               MR. SOLOMON:  I would like a copy, if that is all 

18    right.  Could we just make a copy?  

19               THE COURT:  Right now?  

20               MR. SOLOMON:  Yes.  

21               MR. GLASS:  Perhaps between Mr. Meier's direct 

22    and his cross it might be a little more appropriate.  

23               THE COURT:  Let's do it then, between direct and 

24    cross.

25         A.    I mentioned our census.  It begins with its 
                                                              1968

 1    historical role of a head count, of course adjusted for 

 2    slaves and so on, but the purpose being to allocate 

 3    representation.  It at that time was probably unthinkable 

 4    that it should be anything other than a direct head count 

 5    and you do the best you can and that's what you have got.  

 6               The census has developed a lot since then, and it 

 7    is apparently seen to a substantial extent as a portrait of 

 8    the American people.  I think language like that has been 

 9    used in describing it, and it has become used, of course, 

10    for a great many purposes.  

11               I want to point out all the way along the census 

12    has to a large extent been arbitrary as to what it is.  I 

13    point out the variation in whether to count and how to count 

14    and allocate servicemen overseas, debates about the 

15    residence of college students and how they should be 

16    allocated, the determination that illegals should be 

17    counted.  

18               In my discussions with census bureau personnel, 

19    it is not a matter of broad, general principle that 

20    everybody here should be counted -- after all, declared 

21    visitors are not counted -- but rather it would be so much 

22    harder not to count illegals than to count them, that the 

23    official decision seemed sensible that they should be 

24    counted.  

25               My point is that there is a great deal of 
                                                              1969

 1    arbitrariness.  Some of that is regrettable, but it would 

 2    seem to me beside the point to get overly fussed by the 

 3    arbitrariness in it.  The intent is to count the people as 

 4    best you can, and I think the census on the whole does a 

 5    pretty good job of that.  

 6               I point out, however, that with respect to all 

 7    these purposes, an entirely sampled census, that is to say 

 8    no attempt at an individual-by-individual head count, would 

 9    seem to be quite appropriate, and we don't do that and we 

10    don't propose to do that.  I think that is basically for two 

11    reasons.  

12               Number one, whether wisely or not, we feel we 

13    want good counts in very local areas, right down to the 

14    block or something close to it, what the census uses in 

15    place of that in rural areas.  That is the kind of problem 

16    which defeats sampling, except in very special cases where 

17    there are covariants that you can use to supplement the 

18    sampling, and that is not the case here.  If that is what 

19    you want, then you have to go count them.  

20               The other thing is that the census relies on the 

21    voluntary cooperation of the citizenry, and at least until 

22    recently I think we have seen the census as what my 

23    colleague Bill Kruskal calls a national ceremony.  It is 

24    something in which we rely upon the citizens to willingly 

25    cooperate in carrying through this process.  And, indeed, we 
                                                              1970

 1    have had in various censuses what I would call a benevolent 

 2    boosterism in which various agencies, civil rights groups, 

 3    and others have encouraged people to participate so that 

 4    they be fully counted and whatever benefits go with that 

 5    full counting should accrue.  

 6               It is my view that if the census becomes an 

 7    arcane and almost impenetrable calculated result rather than 

 8    a fairly direct count -- which I believe, despite imputation 

 9    and other things, it still is -- it seems to me that it is 

10    likely that we will begin to lose that citizen enthusiasm.  

11    Most especially, if we say we don't really need your 

12    cooperation because we can sample and adjust for your lack 

13    of cooperation, it seems to me the census is likely to 

14    degrade substantially over time.  

15               It is for all these reasons, quite independently 

16    of whether the proposed methods can somewhat improve the 

17    count, I think the entire enterprise is misguided.  

18               MR. SOLOMON:  Your Honor, I am going to move to 

19    strike the last answer.  I wanted to see the sheets of paper 

20    because there was testimony given at the deposition that the 

21    professor stated to me was not based on statistics.  I 

22    called it religion.  Maybe we can dispute about what it is.  

23    But it is not statistics.  What we have heard, I submit, is 

24    not statistics, and this witness has been qualified only in 

25    statistics.  
                                                              1971

 1               MR. GLASS:  Your Honor, we have had -- 

 2               THE COURT:  What is the harm?  Why are you 

 3    exercised?  

 4               MR. SOLOMON:  The harm?  

 5               THE COURT:  Yes.  

 6               MR. SOLOMON:  In this courtroom I had supposed 

 7    that before someone would get up and offer an opinion, they 

 8    should be qualified.  This man is qualified in statistics, 

 9    not in religion or sociology or boosterism.  

10               MR. GLASS:  Your Honor, we have one of the most 

11    eminent statisticians in the country here.  He has long 

12    experience and familiarity with the census.  This is an 

13    overview question.  It seems to me entirely appropriate that 

14    he be permitted to offer his opinion.  

15               MR. SOLOMON:  If he would stick to statistics.  

16               THE COURT:  The objection is overruled.  

17    Continue.

18    BY MR. GLASS:

19         Q.    Mr. Meier, you have also certain opinions, do you 

20    not, on the limitations of statistics and the resolution of 

21    controversies?

22         A.    Yes, I do have strongly stated opinions on that 

23    question.

24         Q.    Could you share those with the Court, please?  

25               MR. SOLOMON:  Objection, vague.  I cannot imagine 
                                                              1972

 1    what we are about to get.  

 2               THE COURT:  I can.  Overruled.

 3         A.    I think it is best to explain the issue here by 

 4    recognizing that statisticians can largely be divided into 

 5    two camps.  

 6               THE COURT:  Those who divide into camps and those 

 7    who don't.  

 8               THE WITNESS:  They are all pretty much one place 

 9    or the other, your Honor.

10         A.    There are those who take statistics as a powerful 

11    tool for analyzing data, pretty much any data.  Although in 

12    some situations the underlying assumptions that would 

13    justify statistical inference are either plainly false or at 

14    least not verifiable, nonetheless that group of 

15    statisticians, an honorable group of people, let me say, 

16    believe that illumination will be provided by carrying out 

17    such statistical analyses anyhow, and better that we should 

18    get whatever illumination such analyses can give us than to 

19    put the enterprise aside.  

20               The other group is concerned about the 

21    limitations in a very serious way.  In a number of important 

22    subdomains, such as scientific sampling, the assumptions 

23    needed to validate the statistical method are well 

24    established, and in those cases statistics is a powerful 

25    tool which that camp hopes the world will accept as a solid 
                                                              1973

 1    basis for inference.  That group of people worries extremely 

 2    about the degradation of the credibility of statistics by 

 3    applying it in situations where it is very liable to error 

 4    and where it is very hard to distinguish the likelihood of 

 5    error.  

 6               David Freedman, I believe, and certainly I belong 

 7    to the second camp.  My professor, John Tukey, and my 

 8    colleague or ex-colleague Steve Fienberg clearly belong to 

 9    the other camp.  That does define an important difference.  

10               It is my view that we have an example before us 

11    that we should seek not to emulate, and that is the 

12    testimony of psychiatrists in the court of law.  There we 

13    have a group of people with considerable ability and 

14    considerable potential for doing good for people but who 

15    have been willing to go beyond what many think their 

16    profession is able to do and to testify in court on subjects 

17    that their skills simply cannot effectively clarify.  

18               I have written on this subject, partly in a paper 

19    having to do with discrimination, what happened in 

20    Hazelwood, and more fully in an address that I made to the 

21    American Statistical Association which I titled "Damned 

22    Liars and Expert Witnesses."  

23               I guess that's my answer.

24         Q.    May I refer you to Defendants' Exhibits 80 and 

25    81, which are on the desk in front of you.  May I ask if 
                                                              1974

 1    those two documents are the two papers to which you have 

 2    referred, the "Damned Liars and Expert Witnesses" article 

 3    and the article on what happened in Hazelwood?

 4         A.    They are the two articles, yes.  

 5               MR. GLASS:  Your Honor, we would move the 

 6    admission of these two articles.  

 7               MR. SOLOMON:  No objection.

 8               THE COURT:  They are admitted as 80 and 81, 

 9    Defendants' Exhibits. 

10               (Defendants' Exhibits 80 and 81 for 

11    identification were received in evidence)

12         Q.    As to the proposed statistical adjustment before 

13    the Court, Mr. Meier, you have certain views, do you not, 

14    with respect to the presumption of accuracy within the P 

15    sample?

16         A.    I'm sorry.  Presumption of accuracy --  

17         Q.    -- of accuracy within the P sample.

18         A.    Yes.  My understanding is that the P sample is 

19    believed to be accurate because after it was originally 

20    taken, a second expert group went out and did a sample of 

21    those areas to verify the work that had been done before, 

22    came up with results very close to those of the original P 

23    sample, and this was taken as good evidence that the 

24    original P sample was correct.  

25               It is my view that the kinds of errors that are 
                                                              1975

 1    likely to be made are at that stage in the game likely to be 

 2    errors that an additional moiety of expertise is not going 

 3    to discover.  In short, it seems to me the same set of 

 4    errors was very likely present in the follow-up as was in 

 5    the original P sample and that, therefore, the presumption 

 6    of accuracy is really not sustained.  

 7               In all fairness, it is not clear to me how you 

 8    could really convince yourself that you did have an accurate 

 9    P sample.  But the difficulty of doing it doesn't make it 

10    accurate.

11         Q.    You also have certain views, do you not, with 

12    respect to the validity of the assumptions that underlie the 

13    census bureau's smoothing model?

14         A.    Yes, I do. 

15               MR. SOLOMON:  Pardon me, your Honor.  I do have 

16    to object.  This is one of the areas where the witness 

17    disclaimed any expertise at all.  He stated at his 

18    deposition that he doesn't claim any expertise in smoothing 

19    or in presmoothing.

