                                                              1581

 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 21, 1992
                                              9:30 a.m.
20    
      
21    
      Before:
22    
                 HON. JOSEPH M. McLAUGHLIN,
23    
                                              Circuit Judge
24    

25    
                                                              1582

 1               THE COURT:  Mr. Zimroth, I understand you have 

 2    some preliminary matters?  

 3               MR. ZIMROTH:  Yes, your Honor.  There are two 

 4    matters concerning documents that we wanted to see if we 

 5    could clarify and put aside before we go on.  

 6               The first is, your Honor, we think that there are 

 7    a substantial number of documents that we think are self- 

 8    authenticating that we intend to introduce at the trial.  

 9               THE COURT:  What is a substantial number?  2,000, 

10    200?  

11               MR. ZIMROTH:  Maybe hundreds, your Honor.  

12               THE COURT:  Hundreds?  

13               MR. ZIMROTH:  Yes.  But what we are attempting to 

14    do is to weed them out so that we can offer only the most 

15    important ones and eliminate the others.  What we propose to 

16    do is to try to weed them out and then present that list to 

17    the government so they can then see precisely which ones we 

18    want to offer and either agree with us that they are 

19    admissible or disagree with us that they are not admissible, 

20    and then we can present to your Honor only those, I hope, 

21    handful about which there is a disagreement. 

22               I propose that we get that list to Mr. Milletpy 

23    Tuesday or Wednesday of next week.  I think after we finish 

24    out our weeding-out process there won't be a huge number of 

25    them.  And of those, I hope there will only be a small 
                                                              1583

 1    handful that we have a dispute about.  

 2               MR. MILLET:  We are certainly willing to try, 

 3    your Honor.  

 4               THE COURT:  I can't ask for any more than that.  

 5               MR. ZIMROTH:  A second point, your Honor, during 

 6    the direct testimony of Eugene Ericksen, who was the first 

 7    witness at the trial, we marked and your Honor admitted into 

 8    evidence several so-called ethnographic studies.  When your 

 9    Honor admitted them into evidence, you said that that was 

10    subject to our being able to prove that those ethnographic 

11    studies were available to the Secretary prior to July 15th.  

12               THE COURT:  The decision date.

13               MR. ZIMROTH:  The decision date.  Although we 

14    think we could prove that with respect to each of those, 

15    that proof would be extremely cumbersome and time-consuming.  

16    Frankly, we are trying to streamline the trial.  

17               There are about half of those ethnographic 

18    studies that on their face are dated prior to the decision 

19    date, so those, I think, are admissible on their face.  The 

20    ones that are not, and I would give you the numbers now, in 

21    light of the government's objection and your Honor's ruling, 

22    we will withdraw.  

23               Let me just read those into the record.  

24               THE COURT:  Is this all acceptable to you, the 

25    ones that predate the decision are admissible, those 
                                                              1584

 1    thereafter are withdrawn?  

 2               MR. SITCOV:  Some of those say "Draft," your 

 3    Honor.  I don't have them in front of me.  There is no way I 

 4    can tell from them whether anybody saw them in that form 

 5    before the decision.  

 6               THE COURT:  Let's today get rid of the ones that 

 7    postdate the 15th, which you will now dictate in the record.  

 8               MR. SITCOV:  That is fine.  

 9               THE COURT:  Let the record stand as it is with 

10    the predated ones, subject to your review, and you can 

11    attack them at some later time.  

12               MR. SITCOV:  That is fine.  

13               MR. ZIMROTH:  That is fine, your Honor.  I would 

14    say this particular argument that Mr. Sitcov made was made 

15    to your Honor previously, that is, the difference between 

16    draft and nondraft, and your Honor said that might go to the 

17    weight but it would not go to its admissibility, and you 

18    admitted them subject only to what I have said this morning.  

19    So obviously if Mr. Millet -- 

20               THE COURT:  If I have already ruled on it, I will 

21    adhere to that ruling.  It sounds right.  