20         Q.    Mr. Meier, can you tell the Court what knowledge 

21    or expertise you bring to bear on the subject of smoothing 

22    and presmoothing?

23         A.    Difficult question, I guess.  In all fairness, I 

24    have not made that a major topic of study for myself.  But a 

25    certain familiarity with these things is part of what any 
                                                              1976

 1    statistician knows.  As a statistician, with whatever 

 2    qualification you people wish to put on it, I do know 

 3    something about the matter, yes.

 4         Q.    With that as a prelude, would you share with us 

 5    your views on the assumptions underlying the bureau's -- 

 6               THE COURT:  Let's dispose of the objection first. 

 7               MR. SOLOMON:  I read from page 128 of Dr. Meier's 

 8    deposition.

 9              "Q.    Do you consider yourself an expert on the 

10    issues of census adjustment?

11              "A.    What is an expert is a kind of diffuse 

12    matter.  I am not an expert on presmoothing.  I am not even 

13    on expert on smoothing.  Although, I know generally what 

14    both are about and more about the latter than the former."  

15               I think we have had lots of testimony from 

16    experts in this matter on smoothing, and this does go 

17    beyond, I believe, this witness's expertise.  

18               THE COURT:  I am going to sustain the objection.

19         Q.    Mr. Meier, you have certain views, do you not, 

20    with respect to the role of loss functions in statistical 

21    theory and practice?

22         A.    I do.

23         Q.    Would you share those with the Court, please.

24         A.    Yes.  It has been said, and I was privy to some 

25    of the testimony recently given, that loss functions are 
                                                              1977

 1    very important in statistics, and I agreed with that:  They 

 2    are.  They are particularly important in the development of 

 3    certain kinds of statistical theories that are important in 

 4    our understanding the process of inference and give us some 

 5    guidance on methods of inference.  

 6               The motion, however, that loss functions are 

 7    typically used in the affairs of applied statistics is odd.  

 8    They are not.  I take the example that -- I'm sorry, the 

 9    name slips me at the moment, the economist.

10         Q.    Dr. Fisher?

11         A.    Yes.  

12               -- Franklin Fisher used, in which he described 

13    the problem that a drug regulator would have in deciding 

14    whether to approve a certain drug.  He said, well, the drug 

15    might save some lives, but it also has side effects and it 

16    might cost some lives, and we have to put together a loss 

17    function and optimize.  

18               In some very abstract sense, and he says that 

19    might be explicit or implicit, in some very broad sense 

20    implicitly I suppose that is true, but we take it very 

21    seriously to look at each component very carefully.  They 

22    are rarely of a kind that you can balance.  The side effects 

23    may be, for example, discomfort, headache, things of that 

24    kind.  The benefits may be the saving of life.  How many 

25    headaches are worth one death?  Very hard to put together, 
                                                              1978

 1    and we don't attempt to put them together.  

 2               In some principle, broad theory, we might insist 

 3    that it has to be done, and I would agree that implicitly we 

 4    are doing something like that.  But we do not use loss 

 5    functions.  We don't refer to loss functions.  On this I can 

 6    speak with authority because I have spent a great deal of 

 7    time in problems of just that sort.  

 8               The same is true generally in applications of 

 9    statistics.  So a claim that the loss function analysis has 

10    some special status in deciding this issue seems to me to be 

11    clearly in error.  

12               There has been testimony, and there is no need 

13    for me to repeat it, on the arbitrariness of loss functions.  

14    Squared error loss, particularly if it has to do with the 

15    total count rather than the shares, seems to me an almost 

16    bizarre loss function in this situation.  But it would be 

17    very hard to come up with a loss function on which there 

18    might be common agreement.

19         Q.    Mr. Meier, you may already have covered this, but 

20    you have certain views, do you not, about the problems 

21    caused by the multiplicity of options in the proposed 

22    statistical adjustment of the census?

23         A.    Yes, I do have very serious problems.  If I may 

24    say, apart from the general issues that I raise and were 

25    objected to, this is perhaps the most serious in the 
                                                              1979

 1    technical domain.  

 2               I come at this problem with a background from the 

 3    area in which I have spent most of my effort applying 

 4    statistics, the analysis of clinical trials, because the 

 5    same issue arises there.  If we are comparing two drugs, A 

 6    and B, for some therapy, we may find that it appears in this 

 7    study that drug A is somewhat better than drug B and we have 

 8    assigned patients at random which justifies carrying out a 

 9    statistical test, and we do that, and we find perhaps that A 

10    is significantly better than B and we are likely then at the 

11    FDA to say, well, we would approve A over B.  

12               Then someone comes along and says, hey, wait a 

13    minute, wait a minute, in group B you had more older 

14    patients than you had in group A, not significantly more but 

15    somewhat more.  Shouldn't you adjust for age?  And, oh, by 

16    the way, there were more women also, and you had better 

17    adjust for sex.  

18               Let me tell you that there are an almost 

19    limitless number of variables which will inevitably be 

20    somewhat different between the groups and which will give 

21    rise to the question shouldn't you adjust.  This is an issue 

22    of which I have made a study.  At least with respect to 

23    those studies, and there are a number, for which the data is 

24    public, the effect of adjustment has two components.  

25               Number one, with different adjustments we can 
                                                              1980

 1    move the estimate and the significance level around quite a 

 2    bit.  Number two, the adjustments provide no increase in 

 3    accuracy whatever.  

 4               The result is that if we encourage adjusted 

 5    analysis in preference to the unadjusted analysis, we have 

 6    put ourselves very much at the disposal of the investigator, 

 7    the person writing the paper, who has a great option to lead 

 8    us where he wants us to go, and this without any gain being 

 9    made in precision.  This I see as strongly analogous to what 

10    we are dealing with here.  

11               The method of adjustment that is used requires 

12    the determination of, what is it, 1352 or some such number 

13    of strata.  Those strata are put together not in an 

14    irrational manner but in an arbitrary manner and could well 

15    be put together in other ways.  There is no showing that I 

16    know of that putting them together in other ways would leave 

17    the number of shifts of representatives from one state to 

18    another unaltered.  

19               In effect, we invite ourselves to be in a 

20    situation in which the census presents a count which is not 

21    only a count plus some modest imputations but something that 

22    depends on a very arcane, difficult to understand adjustment 

23    method, one among many, for which we could make other 

24    choices.  That seems to me extremely unfortunate, likely to 

25    degrade the credibility of the census.  And in my view, even 
                                                              1981

 1    a substantial increase in precision -- and I don't believe 

 2    there is one -- would have to overcome a strong view that we 

 3    ought not to play that game.

 4         Q.    One final question, Mr. Meier.  One of your 

 5    fellow instructors at the University of Chicago law school, 

 6    Judge Posner of the Seventh Circuit, has recently 

 7    characterized litigation that seeks the statistical 

 8    adjustment of the census as litigation in which the 

 9    plaintiffs ask the Court to, quote, take sides in a dispute 

10    among statisticians, demographers, and census officials 

11    concerning the desirability of making a statistical 

12    adjustment based on the census head count."  Does that 

13    strike you as a fair characterization of this lawsuit?  

14               MR. SOLOMON:  Objection.  

15               THE COURT:  I will permit it.

16         A.    It does strike me as a fair characterization.  I 

17    would say it doesn't go nearly far enough.  But it does, 

18    indeed, get down to the question of who should judge whether 

19    a statistical analysis is competently put together or not.  

20    I think, with all due respect, the court is not well placed 

21    to do that.  But it does not take up the other issues that I 

22    began my statement with, which to my mind are at least as 

23    important.  So I quite agree with it.  As I say, it doesn't 

24    go far enough.  

25               MR. GLASS:  Thank you.  No further questions.  
                                                              1982

 1               THE COURT:  Thank you, Mr. Glass.  

 2               Mr. Solomon, let's have some fun.  

 3    CROSS-EXAMINATION 
                        

 4    BY MR. SOLOMON:

 5         Q.    I wonder if we could have a copy of those notes.

 6         A.    Yes.  

 7               MR. SOLOMON:  I wonder if I could have a minute 

 8    to read them.  

 9               MR. MILLET:  While Mr. Solomon is reading, I am 

10    reminded that I may have omitted moving into evidence 

11    Defendants' Exhibit 33, which was Dr. Fay's paper.  If I 

12    have forgotten to do so while Mr. Sherman is still here, I 

13    would like to do so now.  

14               THE COURT:  Dr. Fay's -- 

15               MR. MILLET:  Dr. Fay's draft papers on census 

16    estimation.  

17               THE COURT:  Yes, I remember it.  Any problem with 

18    it? 

19               MR. SHERMAN:  We continue to object to the 

20    admission of the report.  

21               THE COURT:  Would you favor me with a ground?

22               MR. SHERMAN:  Yes.  As Dr. Fay testified, it is 

23    not the final version of his conclusions.  Therefore, we 

24    believes his analysis as he testified today and at his 

25    deposition shouldn't be admitted for the truth.  It is not a 
                                                              1983

 1    bureau document.  It was written for purposes, apparently, 

 2    of publication.  Therefore, I don't think it falls under any 

 3    of the exceptions to hearsay, which the bureau documents 

 4    that have been introduced in this case might fall under.  

 5               THE COURT:  He did indicate that it was a summary 

 6    of his testimony.  The conclusions he articulated here were 

 7    the same ones that he has in that draft article.  

 8               MR. SHERMAN:  As a matter of fact, your Honor, I 

 9    don't think that is the case.  What he testified to are his 

10    current conclusions.  They aren't the conclusions that are 

11    in the draft paper.  In fact his analysis has changed.  One 

12    of the questions I asked on cross was whether his analysis 

13    had changed since his deposition, and he said yes.  His 

14    current analysis, which he did testify to, isn't reflected 

15    in the draft.  

16               THE COURT:  That was not my understanding.  Mr. 

17    Millet, am I wrong?  

18               MR. MILLET:  Your Honor, I think Mr. Sherman is 

19    not too far from the mark.  Frankly, your Honor, one of the 

20    reasons we wish to offer that exhibit is to show that this 

21    is an area that is very complicated and requires study and, 

22    as the data are further refined and further reviewed over 

23    time, the estimates of this important issue can change. 

24               THE COURT:  I surely don't need it for that 

25    purpose.  I will exclude it.  
                                                              1984

 1               MR. SHERMAN:  Thank you, your Honor.  

 2    BY MR. SOLOMON:

 3         Q.    Professor Meier, I notice one thing on your 

 4    resume I wanted just to clarify.  It stated that you were a 

 5    fellow of the Royal Statistical Society.  That is not the 

 6    Royal Society, is it?

 7         A.    No, certainly not.  Fellowship in the Royal 

 8    Statistical Society is the same as membership.

 9         Q.    You mentioned John Tukey.  He was your Ph.D 

10    adviser, correct?

11         A.    That's right.

12         Q.    You don't have any problem describing him as one 

13    of the giants in the field of statistics, do you?

14         A.    Characterizing John Tukey is not a trivial task.

15         Q.    You don't have any problem characterizing one of 

16    the characterizations of him being a giant in the field 

17    confident statistics, do you?

18         A.    No.  That is far from complete, however.

19         Q.    Did you read his testimony?

20         A.    I had a chance to look at a very little bit of 

21    it, but in all fairness I can't say I have read it.

22         Q.    What of the testimony have you read in this 

23    trial?

24         A.    I have had -- largely, I must confess, at my own 

25    request -- a number of transcripts poured at me in the last 
                                                              1985

 1    two weeks.  I read Steve Fienberg's direct, pretty much all 

 2    of that.  I had my attention directed to a little bit of 

 3    Franklin Fisher's.  John Tukey's I asked for but I just 

 4    didn't have time.  Literally, I turned bunches of pages and 

 5    saw what kind of thing he was talking about.  I have no 

 6    coherent account of John's testimony.

 7         Q.    You didn't read all of Professor Fisher's 

 8    testimony?

 9         A.    No, certainly not.

10         Q.    You offered the opinions that you did without 

11    reading all of his testimony?

12         A.    The opinions that I offered were based on very 

13    circumscribed and specific parts of the testimony.  I have 

14    no embarrassment in offering opinions on those.

15         Q.    You don't know whether what you read was 

16    complete, do you?  

17               THE COURT:  He knows it wasn't.  He must know it 

18    wasn't.  

19               MR. SOLOMON:  The witness doesn't know that it 

20    was complete on any of the issues that he read.  

21               THE COURT:  All right.

22         A.    All I can say is that he was wasn't talking about 

23    those things in the two pages before and in the two pages 

24    after.  He may have been that he took them up again later.  

25    So in principle you are right, but I would be surprised if, 
                                                              1986

 1    in fact, I hadn't read what he had to say on those subjects.

 2         Q.    You only read what the government gave you with 

 3    respect to Professor Fisher, correct?

 4         A.    They gave me everything, but they directed my 

 5    attention to those particular passages.

 6         Q.    You have talked about the article "Damned Liars 

 7    and Expert Witnesses."  One of the admonitions that you make 

 8    in this article is that, and correct me if I am wrong, an 

 9    expert ought to remain within his area of expertise to 

10    maintain credibility with the court.  Is that a fair 

11    statement?

12         A.    It is.

13         Q.    You also believe that it is preferred for an 

14    expert, when coming before a court, to write a report to 

15    fully explain all of his reasons, correct?

16         A.    It is preferable.

17         Q.    You didn't write a report here, did you?

18         A.    That's correct.

19         Q.    You also believe, on the basis of this article, 

20    among others, that it is important to fully read, read 

21    fully, in fairness, about the area that you are going to 

22    opine on, isn't that fair?

23         A.    I am not sure I know what part of that article 

24    you are referring to.  You may want to direct my attention 

25    to where I say something like that.
                                                              1987

 1         Q.    You and I can have a difference of opinion over 

 2    whether you should be fully prepared to testify about areas 

 3    that you are opining on?

 4         A.    It depends on what you mean by fully prepared.  

 5    If you mean do I have to read the Encyclopaedia Britannica, 

 6    the answer is no.

 7         Q.    When you first got involved in this effort, you 

 8    were contacted by one of the government's lawyers, Mr. 

 9    Sitcov, right?

10         A.    I was first contacted by David Freedman and 

11    subsequently by Mr. Sitcov.

12         Q.    At the time Mr. Sitcov mentioned to you that the 

13    Secretary's decision was going to be against or was against 

14    adjustment, correct?

15         A.    That may be.  In all honesty, I don't recall the 

16    details of that conversation.

17         Q.    Prior to the time that you talked with Mr. 

18    Sitcov, you already had a view that adjustment was unwise 

19    even if it were feasible, correct?

20         A.    That's correct?  Even before you read anything 

21    about the 1990 census, you doubted the feasibility of 

22    adjustment, correct?

23         A.    I wouldn't say at that point I doubted the 

24    feasibility.  I certainly doubted the wisdom.

25         Q.    Do you remember your deposition?
                                                              1988

 1         A.    Somewhat.  Do I have it here, by the way?

 2         Q.    Have you reread it?

 3         A.    Not in the last few days, no.

 4         Q.    If you need it, I will hand up a copy.  I want to 

 5    see whether this refreshes your recollection.  Can you place 

 6    in time when you reached the conclusion that adjusting the 

 7    census is not even technically feasible?  

 8               THE COURT:  Hold on a second.  Do you have a copy 

 9    there for him?

10         Q.    Page 10.  

11               MR. GLASS:  Yes.  

12               THE COURT:  Why don't you give it to him.

13         A.    I'm sorry.  The page?

14         Q.    Look at page 10.  In the answer that includes 

15    line 14, you told me that you doubted the feasibility to 

16    begin with, right?

17         A.    Yes.  It is followed by, "I saw no reason as I 

18    was reading to suspend judgment." 

19         Q.    So both statements were true?

20         A.    Yes, of course.

21         Q.    You were retained by the government several 

22    months before the Secretary announced his decision in July 

23    of 1991, weren't you?

24         A.    In all honesty, I don't really recall.

25         Q.    I ask you to turn to page 7 of your deposition, 
                                                              1989

 1    where I ask you on line 3, "Was the retention prior to July 

 2    15, 1991?"  

 3               You answered, "Yes.  

 4               "How long prior?  

 5               "A.   A few months."  

 6               Does that refresh your recollection?

 7         A.    It does, and I'm sure that's correct.

 8         Q.    You mentioned that you first became involved 

 9    because of Professor Freedman.  He is an old friend of 

10    yours?

11         A.    Yes.

12         Q.    You and Professor Freedman and Mr. Sitcov talked 

13    on the phone about the subject matter of your testimony 

14    before you were retained, correct?

15         A.    I'm sure that must be correct.  I can't imagine 

16    Mr. Sitcov retaining me without talking to me first.

17         Q.    Professor Freedman, in fact, mentioned to you the 

18    importance of having respect for the witnesses to support 

19    the view that he was supporting as one of the reasons why he 

20    wanted you to testify, correct?

21         A.    He was of the view that if there was a strong 

22    case of nonunanimity of the profession in favor of 

23    adjustment, it would be important to show that nonunanimity.  

24    He knew my general views and was, I suppose, confident that 

25    after reading the material I would probably be reinforced in 
                                                              1990

 1    those views.  Yes, he thought it was important that if I did 

 2    continue to hold that view, that I testify to that.

 3         Q.    At page 15 of your deposition, I asked you what 

 4    you and Professor Freedman talked about, I will read only 

 5    part of the answer.  "He mentioned the importance of having 

 6    respectable witnesses to support the view that was 

 7    supporting."  That was accurate when you told me it, wasn't 

 8    it?

 9         A.    I'm sure that is what I said.

10         Q.    Then he went on to urge you as a public citizen 

11    to get involved, correct?

12         A.    Yes, basically.

13         Q.    You are charging the government $350 an hour for 

14    your time, aren't you?

15         A.    As a matter of fact, I have not billed the 

16    government, have not yet decided whether to bill the 

17    government in this case.  That is my rate.  I may not have 

18    told you that I already promised Mr. Sitcov I would not 

19    charge my full rate if I did make a charge.

20         Q.    When did you promise him that?

21         A.    When he retained me.

22         Q.    You didn't tell me that at your deposition, did 

23    you?

24         A.    No, I'm afraid not.  Again, if you will direct me 

25    to the colloquy, where is that?
                                                              1991

 1         Q.    Where you telling me that you are charging $350?

 2         A.    That is my right.

 3         Q.    Yes, all right.

 4         A.    I think that is what you asked me.

 5         Q.    Right.

 6         A.    It may have been you asked what they were going 

 7    to pay me.  If I answered that, that would be a wrong 

 8    answer.  Well, you see it.

 9         Q.    What is your reduced rate for that matter?

10         A.    I may not have been specific, but I may have 

11    mentioned half.

12         Q.    You have spent upwards of a hundred hours already 

13    on this retention, is that right?

14         A.    I think so.

15         Q.    To prepare yourself to testify here, you read the 

16    Secretary's report, didn't you?

17         A.    I did.

18         Q.    You read maybe five other bureau reports, 

19    correct?

20         A.    Quite a bit of stuff, yes.

21         Q.    You read about five of them, is that right?

22         A.    That sounds right.

23         Q.    Which reports did you read?

24         A.    At this point I am not prepared to name them.  

25    There was a group report by several of the people who were 
                                                              1992

 1    on the committee to advise the Secretary, including 

 2    witnesses here who are proponents of adjustment.  I do 

 3    remember that.  