22               MR. ZIMROTH:  It is PX-501 -- 

23               THE COURT:  These are the ones going out?  

24               MR. ZIMROTH:  That is correct.  PX-501, 502, 504, 

25    507, 512, 525, and 528.  
                                                              1585

 1               THE COURT:  Thank you, Mr. Zimroth.  

 2               MR. RIFKIND:  I have two further preliminary 

 3    matters.  Confession being good for the soul, the affidavit 

 4    which I said I would have today is on its way from Boston, 

 5    and I am not confident I will have it here by the end of the 

 6    day.  I would propose, if it is convenient to everyone, to 

 7    submit it tomorrow morning, the affidavit relating to the 

 8    production of certain of the exhibits that I said I would 

 9    give them in two days, two days ago.  

10               THE COURT:  Mr. Millet?  

11               MR. MILLET:  Your Honor, that is fine.  We said 

12    we would endeavor to get our responding affidavit by 

13    Tuesday.  We will still try to do that.  Given the fact that 

14    we are all here and the records are back in Brooklyn, we may 

15    need additional time to respond.  With that understanding, 

16    we have no problem with the additional time.  

17               THE COURT:  Understood.  

18               MR. RIFKIND:  Finally, counsel for Hudson County, 

19    New Jersey, Mr. Segen, who spoke to the Court the other day, 

20    asked if he might say a word at this time.  This fellow from 

21    New York cedes to that fellow from New Jersey.  

22               MR. SITCOV:  May I say one thing about the 

23    ethnographic studies before we hear from Mr. Segen?  

24               THE COURT:  Yes.  Is that what he is going to 

25    talk about?  
                                                              1586

 1               MR. SITCOV:  I don't think so.  

 2               The list that we have of these studies that Mr. 

 3    Zimroth was reading from includes some that were dated after 

 4    the decision that he apparently has not agreed to withdraw.  

 5               THE COURT:  You don't think he has withdrawn all 

 6    post-decision studies?  

 7               MR. SITCOV:  Right.  

 8               THE COURT:  Check it out with him later.  

 9               Mr. Segen.  

10               MR. SEGEN:  Yes, your Honor.  Leon Segen of 

11    Scarinci & Hollenbeck, representing the County of Hudson, 

12    New Jersey.  

13               Your Honor, I had the pleasure of being here on 

14    May 12th and informing the Court that we join in the issue 

15    that had been raised by plaintiffs in regard to adjusting 

16    the entire census count on the basis of numerous errors 

17    including differential undercount, but that we intended to 

18    preserve two additional issues:  The first issue being the 

19    differential undercount being the product of bias was a 

20    separate and distinct claim, and that even if the Court 

21    decided that the entire head count should not be adjusted as 

22    a result of the numerous errors, that nevertheless, the 

23    differential undercount should be adjusted; number two, that 

24    the failure to use adjusted statistics would have the effect 

25    of diluting minority voting strength in violation of the 
                                                              1587

 1    Voting Rights Act.  

 2               At that time your Honor advised me that I should 

 3    monitor the proceedings in order to make certain that 

 4    nothing fell through the cracks.  I have been doing my best 

 5    to do that.  However, there have been some difficulties.  

 6    Plaintiffs' counsel has been under the gun getting prepared 

 7    for the trial every day.  I have been reviewing the 

 8    transcripts and having conversations with them, with those 

 9    limitations.  

10               One thing that I have found as a result of that 

11    conversation that I had with Mr. Goldin of the Corporation 

12    Counsel yesterday is that there may be a gap in regard to 

13    the evidence pertaining to the Voting Rights Act.  It is my 

14    information from Mr. Goldin that Professor Cain was the 

15    essential witness in regard to the political impact of 

16    adjustment or the failure to adjust, and that testimony 

17    indicates that there has been a dilution of the voting 

18    strength but that there is no statement in the record that 

19    the impact of that would be that a minority Congressional 

20    district at some level, be it federal, state or municipal, 

21    etc., would be denied or would not be created as a result of 

22    using the original head count as opposed to the adjusted 

23    statistics.  

24               I have been unable to actually see Professor 

25    Cain's transcript.  Part of the problem, I believe, lies 
                                                              1588

 1    with me.  Part of the problem also lies with the fact that 

 2    at least as of the time that I spoke to Mr. Goldin on 

 3    Wednesday at about 3:30, the Corporation Counsel was not in 

 4    possession of that transcript.  

 5               I am assuming that as to the other claim that we 

 6    have raised, all the testimony that I could possibly have 

 7    hoped for would be in there, because it is rather apparent 

 8    that the first issue involves all of the evidence that 

 9    Hudson County's issue as to a distinct differential 

10    undercount involves.  I would ask the Court for the 

11    opportunity to be able to fully review Professor Cain's 

12    testimony and also to fully review as to the differential 

13    undercount issue.  