 4               I read several reports of David Freedman's.  I 

 5    read some further bureau material and some further material 

 6    by some of the individuals, individuals who had put in 

 7    statements in support of adjustment.  I'm sorry, today I 

 8    cannot tell you exactly which they are.

 9         Q.    With respect to the joint report, I believe you 

10    told me that you had scanned it, isn't that right?

11         A.    If I said so, that's what I did.

12         Q.    You rely on those reports, such of those that you 

13    read, for purposes of your knowledge of the PES here today, 

14    isn't that right?

15         A.    That's right.  

16               MR. SOLOMON:  Your Honor, we had asked for those 

17    reports to be produced to us.  The witness has expressed a 

18    failed recollection, as he did at his deposition, concerning 

19    which reports and which documents they were.  He says there 

20    are about five.  There were not that many.  

21               A request was turned down at the deposition.  I 

22    asked twice.  I was told at the deposition I didn't make a 

23    formal request.  After the deposition I made a formal 

24    request.  Then my request was turned down.  

25               We would like those documents produced under 705 
                                                              1993

 1    of the Federal Rules of Evidence, among other things.  

 2               MR. GLASS:  Your Honor, it has always been my 

 3    understanding that requests for documents were made under 

 4    Rule 34 or under Rule 45, and no such request was made in 

 5    this particular.  And I never understood that an oral 

 6    request for documents at a deposition was an appropriate 

 7    request.  

 8               THE COURT:  In any case, are you now prepared to 

 9    find these things and turn them over to him?  

10               MR. GLASS:  Sure.  

11               THE COURT:  All right.

12         Q.    In connection with the testimony that you have 

13    given, you, in fact, arr exclusively relying on the work of 

14    others; you have not done any independent work yourself, 

15    correct?

16         A.    I have not requested census tapes, I have not 

17    analyzed tables of data, that's correct.

18         Q.    You have not done any independent work.  All you 

19    have done is you have read the various reports that you 

20    can't identify now, right?

21         A.    Not quite right.  I sat with equations and I 

22    spent some time thinking hard about the nature of 

23    justification for the assumptions.  I looked in the various 

24    reports for statements of that kind.  I wouldn't say that 

25    was merely reading.  Again I say if you ask did I do any 
                                                              1994

 1    formal analysis, no, I did not.  It is not the same as 

 2    reading a novel.

 3         Q.    What equations did you look at?

 4         A.    Primarily, the equations 1 and 2 that Freedman 

 5    and others talk about in connection with the smoothing and 

 6    presmoothing.

 7         Q.    I am right that you don't consider yourself an 

 8    expert in smoothing and presmoothing, isn't that right?

 9         A.    Just that you don't carry that too far.

10         Q.    All right.  I only want to carry it to the length 

11    that you said it at your deposition.  You told me at your 

12    deposition you are not an expert on presmoothing, right?

13         A.    Sure, but -- 

14               THE COURT:  Hold on.  Why are we going into this?  

15    I have already ruled that I wouldn't take any testimony on 

16    the subject.  

17               MR. SOLOMON:  Except that the witness just talked 

18    about having looked at formulae of Professor Freedman.  If 

19    your Honor is not going to take any testimony from this 

20    witness on that, I will move on.  Thank you.

21         Q.    With respect to the PES generally, you don't 

22    regard yourself as a technical expert on those subjects, 

23    isn't that a fair statement?

24         A.    All I know of the PES is what I read in those 

25    documents, that's right.
                                                              1995

 1         Q.    You do know and agree that the origin of the 

 2    battle, if you will, the engine that drives the whole 

 3    argument for why adjustment is necessary, is the racial 

 4    differential undercount, correct?

 5         A.    That is my understanding.

 6         Q.    You saw only a passing reference to that issue in 

 7    the Secretary's decision, correct?

 8         A.    Whatever is there.

 9         Q.    Very well.  What's there?

10         A.    I don't recall what the Secretary said about that 

11    issue.  Again, I am a little confused.  You are giving me a 

12    piece of sociology, the driving engine, not a formal 

13    statement of what the contestants were contesting about.  I 

14    answered in terms of that sociology, and that's all.

15         Q.    Did you participate in the design of the PES so 

16    that you have some knowledge of what it was designed to do?

17         A.    I think I have been quite clear that I had no 

18    such participation.

19         Q.    You didn't participate in the planning of the 

20    PES, correct?

21         A.    Of course not.

22         Q.    You didn't participate in the carrying out of the 

23    PES, correct?

24         A.    Certainly not.

25         Q.    Before you testified about the PES, did you 
                                                              1996

 1    review the technical operations of the plans?

 2         A.    I did in fact review it.  And I cannot lay my 

 3    hand on -- I cannot identify the document now that described 

 4    quite well, or I thought it was quite well, what the PES was 

 5    and what the technical operations were.

 6         Q.    You can't identify that document to us now, is 

 7    that right?

 8         A.    No.  But we have agreed to find it for you.

 9         Q.    I asked you at your deposition some questions 

10    about the PES.  Do you recall that I asked you what the P 

11    projects were, and you didn't know what the P projects were?

12         A.    Not on that name.  The P sample, I knew what that 

13    was.

14         Q.    You thought that the P sample were in fact the P 

15    projects or the P studies, correct?

16         A.    If you are referring to something here, please 

17    point me to it.

18         Q.    You mentioned in your testimony to me that there 

19    were about 5,000 and some poststrata, right?  You know that 

20    to be wrong, correct?

21         A.    When did I say 5,000?  Today?

22         Q.    No.  When you were under oath at your deposition.  

23    Look at page 40, please, and tell me.  Maybe I have misread 

24    it.  Lines 20 to 22.

25         A.    It's odd, because I have been aware from the 
                                                              1997

 1    start that it is somewhat over a thousand.  I may have 

 2    misspoken there.  Certainly, it says 5,000, and I know that 

 3    is not right.

 4         Q.    It is on pages 40 and 41 where I asked you 

 5    whether you were familiar with the P projects or P studies, 

 6    you went and described the PES; isn't that a fair 

 7    characterization of what you did?

 8         A.    I think that is probably right.

 9         Q.    What is the E sample?

10         A.    That is the enumeration sample, the enumeration 

11    in the selected subsample.

12         Q.    What does it do?

13         A.    It tells us what the census counted in those 

14    areas.

15         Q.    Is that all it does?

16         A.    There is a whole process of matching and so on 

17    and so forth that goes with it.  Is that what you refer to?  

18         Q.    What is the result of the E sample?

19         A.    What is the result of the E sample?

20         Q.    Yes.  What does the E sample test for?

21         A.    The E sample is used in conjunction with the P 

22    sample.

23         Q.    What does the E sample test for?

24         A.    You are aiming at something that I don't follow.

25         Q.    You made up your mind that even if adjustment 
                                                              1998

 1    were technically feasible, it would be a bad idea for policy 

 2    reasons; you made up your mind about that with respect to 

 3    the 1980 census, is that a fair statement?

 4         A.    I had that view.  You phrase it in such a way as 

 5    to suggest I had more than the view that I have expressed, 

 6    some stronger prejudice.  I had the view that I just gave in 

 7    my testimony.

 8         Q.    Basically, your views have remained the same, 

 9    correct?

10         A.    They have not substantially changed, that's 

11    right.

12         Q.    You mentioned that you were trying to show that 

13    you had had some involvement in this, that you got some 

14    material from Dr. Bailar.  Do you remember testifying to 

15    that on direct?

16         A.    Yes.

17         Q.    Dr. Bailar was at the bureau at the time, right?

18         A.    Yes.

19         Q.    You looked at some of the material that she sent 

20    to you, you decided that you didn't have time to get into it 

21    in any depth, and you didn't, is that a fair statement?

22         A.    Yes.  That is, I made the judgment that that was 

23    not the best allocation of my effort at that time.

24         Q.    Thereafter, you didn't communicate with Dr. 

25    Bailar or anybody else at the census bureau either, did you?
                                                              1999

 1         A.    No.

 2         Q.    Are you aware that the National Academy of 

 3    Sciences sponsored two blue ribbon panels on the issue of 

 4    adjustment?

 5         A.    I am aware of the Committee on National 

 6    Statistics.  I have served on the Committee on National 

 7    Statistics.  You say two blue ribbon panels.  That sounds 

 8    right.  That they have appointed panels.  How blue the 

 9    ribbon is, I wouldn't be prepared to say.  But they honestly 

10    made an effort, no doubt about that.

11         Q.    Fair enough.  You didn't serve on either of 

12    those, correct?

13         A.    No.  Indeed, I wasn't asked to.

14         Q.    You didn't attend any of the meetings that they 

15    had, correct?

16         A.    I wasn't invited.

17         Q.    They were open meetings, weren't they?

18         A.    I repeat, I was not invited.  They may have been 

19    open meetings.  No, I didn't solicit invitationo or show up 

20    uninvited or anything like that. 

21         Q.    There were discussions of loss functions by the 

22    NAS; the NAS are people who you believed to be your 

23    colleagues, correct?

24         A.    Let's be careful.  Not by the NAS.  You must 

25    understand that committees of the National Academy of 
                                                              2000

 1    Sciences are typically without any member of the academy on 

 2    them.  So the academy responds to an executive or 

 3    congressional initiative saying we need to study something 

 4    of this kind.  They ask staff to help them appoint a 

 5    committee.  

 6               The fact that it is an academy committee doesn't 

 7    make it any more elect or elevated than, say, the ASA might 

 8    be.

 9         Q.    Who were the people on the NAS panel who studied 

10    the issue of census adjustment?

11         A.    At one point I knew.  I don't recall now.

12         Q.    You mentioned the ASA panels.

13         A.    Yes.

14         Q.    You didn't serve on any of those panels relating 

15    to possible adjustment of the 1980 or 1990 census, correct?

16         A.    No, I did not.

17         Q.    Did you attend any of those meetings?

18         A.    No, certainly not.

19         Q.    Have you ever had a joint statistic agreement 

20    with the bureau to study the issue of adjustment or any 

21    subject relating to adjustment?

22         A.    I have never had a contract or grant or 

23    what-have-you with the bureau.

24         Q.    With the substantial bibliography that is 

25    attached to your CV, is there an article in there about the 
                                                              2001

 1    census or adjustment of the census?

 2         A.    No.

 3         Q.    Have you ever written an article about the census 

 4    or adjustment?

 5         A.    No, I have not.  It would be there if I had.

 6         Q.    I'm sorry?

 7         A.    It would be there if I had.

 8         Q.    You might have written it and it didn't get 

 9    published.  In any event, you haven't written one, right?

10         A.    Thank you.  No, I have not.

11         Q.    I think when you were rushing through your direct 

12    you talked about, let's see if I can find it, presumption of 

13    accuracy of the P sample, invalid and untestable.  Do you 

14    recall testifying to that?

15         A.    Yes.

16         Q.    By the way, do you know the P sample is not the 

17    whole PES?  Do you know that?

18         A.    I'm not really clear on that issue.

19         Q.    You mean to be referring to the whole PES when 

20    you refer to the P sample here, is that right?

21         A.    Yes, I do.

22         Q.    What reports have you read of the testing that 

23    the bureau did of the PES?

24         A.    I did read a report which described the PES and 

25    which described their testing of it, and I have already said 
                                                              2002

 1    that we will try to find that one for you.  I can't say now.

 2         Q.    That was a report about the PES taken in 

 3    connection with the 1990 census, is that right?

 4         A.    I believe so.

 5         Q.    Do you know that there were tests at the bureau 

 6    carried out of the PES in 1985?

 7         A.    I don't know about the 1985 tests.

 8         Q.    1986?  There were two tests.

 9         A.    No.

10         Q.    1987 there was one test.

11         A.    The chronology of the tests, as you see, I don't 

12    know.

13         Q.    In 1988 there were three tests, and you don't 

14    know anything about those, correct?

15         A.    What I testified to is what I understood from my 

16    reading about the method of verification.  That's all.

17         Q.    The method of verification, in your view, is 

18    somehow untestable, right?

19         A.    I have testified on it.  All I can do is repeat 

20    what I said before, if you would like me to.  

21               (Continued on next page) 

22    

23    

24    

25    
                                                              2003

 1         Q.    Well, but you haven't read any of the reports of 

 2    the Bureau of the Census that were, in fact, carried out, is 

 3    that a fair statement?

 4         A.    I have read a statement of what was done and how 

 5    the verification was carried out.  It may well be that it 

 6    was a summary statement, not any one of these individual 

 7    documents that you are speaking of.  When we identify it we 

 8    will know.

 9         Q.    You are talking in this page and I think you 

10    talked about stratificaion in the capture-recapture 

11    methodology.

12               Are you aware that the Bureau published a number 

13    of reports about the capture-recapture methodology that it 

14    was going to use as part of the PES?

15         A.    I'm aware that the Bureau published on the dual 

16    system methodology, yes.

17         Q.    Are you aware that the NAS panel that you 
                                                           

18    referred to before vetted that design and concluded that if 

19    carried out in the way it was designed, it would lead to an 

20    improvement in the unadjusted counts?  Do you know that?

21         A.    I believe I heard that, but I would have had the 

22    view that they went beyond their competence in reaching such 

23    a conclusion.

24         Q.    Have you gone beyond your competence in reaching 

25    the opposite conclusion?
                                                              2004

 1         A.    No, I'm not.

 2         Q.    The census advisory committee, are you aware that 

 3    they studied the design of the PES and concluded that the 

 4    methodology that was being proposed carried out would work?

 5         A.    Which census advisory committee are you speaking 

 6    of?

 7         Q.    The ASA census advisory committee that you were 

 8    making reference to.

 9         A.    It was a joint statistical advisory committee, 

10    fine.  But those committees do not study, they read reports 

11    and give their opinion that they like the report.

12         Q.    You don't know whether they have done more study 

13    than you, do you?

14         A.    All I can draw on is my service on such a 

15    committee and let me tell you, I have done a heck of a lot 

16    more on this than any of us did at that time.  Maybe they 

17    are working a lot harder now.

18         Q.    You have done a heck of a lot more study of the 

19    census and the PES than they did, is that what you are 

20    saying, or you have just done a heck of a lot more study on 

21    what you think are fair analogies?

22         A.    What you are asking about is what I know of what 

23    census advisory committee members do with the material that 

24    is handed to them.

25               These are very busy people.  They generally read 
                                                              2005

 1    summaries, they generally listen to the person who wrote the 

 2    report and try to assess credibility.

 3               It could be that their behavior has now radically 

 4    changed, that they put aside all other things and devote 

 5    themselves to this topic.  That would be quite at variance 

 6    with my experience with such committees, but I was not on 

 7    that committee and I cannot say.

 8         Q.    Dr. Wolter testified here.  You know Dr. Wolter, 

 9    yes?

10         A.    I do.

11         Q.    And you believe him to be of high --  a very good 

12    statistician, correct?

13         A.    I have favorable regard for his capabilities.

14         Q.    He testified that dozens of reports were 

15    published by the Bureau during the 1980s describing the 

16    methodology that was going to be used as part of the PES.

17               I ask you to assume that, okay?

18         A.    Fine.

19         Q.    Which of those reports have you read?

20         A.    Can't say.  I think I've said several times now 

21    that I read one report that described this material.

22         Q.    And that's what you are now talking again about 

23    in connection with the 1990 census?

24         A.    Yes.

25         Q.    Okay.
                                                              2006

 1               Now, you mentioned that the issue of alternative 

 2    adjustments you found to be particularly devastating.

 3               Did I capture that correctly?

 4         A.    I don't know if I used the term, but it is 

 5    particularly troublesome, yes.

 6         Q.    Before you can make, offer an opinion on whether 

 7    the alternative adjustments would be troublesome, don't you 

 8    need to know what degree of change is within each of the 

 9    alternatives?

10         A.    I would need evidence that the proponent should 

11    give me that it really didn't matter.  There isn't any, at 

12    least I wasn't able to find any.

13         Q.    And you want the proponents to tell you that it 

14    really wouldn't matter, are you saying you want to make sure 

15    that no alternative scheme wouldn't change the apportionment 

16    in the House, is that what I think you said on your direct?

17         A.    I mentioned that as an example.

18         Q.    The apportionment of the House can change by 

19    moving one of 250 million people from one state to another, 

20    isn't that right?

21         A.    I'm now confused at what you are asking me to 

22    accept.

23         Q.    Isn't it correct that the apportionment can turn 

24    on --

25         A.    It can turn --
                                                              2007

 1         Q.    On the movement of one person.

 2         A.    It can turn, sure.

 3         Q.    Surely it is not reasonable to say that the PES 

 4    has not shown itself to be robust among alternative 

 5    assumptions if what you are testing it against is the 

 6    movement of one person, is it?

 7         A.    Oh, no, no, you quite misunderstand the thrust of 

 8    my comment.  That was merely an example of the gross of the 

 9    kind of thing that might happen if, indeed, my suspicious if 

10    differ ways of stratifying would lead to substantially 

11    different results.  It was not the one and only assessment 

12    of the multiplicity effect.

13         Q.    But you do believe that because it would change a 

14    seat in the House, that is one of the things that you would 

15    rely on to suggest that there is a very serious problem 

16    here, is that right?

17         A.    It would be one of the things of interest, 

18    certainly, and in particular would be of interest because if 

19    there are alternatives, and I can only believe in the future 

20    that interested parties will have the tapes to play with 

21    themselves, we will have state by state a claim that on some 

22    ground or another this particular adjustment is the best one 

23    to use, it's loss function shows the minimum loss and, by 

24    the way, it gives another seat to New York or to California 

25    or to Pennsylvania.
                                                              2008

 1               My expectation, and that is a full expectation, 

 2    if we go down this route is that we will be wound in such 

 3    litigation at least half the time between each census and 

 4    there will be a color of validity to each of these analyses.

 5         Q.    I want to ask you a hypothetical.

 6               I want you to suppose that there is an estimate 

 7    right here and I want you to suppose over here that there 

 8    are six alternative estimates.  I want to ask you to assume 

 9    that each of the six is better than this one.  All right?  I 

10    am stacking the deck.  Each of the six is better than this 

11    one.

12               You are not saying that the existence of those 

13    six precludes using any of them to improve the initial one, 

14    are you?

15         A.    I pretty much am saying that -- 

16         Q.    Okay?

17         A.    --  unless that improvement in accuracy is so 

18    large that it is require us to put aside another very 

19    important good, that is to say, uniqueness and credibility, 

20    and I wouldn't lightly put that aside.  

21               It would have to be a substantial gain, and, 

22    furthermore, we would need some criterion for picking among 

23    the six, and that would be tremendously important.

24         Q.    From a statistical point of view, you agree that 

25    even averaging the results of what are known to be better 
                                                              2009

 1    estimates will do better than ignoring the better estimates?  

 2    Don't you agree with that?

 3         A.    Again, it will be another version of those six.  

 4    And may I point out that those six are from six interested 

 5    parties who made the effort to get these things together.  I 

 6    would hardly want the census to be the average of what six 

 7    interested parties decided it ought to be.

 8         Q.    And that goes not only to statistics, but your 

 9    notion that the census is a national or American ceremony, 

10    correct?

11         A.    It has nothing to do with the ceremonial aspects 

12    at all.  