14               For that purpose I am here today to ask the 

15    Court, given that it is my understanding that plaintiffs 

16    intend to rest today, that the Court allow Hudson County to 

17    reopen as to the issue of the Voting Rights Act if a 

18    stipulation or some other means cannot be created for 

19    developing that one factual issue, the impact on 

20    Congressional districts, and for any other lacunae that 

21    might have occurred.  

22               I have been attempting to monitor, your Honor, I 

23    believe the case has been going at a rather steady clip, at 

24    a rather impressive clip, and it has posed certain 

25    difficulties in terms of my ability to be able to be right 
                                                              1589

 1    on top of things.  I believe that in regard to the Voting 

 2    Rights Act and Professor Cain, there is a very good reason 

 3    why I find myself a little bit behind at this point in time.  

 4    I would just ask your Honor to provide Hudson County with 

 5    this opening should it become necessary.  

 6               THE COURT:  Mr. Millet?  

 7               MR. MILLET:  Your Honor, with all due respect, 

 8    the defendants would have to vigorously oppose that.  We are 

 9    on the verge of starting our case today, I believe.  We 

10    conducted our discovery on behalf of defendants as to all 

11    the plaintiffs.  All the plaintiffs have been represented by 

12    counsel here today.  All the intervenors took the case as 

13    they found it.  For another party to come in at 11:59 and to 

14    say they want to start putting on new evidence for which 

15    there has been no discovery is absolutely prejudicial to us.  

16               As to the voting rights issue, your Honor, again, 

17    I frankly have not had the opportunity to research it.  It 

18    is my understanding that the Voting Rights Act does not 

19    operate against the federal government.  I think that is a 

20    legal issue, as we said at the beginning of the trial.  I 

21    think that is something that can be handled in a post trial 

22    memorandum.  But we certainly oppose any effort by any of 

23    the individual plaintiffs at this very late hour to ask to 

24    put on new evidence.  

25               THE COURT:  Mr. Segen, I have no problem with 
                                                              1590

 1    your putting in papers later on, but I must say it gives me 

 2    pause to let you reopen the hearing and start calling 

 3    witnesses at some point.  

 4               MR. SEGEN:  Your Honor, let me clarify what I 

 5    envision as the maximum that would occur.  It is my 

 6    understanding that in regard to the Voting Rights Act issue, 

 7    the only gap involves no statement about the impact on 

 8    minority Congressional districts.  All that that could 

 9    possibly involve would be one question and answer from 

10    Professor Cain just to fill that gap, and I suspect there 

11    are many different ways that it could be done.    

12         Furthermore, it is my understanding, both from the 

13    proceedings on May 12th and from Mr. Millet's statement 

14    today, that the issue is not being joined in regard to the 

15    impact on the Congressional districts, but rather the issue 

16    is being joined by defendants in regard to merely the issue 

17    of whether the Voting Rights Act binds the federal 

18    government.  If that is the case, I would request that they 

19    enter into a stipulation covering that.  

20               THE COURT:  Who is "they"?  

21               MR. SEGEN:  Defendants.  As I understand it from 

22    Mr. Goldin's statements to me, the transcript rings out with 

23    numerous statements to the effect that the impact -- 

24               THE COURT:  Let me interrupt you.  What is the 

25    status of Cain's transcript at the moment?  
                                                              1591

 1               MR. RIFKIND:  It was distributed in due course 

 2    after it was taken.  

 3               THE COURT:  When?  

 4               MR. RIFKIND:  Two days ago.  We got the 

 5    transcript yesterday.  

 6               THE COURT:  You have had two days to read it.  

 7    You have had two days to read his transcript.  

 8               MR. SEGEN:  Your Honor, I would indicate to you 

 9    that I spoke to Mr. Goldin yesterday.  

10               THE COURT:  Who is Mr. Goldin, which one of this 

11    great team?  Yes, okay.  

12               MR. SEGEN:  It was at about 3:00 3:30 p.m.  I did 

13    not know that Professor Cain had already testified.  He 

14    informed me, when I asked him to review the transcript, that 

15    the Corporation Counsel had not yet received the transcript 

16    but that he thought that Cravath might have the transcript.  