13               And allow me, allow me to correct you, by the 

14    way.  Statistics is not simply adding numbers, the 

15    methodology, the cooperation of the parties to a process is 

16    very much a part of statistics, and I do want to emphasize 

17    that.

18         Q.    Let me ask you, read you something and see if you 

19    agree with it.

20               Suppose two or more technicians have performed 

21    assays on several samples, you agree that it is desired to 

22    average their results in the best possible manner allowing 

23    for the fact that technicians differ in precision.

24               Do you agree with that?

25         A.    As a perfectly general statement, that is 
                                                              2010

 1    preposterous.  I mean, what are the ways in which these 

 2    assays arose?  Was it assayers in the one case the mining 

 3    company, in the other case the metallurgical company?   All 

 4    kinds of differences arise there.

 5               Did they present all the assays they had actually 

 6    carried out or only the ones more favorable to their side? 

 7               Such a question cannot be answered in general.  

 8               MR. SOLOMON:  May I hand the witness an exhibit 

 9    that we have marked.  

10               THE COURT:  Yes.

11               (Handing to the witness)

12               (Pause)

13         Q.    This is PX 756.  Let me ask you to look at that.

14         A.    I haven't seen this one in a long time.

15         Q.    Okay.

16               You do recognize it as an article that you wrote?

17         A.    You are quoting from my second paragraph --

18         Q.    All I am asking you now, sir, is whether that was 

19    an article that you wrote?

20         A.    This is an article that I wrote.

21         Q.    I am going to read into the record the second 

22    paragraph.

23               "For example, suppose two or more technicians 

24    have performed assays on several samples from a homogeneous 

25    material, it is desired to average their results in the best 
                                                              2011

 1    manner possible, allowing for the fact that technicians 

 2    differ in precision." 

 3               That's what you wrote, correct?

 4         A.    That is what I wrote, but it is not what you 

 5    quoted to me.

 6         Q.    Okay.

 7               You agree with me that the count for whites in 

 8    the unadjusted census is biased, the count for blacks in the 

 9    unadjusted census is biased, but the amount of bias is not 

10    the same?  Would you agree with that?

11         A.    I have no reason to doubt that that's true.

12         Q.    I want to put aside the lack of knowledge that 

13    you have, the knowledge that you talked about with respect 

14    to the PES.

15               You testified that you have a lot of experience 

16    in sampling, correct?

17         A.    Yet, quite a bit.

18         Q.    And at your deposition you and I went through a 

19    number of examples of where sampling is a good thing.

20               Do you remember that?

21         A.    Yes.

22         Q.    All right.  If you wanted to know how many people 

23    had been hospitalized with such and such a condition, it's 

24    been your view that you recommend sampling over trying to 

25    count them, correct?
                                                              2012

 1         A.    In many such cases, yes.

 2         Q.    And when you want to know what fraction of 

 3    readers in a university library were doing work related to 

 4    support research, it's been your point, sample them, don't 

 5    count them?

 6         A.    In that case that is exactly what we did in the 

 7    University of Chicago.

 8         Q.    And you believed you would get a more accurate 

 9    result that way?

10         A.    That's right.

11         Q.    Something as mundane as trying to find out 

12    millions and millions of tin cans, it's your view you would 

13    get a better result by sampling than counting, correct?

14         A.    Absolutely.

15         Q.    The railroad cases you talked about, in that case 

16    you believe that it is better to get as much information 

17    that you have in the form of a census or census data, hard 

18    data on computer and then sample the missing data so that 

19    you will improve the overall result, correct?

20         A.    There is no other way to do it.  On a typical 

21    railroad, that's a million shipments that you are looking at 

22    and in real time you could not process them all.

23         Q.    I asked you at your deposition for any analogize 

24    in the noncensus context, because that's all you can talk 

25    about, where you claim that the methods that you thought 
                                                              2013

 1    were being used in the PES were shown not to work. 

 2               Do you remember that?

 3         A.    Sounds right.

 4               Can you direct me to the pages?

 5         Q.    Yes.  Look at pages 113 and 114.

 6               Reference to the first is on 114, the second is 

 7    on 115.

 8         A.    This has to do with the utility of adjustment, 

 9    not anything about dual system.

10         Q.    Well, do you think that one of the things that 

11    the dual system estimate does is adjust the census numbers 

12    or --

13         A.    It's used as a piece of setting up an adjustment, 

14    yes.

15         Q.    And do you rely on the clinical work, the 

16    analysis of clinical tests that you have done to suggest 

17    that adjustment has not been shown to work or has been shown 

18    not to work?  Do you rely on the clinical studies for that?

19         A.    The appeal or examining a method in one, two, 

20    three other areas of statistical application is one of the 

21    very useful ways of making a judgment about what is likely 

22    to be able to do in a new domain, and the analogies are far 

23    from precise, but the conceptual issues are very much the 

24    same.

25         Q.    In fact, you relied on some of the analysis that 
                                                              2014

 1    you have done in clinical studies to offer the opinions that 

 2    you have here today, correct?

 3         A.    They influence my opinion, yes.

 4         Q.    And you pointed out to me in your CV two articles 

 5    when I asked you at your deposition to tell me what you were 

 6    basing your opinion concerning adjustment on, correct?

 7         A.    I see the Beach article.  Where's the other one?

 8         Q.    115.

 9         A.    That looks as if I am referring to my article on 

10    stratificaion in the clinical trials and --  oh, yes, that's 

11    right, that's the second paper on the same general topic.

12         Q.    And did you have any discussions with the 

13    government about using these articles as part of your 

14    testimony since you identified them to me at your 

15    deposition?

16         A.    Discussion with the government?

17         Q.    Yes, sir.

18         A.    We had some conversation about what I was going 

19    to say at deposition, so I probably did mention them, yes, I 

20    would think so.  But, again, memory is not sharp on that.

21         Q.    Okay.

22               I want to show you the two articles for just a 

23    minute.  We marked them for identification as PX 754 and 

24    755.  

25               MR. SOLOMON:  May I?  
                                                              2015

 1               THE COURT:  Yes.

 2               (Handing to the witness)

 3               (Pause)

 4         Q.    Let's look at 754 first.

 5               That is the article that you pointed me to in 

 6    your deposition, correct?

 7         A.    Right.

 8         Q.    And one of the conclusions that you reach in this 

 9    article is that post-stratification is important and useful 

10    in analysis of samples, right?

11         A.    Would you point me to that?  What page?

12         Q.    All I was reading from was the abstract just now.

13         A.    You are referring to 754?

14         Q.    Is that the stratificaion in the design of the 

15    clinical trials?

16         A.    No, it's not.

17         Q.    Then I apologize.

18         A.    You mean 755.

19         Q.    Let's start, then, with 755, then, please.

20         A.    Yes.

21               Yes, I see what you are referring to.

22         Q.    In this article, you talk about and really 

23    contrast the use of stratificaion in clinical trials, which 

24    you testified to on direct, and the use of stratificaion in 

25    sampling, don't you?
                                                              2016

 1         A.    Oh, yes, certainly.

 2         Q.    Okay.

 3               Look, please, at page 359.

 4         A.    Yes.

 5         Q.    You there talk about sample survey model and you 

 6    say, "The notion that stratificaion can contribute 

 7    substantially to improved precision arises largely from the 

 8    experience of sample surveys in which spectacular gains 

 9    sometimes arise from adroit stratificaion.  Analysis shows 

10    that these gains are of two separate types:

11               One, when stratum means differ appreciably, the 

12    use of stratificaion in analysis removes the between stratum 

13    variation from the error." 

14               Do you see that?

15         A.    Yes, indeed.

16         Q.    Do you believe that to be true?

17         A.    Of course.  I wrote it.

18         Q.    Do you know that the Census Bureau did that with 

19    respect to the PES?

20         A.    Oh, they did lots of stratificaion, indeed.

21         Q.    And do you know that that is one of the reasons 

22    they did it?

23         A.    Of course.  It would have been a meaningless 

24    exercise without.

25         Q.    Let me go on, "But more powerful by far, when 
                                                              2017

 1    within stratum variability varies widely, appropriate use of 

 2    different sampling rates can lead to very high gains in 

 3    efficiency." 

 4               Did the Bureau do that with respect to the --

 5         A.    Of course they did.  The Bureau is extremely good 

 6    in sampling methodology.  This kind of thing is their bread 

 7    and butter.

 8         Q.    And at the end of this paragraph hear you say, 

 9    "The lessons learned from sample surveys do not apply to 

10    clinical trials," right?

11         A.    This particular lesson or the lessons of this 

12    domain don't apply to clinical trials, because we are not 

13    privileged there to do the differential kind of sampling 

14    that was done in sample surveys.

15         Q.    Let's look very quickly at the other article.  

16    That is also an article by you that you told me about at 

17    your deposition?

18         A.    Yes, it is.

19         Q.    Let's go right to the conclusion.  That is on 

20    page 174 S.

21               "One principle seems to be supported both by 

22    simulation and the empirical findings - adjustment improves 

23    precision only if R squared and N are fairly large." 

24               The N in the PES is very large, isn't it?

25         A.    What the effective N for the PES would be may be 
                                                              2018

 1    a complex question, but I'm prepared to accept that if you 

 2    give that to me as a presupposition.

 3         Q.    You don't have an opinion on it?

 4         A.    The issue is change and the question has to do 

 5    with how many cases there are in which the two do not agree 

 6    with one another.

 7               What the proper measure of relative precision is 

 8    going to be there is not transparent to me.  I don't want to 

 9    affirm that the total size of the sample is the relevant 

10    index, neither do I deny it, and I'm prepared to accept it 

11    for purposes of this discussion.

12         Q.    What about the R squared, what is the R squared 

13    among the post-strata?

14         A.    I guess I have to know what the variables are 

15    that we are comparing.  That, as a general question, I can't 

16    answer it that way.

17         Q.    You don't know whether it was large?

18         A.    R squared for what?

19         Q.    What was the R squared among the 1,392 

20    post-strata?  Do you know, sir?

21         A.    R squared is not a characteristic of the stratum, 

22    it has to do with the relation sizes between the variables.  

23    Which variables?

24         Q.    Which variables were in the 1,392 --

25         A.    Counselor, we are somehow talking at 
                                                              2019

 1    cross-purposes.  You are not asking a statistical question 

 2    that I am recognize.  I'm sorry.

 3         Q.    Could you tell me what variables went into the 

 4    1,392 post-stratum, please?

 5         A.    What --  do you mean what characteristics of the 

 6    individuals picked up in the samples were measured, such as 

 7    age, address, so and and so forth? 

 8               I'm sorry, but the R squared referred to here is 

 9    not a piece that is relevant to our discussion about the 

10    census, in my opinion.  

11               I don't think we are going to --  I'll be glad to 

12    tell you whatever R squared might be.  It's probably quite 

13    large.  That's not the nature of my objection to the process 

14    of adjustment.

15         Q.    You agree that the PES is a real sample, right?

16         A.    Yes, indeed.

17         Q.    You also agree that the Bureau is very competent 

18    in the area of drawing samples, correct?

19         A.    Absolutely.

20         Q.    And if we really, really cared just about 

21    accuracy, that's all I want you to think about at the 

22    moment, just accuracy, and what we were concerned about was 

23    allocating seats in Congress, you agree that taking a sample 

24    such as the PES rather than a census could do a much better 

25    job, isn't that right?
                                                              2020

 1         A.    It wouldn't be a sample of anything like the PES.  

 2    The strata in the PES cross state lines.  You couldn't 

 3    possibly be thinking of doing that kind of sample for 

 4    allocation, it would be a different sort of sampling that 

 5    you would do.

 6         Q.    You agree that a sample will give you a more 

 7    accurate result with respect to allocating seats in Congress 

 8    than a census, correct?  

 9               MR. GLASS:  Asked and answered.  

10               THE COURT:  Sustained.  

11               MR. SOLOMON:  Your Honor, the first question had 

12    the word PES in it and the witness avoided, I believe, the 

13    answer by relying on that.  

14               MR. GLASS:  I object to that, your Honor.  

15               THE COURT:  Yes.  Sustained.

16               Move on.

17         Q.    If we were concerned about improving the shares 

18    of substate areas, it would be your opinion that taking the 

19    sample would give you better results than a head count, 

20    isn't that right?

21         A.    Would you say that again?

22         Q.    Yes.  If we were concerned about improving the 

23    shares --  you know what I mean by improving the shares?

24         A.    Oh, I do indeed, but I thought you said substate 

25    areas and now I'm lost again.
                                                              2021

 1         Q.    Okay.  

 2               THE COURT:  You did say that the first time.  You 

 3    did say substate.  He asked you to restate it and you left 

 4    that word out.  

 5               MR. SOLOMON:  Okay.  Let me try to be clear.

 6         A.    Good.

 7         Q.    I want to know whether it is your opinion that if 

 8    we were concerned about the accuracy among substate areas --

 9         A.    Which kinds of substate areas?

10         Q.    Any shares among substate areas?

11         A.    Well, blocks, census tracks, what?

12         Q.    Does it matter?

13         A.    It sure does.

14         Q.    Tell me whether it is your opinion that taking 

15    the sample would be better than taking the census with 

16    respect to any substate area with respect to its shares?

17         A.    In my direct testimony I pointed out that one 

18    reason we do a head count is when we get down to the level 

19    of blocks, sampling is really helpless.  We have to do a 

20    count to get anything meaningful at that level.  It really 

21    depends upon how big a chunk you are looking at.

22         Q.    Let me ask you, please, to look at page 206 of 

23    your deposition.  Let me ask you to start at page 205.

24         A.    All right.

25         Q.    I asked you if you had any objection on a 
                                                              2022

 1    theoetical basis to taking estimates for areas for which we 

 2    have direct sampling estimates and applying those areas to 

 3    which we don't have direct sampling estimates, and your 

 4    answer, "Depends on what purpose.  If, for example, we saw 

 5    the only purpose of the census was to allocate seats in 

 6    Congress, then I would not have a problem with saying we are 

 7    not going to count everybody.  We are going to do this by 

 8    sampling methods, we can do a much better job.  And I would 

 9    agree that it would be a much better job and we would get a 

10    better allocation for purposes of allocating Congressional 

11    seats if we just turned to sampling.

12               I then asked you, "Am I correct that one of the 

13    things that the PES does is take direct point estimates for 

14    certain areas and apply those to areas for which there are 

15    no direct point estimates?  "

16               And you said, "Often that's the nature of it."

17               MR. GLASS:  Was there a question at the end of 

18    this?

19         Q.    I then said, "In theory, you don't have any 

20    objection to that part of the PES?" 

21               You then said, "I have no objection to that 

22    beyond my objection to degrading what is perceived and 

23    thought of as a count.  It isn't a count.  It may be the 

24    best way to allocate seats, but it isn't a count, and to the 

25    extent that we say we do want a count, it doesn't provide 
                                                              2023

 1    it.

 2               "Q.    From a statistical point of view, you 

 3    don't have any objection?

 4               "A.    The statistical point of view has to be 

 5    with respect to a particular objective and if the objective 

 6    is to say how many people are in this state, then my answer 

 7    is agreeable.  I have no objection.  But the objectives of 

 8    the census are much more than that and when we look to the 

 9    other objectives, then I do have an objection." 

10               Do you recall that testimony?

11         A.    Yes, indeed.

12         Q.    Let's then turn to the other objectives that you 

13    talked about in your direct.

14               These other objectives that you were talking 

15    about are not in the realm of statistics, is that a fair 

16    statement?

17         A.    Not at all.  

18               THE COURT:  It's not at all a fair statement or 

19    not at all? 

20               THE WITNESS:  That's right, not at all a fair 

21    statement.

22         Q.    Let me ask you to turn to page 102.

23               "Q.    The comment that you made about the census 

24    being an American ceremony, you are not offering statistic 

25    or statistical based opinion when it comes to that?
                                                              2024

 1               "A.    Oh, no." 

 2               That's what you answered, correct?

 3         A.    About the business of the ceremony, yes.

 4         Q.    Okay.

 5               What, in fact, you are saying is that the census 

 6    somehow should be the best practical count that the Bureau 

 7    is able to come up with, is that a fair statement?

 8         A.    I said something very close to that, yes.

 9         Q.    Are you aware of estimates made by other 

10    governmental entities concerning how much error there is in 

11    the unadjusted counts?

12         A.    You say other than the census?

13         Q.    Yes, sir.

14         A.    I'm sure there are such estimates.  I'm not 

15    familiar with them.

16         Q.    Based on Census Bureau data, are you aware that 

17    the GAO estimated that there could be upwards of 25 million 

18    errors in the census?  

19               THE COURT:  The 1990 census?  

20               MR. SOLOMON:  Yes, sir, PX 16.

21         A.    If you tell me that's so, I'm prepared to believe 

22    it.  I don't have that knowledge of my own.

23         Q.    Do you agree that the existence of between 14 and 

24    25 million errors degrades the integrity of the census?

25         A.    Careful with the words.
                                                              2025

 1               The existence of error certainly does degrade the 

 2    census.  The integrity has more I think to do with the 

 3    clarity and straightforwardness of the methodologist and the 

 4    competence with which it is carried forward.

 5               I can conceive a census that is seen as done with 

 6    full integrity that acknowledges many things it cannot 

 7    measure and acknowledges the kind of errors, something that 

 8    the census routinely and typically does and should be very 

 9    much commended for.

10         Q.    You acknowledge that the census for a long time 

11    has used imputation, right?

12         A.    Has used certain kinds of imputation, that's 

13    correct.

14         Q.    And the imputation that is used in the census is, 

15    in your words, in the double digits, right?

16         A.    My words double digits?  I don't recall saying 

17    that.

18         Q.    Okay.  What is the estimate of the imputation 

19    that is used in the census?  You don't know?

20         A.    Now you are confusing me.  By imputation I mean a 

21    method, a method for inferring a count for a dwelling unit 

22    that is believed to be occupied but nobody had access to the 

23    people in it, for example.

24         Q.    How much imputation occurs in the census?

25         A.    I have read that.  I'm not prepared to give a 
                                                              2026

 1    number now.

 2         Q.    More than you would like, isn't that right?

 3         A.    Always more than we would like.  We would rather 

 4    have a count that's direct head count if we could get it.

 5         Q.    You disagree with the Bureau that the use of 

 6    imputation improves the counts?

 7         A.    I would have to study in more detail than I have 

 8    the extent of the imputations that the census now uses.

 9               I think the census is on dangerous grounds 

10    whenever it imputes, for example, the kind of thing that I 

11    just said and it is not immediately clear to me that the 

12    gain in doing that is worth the candle, but I have not 

13    studied that enough to take a firm position on it.

14         Q.    You have studied enough to be able to tell me 

15    that in your opinion the census is not a head count, right?

16         A.    It's not a pure head count.

17         Q.    Okay.  You agree with me that in your opinion, 

18    the census is sort of a hazy notion of a number?

19         A.    It's better than that.

20         Q.    It's better than that.  Okay.

21               Let's look at page 139 of your deposition I.

22               Just after the answer where you say there is 

23    error in every stage of the system:

24               "Q.    You agree that the it is not a head count?

25               "A.    Not in that sense, certainly not." 
                                                              2027

 1               You agree with that, don't you?

 2         A.    I guess we do.  We would have to go back to see 

 3    the previous colloquy and see what sense we are talking 

 4    about.  