17    I decided that I would come down to court and make my 

18    application and then go to Cravath and see whether they now 

19    have the transcript.  However, Corporation Counsel, at 

20    least, as of 3:30 yesterday did not have it.  

21               THE COURT:  Is there any darkness you can shed 

22    about this?  

23               MR. GOLDIN:  I think it is going to be more 

24    obscurity, your Honor, but let me belabor it.  

25               I spoke to Mr. Segen yesterday.  I didn't have 
                                                              1592

 1    the transcript.  He didn't have the transcript.  He was 

 2    concerned about what might be in the transcript.  I 

 3    described generally the thrust of Dr. Cain's testimony.  I 

 4    suggested to Mr. Segen that he ought to take an opportunity 

 5    to read that transcript, that after he had reviewed it, 

 6    determine in light of the evidence that had been presented 

 7    orally and the exhibits that had been admitted, that he had 

 8    some remaining concern about the corpus of evidence he would 

 9    need to prove his separate claims, that that would be an 

10    appropriate time to raise the issue.  I specifically 

11    informed him that we were taking no position with respect to 

12    the claims that he was asserting, or the appropriateness of 

13    attempting to introduce additional evidence in light of them 

14    at this time.  

15               I had understood when we spoke that Mr. Segen was 

16    going to write a letter or make in other written statement 

17    that would be submitted to the Court simply to apprise the 

18    Court of the fact that he was continuing to review the 

19    evidence and might at some future time be taking some future 

20    step to reserve or protect his rights, and was not prepared 

21    for this morning's statement.  I don't know what the pending 

22    application is.  That is probably as much as I can do to 

23    clarify it.  

24               THE COURT:  Can we give Mr. Segen a copy of the 

25    transcript?  
                                                              1593

 1               MR. RIFKIND:  He now has it.  

 2               THE COURT:  Go out and read it and sin no more.  

 3    In about an hour or two we will talk to you again and see 

 4    where you are.  

 5               MR. SEGEN:  Thank you, your Honor.  

 6               THE COURT:  Are we ready, Mr. Sitcov?  

 7               MR. SITCOV:  Yes, sir.  

 8               THE COURT:  Dr. Fienberg, would you resume the 

 9    stand.  

10    STEPHEN FIENBERG      resumed 
                                    

11    CROSS-EXAMINATION 

12    BY MR. SITCOV:

13         Q.    Dr. Fienberg, you testified yesterday about 

14    Secretary Mosbacher's decision not to adjust the 1990 

15    decennial census.  Do you remember that?

16         A.    Yes.

17         Q.    You were talking about the Secretary's decision 

18    document which analyzes his decision on the basis of the 

19    guidelines that were adopted in connection with this 

20    litigation, is that right?

21         A.    That's correct.

22         Q.    It is your view that the adjustment plaintiffs 

23    seek satisfies the guidelines?

24         A.    The view I expressed yesterday was that the 

25    reasons that the Secretary gave for rejecting the 
                                                              1594

 1    recommendation of the census bureau were ones that I thought 

 2    were not valid.

 3         Q.    My question to you is, is it your opinion that 

 4    the adjustment that the plaintiffs are seeking satisfies the 

 5    guidelines?

 6         A.    I think that if you were to take that evidence, 

 7    put it in the context of the research program and the 

 8    evaluation done leading up to the 1990 census, and evaluate 

 9    the analyses presented by the census bureau and the other 

10    evidence to which the Secretary referred, that if I were 

11    asked to conclude if it were in accord with the guidelines, 

12    I would.  

13               THE COURT:  Would say that it did?  

14               THE WITNESS:  I would say that it did.

15         Q.    You didn't want the Court to adopt the 

16    guidelines, did you?

17         A.    I believe when the guidelines were issued that 

18    they were vaguely worded, were subject to use technical 

19    terms in ways that were hard to interpret.

20         Q.    Do you see there is a small stack of documents 

21    right there.  Could you turn them over.  Do you see that the 

22    first document is Defendant's Exhibit 901?

23         A.    Yes.

24         Q.    That is an affidavit that you prepared and it was 

25    submitted by the plaintiffs at the time they challenged the 
                                                              1595

 1    guidelines, isn't that right?