 5               Do you really want to do that?

 6         Q.    I am perfectly happy for you to read the prior 

 7    question and answer.  I like that even better, but it's just 

 8    not what I am talking to you about.

 9               (Pause)

10         A.    What I am agreeing to is that it is not a head 

11    count in the sense that we do not have a census taker 

12    standing and talking and identifying each member of the 

13    family.  

14               The census has for a long time depended on 

15    respondents from the household, one respondent to give the 

16    information for every one in that household.  

17               When you pressed me on the matter is it a head 

18    count, not in the sense that everybody has been physically 

19    looked at, not in that sense, but most of us would call it a 

20    head count nonetheless.

21         Q.    Do you recall telling me at your deposition that 

22    your view was that there is a sort of hazy notion of a 

23    number, there isn't really a true number we are trying to 

24    get at, and, of course, we are trying to get more at shares 

25    than --  
                                                              2028

 1               MR. GLASS:  Can you refer us to a page number?

 2         Q.    --  Than we are trying to get to numbers anyway?

 3               We are on page 103.

 4               It would speed things along if that refreshed 

 5    your recollection:?

 6         A.    Ah.

 7               I would like to look at the previous paragraphs.

 8               (Pause)

 9               Ah, yes.

10               This answer is the final part of a response in 

11    which I point out the arbitariness what we count in the 

12    census.  I raise the question should we be counting the 

13    wetbacks, are they there --

14         Q.    The what?

15         A.    I beg your pardon?

16         Q.    I didn't hear you,.  The wetbacks?

17         A.    You see it there just as I do, counselor.

18         Q.    By that you mean Mexican-Americans?

19         A.    I said so in the answer.  I don't understand 

20    this.  

21               MR. GLASS:  YouR HONOR, I object to the tone of 

22    the questions.  

23               THE COURT:  Stop with the innocence outraged 

24    performance, will you.  Let's get this over with? 

25               The direct examination took 20 minutes, the 
                                                              2029

 1    cross-examination has been going on for an hour and a half.  

 2    This witness is not your witness.  

 3               MR. SOLOMON:  No, your Honor --  

 4               THE COURT:  Move on.  

 5               MR. SOLOMON:  I would like an answer to the 

 6    question.  

 7               THE COURT:  What was the question?  

 8               MR. SOLOMON:  The question was whether the 

 9    witness, whether his recollection is refreshed that the 

10    census in his view is sort of a hazy notion of a number.

11         A.    I was trying to answer that when you interrupted 

12    me, counselor.

13               Let me go on.

14               There are the wetbacks, the people that come from 

15    Mexico when a certain crop is in season, they are there, all 

16    right, but they are not going to stay much beyond the census 

17    date because the crops will be picked.  We make arbitrary 

18    decisions about those.

19               And I go on to talk about my conversations with 

20    census staff arguing that it's important to count them 

21    because it's so hard not to count them rather than any 

22    printed determination of whether they belong in or out, and 

23    it was in the light of that context that I referred to a 

24    hazy number, and with that refreshment, I guess I would have 

25    to go back and say, yes, it remains hazy because of all of 
                                                              2030

 1    arbitariness.

 2         Q.    Do you agree that the concept of a census --  do 

 3    you agree that the concept that a census must rigorously 

 4    count everyone is false and has been false from the 

 5    beginning?

 6         A.    Do I agree that a census --  please continue.

 7         Q.    Do you agree that the concept that the census 

 8    must count everyone is false and that that concept has been 

 9    false from the beginning?

10         A.    The concept that the census does count everyone 

11    is certainly false.  No one would imagine that.

12               The concept that the census attempts to count 

13    everyone is not false, they try as best they can.

14         Q.    You talked about the arbitrary decisions that are 

15    made.  You agree with me that some of those arbitrary rules 

16    have an effect on the apportionment, don't you?

17         A.    Of course they do.

18         Q.    And you believe that many of them have dramatic 

19    effects on the counting, don't you?

20         A.    I would be hard put to talk about how dramatic 

21    effects they are.  That they might have effects that they 

22    are cognizable, I am prepared to say, yes, I think they may.

23         Q.    That the census has arbitrary aspects to it you 

24    maintained is wellknown, correct?

25         A.    It's well known and it's unavoidable.
                                                              2031

 1         Q.    One of the examples that you gave was where you 

 2    count college students, correct?

 3         A.    That's right.

 4         Q.    Where are college students counted, by the way?

 5         A.    The truth is I don't know where they are counted 

 6    in 1990.

 7         Q.    You also testified that it was important that the 

 8    census be the sum, the whole equal to the sum of its parts.

 9               Is that a fair statement?

10         A.    Yes, indeed, that an important aspect of the 

11    census when reported is that if you add up the counties in 

12    Illinois you get the total for Illinois.

13         Q.    Do you know whether the census as reported, in 

14    fact, adds up in the way you just said?

15         A.    I believe it's made to do that, yes.

16         Q.    Do you know where servicemen counted overseas is 

17    placed?

18         A.    Where it is currently I don't know.  

19               MR. GLASS:  Your Honor, I have to object to the 

20    relevance of this.  

21               MR. SOLOMON:  He testified on direct.  

22               THE COURT:  That what?  

23               MR. SOLOMON:  Your Honor, he testified on direct 

24    that it was very important to the concept of a census that 

25    the whole equal the sum of its part.
                                                              2032

 1               That is not true now and I sought to establish 

 2    that.  

 3               THE COURT:  I never heard him say the whole had 

 4    to equal the sum of its parts. 

 5               THE WITNESS:  No, I did say that the counts for 

 6    each county of Illinois when added together should add to 

 7    the total of Illinois.

 8         A.    Counselor, are you telling me that is not the 

 9    case?

10         Q.    That is not the case, sir.

11         A.    I stand corrected.  That's the misfortune in the 

12    census, then.  I had thought they would not permit that.

13         Q.    The census as an American ceremony, can you tell 

14    us whether there is any scientific basis for the belief that 

15    you offer the court on that?  

16               MR. GLASS:  Your Honor, I have to object.  This 

17    is getting ludicrous.  This is questioning for the sake of 

18    questioning.  It is burdensome.  

19               THE COURT:  I will permit it.

20               Go ahead.  

21               MR. SOLOMON:  I will be done shortly, sir.

22         A.    I think that question of ceremony is relevant to 

23    our consideration of the census.  

24               When you say is there a scientific basis for it, 

25    I'm at a loss to answer.  It is a piece of sociology of the 
                                                              2033

 1    census and, indeed, the relationship of government to the 

 2    people.  It's highly relevant to the problem that we are 

 3    dealing with.  It is not a matter of mathematical art.

 4         Q.    Are you aware that there are many portions of the 

 5    population that do not believe they are entitled to 

 6    participate in the ceremony?  

 7               MR. GLASS:  Objection.  

 8               THE COURT:  Sustained.

 9         Q.    Are you aware that the Secretary purports to rely 

10    on a study which --  

11               MR. GLASS:  Objection.  

12               THE COURT:  Let me hear the question.

13         Q.    --  which finds that there are large segments of 

14    the population that would prefer an adjustment?

15         A.    Am I aware that the Secretary relies on a study 

16    that says large proportions of the population would prefer 

17    an adjustment?

18         Q.    Yes, sir.

19         A.    I didn't recall that.

20         Q.    Are you aware that in a report that the Secretary 

21    relies on, a greater percentage of the polled people prefer 

22    an adjustment and a very great percentage of racial and 

23    ethnic minorities prefer an adjustment?  Were you aware of 

24    that?

25         A.    I can believe that and I would believe that they 
                                                              2034

 1    have been grossly misled about what the effect of that 

 2    adjustment might be for them.

 3         Q.    What is the basis for that view?

 4         A.    Because I read The New York Times and I've read 

 5    much of the testimony, the allegations of the necessity of 

 6    this in order to get fair shares into the cities, et cetera, 

 7    et cetera, et cetera.

 8               There is, in my view, no political barrier to get 

 9    whatever fair shares the Congress chooses to allocate and to 

10    lay such burdens on the census is, in my mind, bizarre.

11         Q.    Have you at any time suggested to the Census 

12    Bureau any particular way that it could ameliorate the 

13    differential undercount?

14         A.    Have I suggested to the Census Bureau?

15         Q.    Yes.

16         A.    No, I have not.

17         Q.    In fact, it is your view the way to ameliorate 

18    the differential undercount is to count better, isn't that 

19    right?

20         A.    Pretty much.

21         Q.    Thank you.  

22               THE COURT:  Is there any redirect?  

23               MR. GLASS:  Only one question, your Honor.  

24    REDIRECT EXAMINATION 

25    BY MR. GLASS:
                                                              2035

 1         Q.    Mr. Meier, does your view about the 

 2    appropriateness of statistical adjustment of the census 

 3    depend on whether the number of post-strata is 1,392 or 

 4    5,000?

 5         A.    Not at all.

 6         Q.    Thank you.  

 7               MR. GLASS:  Thank you.  

 8               THE COURT:  Thank you very much, Mr. Meier.

 9               You may step down.

10               (Witness excused) 

11               MR. SOLOMON:  The documents that we asked for, 

12    your Honor, if we might have an estimate of when we will get 

13    them.  

14               MR. GLASS:  We will caucus with Mr. Meier and we 

15    will let you know as soon as possible.  

16               MR. SOLOMON:  I shouldn't think there is going to 

17    be a need to ask the witness any more questions because of 

18    that, but the sooner they can get it to us I think the 

19    better off we will all be.  

20               THE COURT:  What's next?  

21               MR. SITCOV:  How about the weekend.  

22               THE COURT:  We will recess and resume Tuesday, I 

23    guess, at 9:30.

24               (Adjournment taken to Tuesday, May 26, 1992, at 

25    9:30 a.m.) 