 2         A.    That is correct.

 3         Q.    You prepared this affidavit yourself, didn't you?

 4         A.    Yes, I did.

 5         Q.    Could you turn to page 6 of the affidavit.  On 

 6    page 6 you stated, did you not, under oath, that the 

 7    guidelines do not, in fact, set forth relevant technical and 

 8    statistical grounds for a decision about whether to adjust 

 9    the 1990 decennial census?  Isn't that right?

10         A.    Exactly where on the page?  I just want to make 

11    sure that I am with you.  Where on the page?

12         Q.    It is in the paragraph marked "Second."  About 

13    the sixth or seventh line, you say the guidelines as 

14    published do not, in fact, set forth relevant technical and 

15    statistical grounds for such a decision?

16         A.    Yes, I do.

17         Q.    In that same paragraph you also said, "Although 

18    the guidelines do contain a good deal of technical-sounding 

19    jargon, they do not contain anything that could properly be 

20    called technical grounds for decision on adjustment," isn't 

21    that right?

22         A.    Yes.

23         Q.    Do you see the next paragraph?  You say, "The 

24    guidelines also do not set forth nontechnical or policy 

25    grounds for the decision to adjust," isn't that right?
                                                              1596

 1         A.    Yes.

 2         Q.    Could you turn to page 7, please.  Do you see 

 3    paragraph 10, which begins at the bottom?  Do you see it?

 4         A.    Paragraph?

 5         Q.    10.

 6         A.    10, yes.

 7         Q.    You say there, "The assertion that the technical 

 8    grounds for adjustment are contained in these guidelines 

 9    (Introduction at 5) is false.  The guidelines do not specify 

10    any statistical or technical ground for adjustment," isn't 

11    that right?

12         A.    Yes.

13         Q.    Could you turn to page 9, please.  You can look 

14    at page 8, if you want, where it talks about the paragraph 

15    12, the explanation of guideline 3.  After your explanation 

16    you say, in the middle of page 9, "None of this constitutes 

17    anything that a statistician could call the technical 

18    grounds for an adjustment decision."  Do you see that?

19         A.    Yes.

20         Q.    You also testified on page 12 that the guidelines 

21    were devoid of substantive content, didn't you?  Look at 

22    paragraph 15.  Do you see it says, "Though devoid of 

23    substantive content"?

24         A.    Yes.

25         Q.    In fact, later on in that same paragraph you say, 
                                                              1597

 1    "As the guidelines currently read, however, no professional 

 2    statistician could interpret such words and phrases with any 

 3    degree of clarity," isn't that right?

 4         A.    Yes.

 5         Q.    Would you turn to page 15, please.  Do you see 

 6    paragraph 18?  There you say that, "The character of several 

 7    of the guidelines is such that no one, statistician or 

 8    otherwise, could determine whether they are satisfied or the 

 9    significance that should be attached to them," isn't that 

10    right?

11         A.    That's what I said.

12         Q.    These statements aren't true, are they?

13         A.    At the time that I wrote them, I had not seen the 

14    technical descriptions of the full set of studies that the 

15    census bureau proposed to do.  I had not seen the way in 

16    which they would be structured either by the census bureau 

17    or by the Secretary.  And I think, if read separately at the 

18    time at which I drafted the affidavit, that those statements 

19    were correct.

20         Q.    You didn't tell the Court that there was anything 

21    that you needed to see in order to make the guidelines 

22    understandable to a professional statistician, did you?

23         A.    I indicated that one needled to see the technical 

24    discussion.  I think that was in one of the earlier 

25    paragraphs that would clarify what those words meant in the 
                                                              1598

 1    context of specific statistical procedures and specific 

 2    analyses that were planned and that would be used as the 

 3    basis for carrying out the assessment vaguely indicated by 

 4    the words in the guideline.

 5         Q.    In fact, it is now your opinion that you had the 

 6    ability even then to impart that kind of knowledge, isn't 

 7    that right?

 8         A.    I'm not sure I understand the question.  Are you 

 9    asking if I had the opportunity to come and rewrite the 

10    guidelines, that I would have been able to impart what one 

11    might interpret those words to mean?

12         Q.    No.  You didn't advise the Court in your 

13    affidavit that a professional statistician with knowledge of 

14    the history of adjustment could interpret the words and 

15    phrases in the guidelines with a degree of clarity, did you?

16         A.    I indicated that if one were to take that 

17    history, to put it together with the actual plans that were 

18    made subsequent, I believe, to the publication of the 

19    guidelines in preparation of this document, of specific 

20    studies to be carried out by the census bureau to address 

21    items that were listed within the guidelines and integrated 

22    with the actual results of those studies and other documents 

23    produced and prepared by the census bureau over the 

24    intervening period, that I thought I could provide that kind 

25    of clarification.  
                                                              1599

 1               At the time I read these without that material in 

 2    hand, I think that these statements are correct and speak 

 3    for themselves.

 4         Q.    Do you have the Secretary's decision in front of 

 5    you?  It is Plaintiff's Exhibit 9?

 6         A.    No, I don't.  All right.

 7         Q.    Could you turn to page 2-25 of the Secretary's 

 8    decision.  Do you see about mid-way down on the page, in the 

 9    second full paragraph, you see the Secretary says, "I 

10    conclude that the constitutional and legal purposes for the 

11    census must take preference, and accuracy should be defined 

12    predominantly in terms of getting the proportional 

13    distribution of the population right among geographical and 

14    political units.  This argues for putting aside the judgment 

15    of accuracy based upon getting absolute numbers right 

16    (numeric accuracy) and instead focusing on the question of 

17    whether there is convincing evidence that the accuracy of 

18    popuation distribution in the adjusted numbers (distributive 

19    accuracy) is superior to the distributive accuracy of the 

20    actual enumeration.  The quality of the adjusted counts 

21    themselves must be examined to address this important issue 

22    squarely." 

23               You believe that the Secretary erred in deciding 

24    that distributive accuracy should be the sole criterion for 

25    deciding whether or not to adjust, isn't that right?
                                                              1600

 1         A.    I believe that looking at information in addition 

 2    to the information on distributive accuracy which in fact 

 3    shows that the corrected counts are superior to the raw 

 4    counts is important in coming to an assessment about the 

 5    quality of the data upon which the Secretary was relying.

 6         Q.    You are aware, are you not, that some of these 

 7    same plaintiffs sued some of the same defendants, in fact, 

 8    to have an adjustment in the 1980 census?

 9         A.    Yes, I am.

10         Q.    I am going to read you something from the 

11    decision in that case.  In that case the court said, "An 

12    adjustment action presents one issue to be resolved by the 

13    court:  Whether the plaintiffs have sustained their burden 

14    of proving that a statistical adjustment of the 1980 census 

15    will result in a more accurate picture of the proportional 

16    distribution of the population of the United States on a 

17    state-by-state and substate-by-substate basis than the 

18    unadjusted census."  That passage said the states adopted 

19    the same test for accuracy for adjustment measures as the 

20    Secretary stated on page 2-25 of his decision, doesn't it?

21         A.    Within a word or two, I guess yes.

22         Q.    So your opinion is that the Secretary erred when 

23    he adopted the same tests of accuracy as the court adopted 

24    in the Cuomo case, is that right?

25         A.    No.  I said that I think that one should, in 
                                                              1601

 1    fact, look at distributive accuracy.  But to look at that 

 2    alone, without looking at the other information which leads 

 3    up to the information on distributive accuracy, such as the 

 4    components of error and how they are represented in terms of 

 5    the differential undercount, and in particular there are 

 6    uses of census data that require numeric accuracy, that 

 7    there are other things to look at in addition.  I didn't say 

 8    that one shouldn't look at this and indeed should not do 

 9    tests that try to focus on this.

10         Q.    In fact, when I took your deposition last month 

11    and I read you that same passage from the Cuomo decision, 

12    didn't you tell me you couldn't even understand what it 

13    said?

14         A.    I don't recall the details of my remarks.  If you 

15    would like to show them to me, I would be happy to.  

16               THE COURT:  Do you understand it today?  Do you 

17    understand the language today?  

18               THE WITNESS:  Yes.  

19               THE COURT:  You are saying you didn't understand 

20    it then?  

21               MR. SITCOV:  He testified he couldn't understand 

22    it.  

23               THE COURT:  All right.  So?  Let's go on.

24         Q.    You agree, don't you thatisticistical analysis 

25    are most widely accepted when the question is defined to 
                                                              1602

 1    descriptive statistics, isn't that right?

 2         A.    That is a statement that looks back reflectively 

 3    over something like 50 to 60 years of the use of statistics 

 4    in legal proceedings.  And in the context of looking back 

 5    over the use up through the mid eighties when that statement 

 6    was drafted, the answer is yes, I do.

 7         Q.    Would you agree that the effectiveness with which 

 8    statistical knowledge can address issues in litigation 

 9    depends on the type of statistical usage?  Isn't that right?

10         A.    Sounds like a familiar phrase, sentence.  I 

11    wouldn't disagree with it.

12         Q.    You also agree that projections of future 

13    outcomes based on earlier events compound the uncertainties 

14    of causal inference with additional problems of changing 

15    conditions and factors over time, isn't that right?

16         A.    Yes, that sounds like another statement from the 

17    same source.

18         Q.    Do you agree that projections to areas outside 

19    the sample based on outcomes for areas inside the sample 

20    compound the uncertainties of causal inference with the 

21    additional problems of changing conditions and factors over 

22    space?

23         A.    That means projecting beyond the population from 

24    which one drew the sample, yes.

25         Q.    Would you agree that greater uncertainty emerges 
                                                              1603

 1    in causal inference because real-world conditions rarely 

 2    provide the equivalent of experimental control?  Isn't that 

 3    right?

 4         A.    That's correct.

 5         Q.    It is correct, isn't it, that inferences 

 6    regarding cause and effect in legal cases depend heavily on 

 7    subject matter assumptions?

 8         A.    That's correct.

 9         Q.    You agree, don't you, that only when randomized 

10    controlled experiments have been done directly on the cause 

11    and effect of interest can statistical inference provide a 

12    vehicle for avoiding substantive assumptions?

13         A.    Sorry.  Could you repeat that one?

14         Q.    You agree, don't you, that only when randomized 

15    controlled experiments have been done directly on the cause 

16    and effect of interest can statistical inference provide a 

17    vehicle for avoiding substantive assumptions?

18         A.    Even randomized trials have assumptions.  I guess 

19    if I had said that, if I had written it in that form, I 

20    probably would today disagree with it in a technical sense.

21         Q.    You did write that, didn't you?

22         A.    I believe that it is extracted in a report of 

23    which I was a member of the group that drafted the language, 

24    yes.

25         Q.    The PES was not a randomized controlled 
                                                              1604

 1    experiment, was it?

 2         A.    No.  It was a random sample selected from the 

 3    population of the United States.

 4         Q.    The P studies weren't randomized controlled 

 5    experiments, were they?

 6         A.    No, they were not.

 7         Q.    The evaluation and review the bureau conducted to 

 8    consider the accuracy of the PES was not a randomized 

 9    controlled experiment, was it?

10         A.    No, it was not.

11         Q.    It is your view, isn't it, that judges are likely 

12    to be unfamiliar with the language of statistics and 

13    statistical concepts?  Isn't that right?

14         A.    It was a statement that was part of that report 

15    and was true of a substantial number of people in the 

16    judiciary at the time it was written.

17         Q.    That was your opinion about a month ago when I 

18    took your deposition, wasn't it?

19         A.    And it is still true of many judges.

20         Q.    It is also your view, isn't it, that people 

21    without special training often have difficulty evaluating 

22    statistical data and the inferences drawn from statistical 

23    data?

24         A.    Yes.

25         Q.    It is also your view that model misspecification 
                                                              1605

 1    is often an important issue in trials of statistical 

 2    matters, isn't that right?

 3         A.    Yes.

 4         Q.    You also take the view that models selection is 

 5    often treated blithely by the courts, isn't that right?

 6         A.    I have seen several instances where that is the 

 7    case, yes.

 8         Q.    It is also true that it is your opinion that in 

 9    evaluating statistical evidence, it is important to explore 

10    the adequacy of the statistical models used, isn't that 

11    right?

12         A.    Yes.

13         Q.    It is also your view that it is important for 

14    courts to explore the adequacy of statistical models used 

15    when considering statistical issues, isn't that so?

16         A.    Yes.

17         Q.    You believe that there are many instances where 

18    courts fail to understand the import of statistical 

19    testimony, isn't that right?

20         A.    Yes.

21         Q.    It is also your view that a close look at how the 

22    courts deal with statistical evidence would reveal 

23    considerable confusion and flawed judgments, isn't that 

24    right?

25         A.    I believe that that statement was made in 
                                                              1606

 1    reference to some specific cases in the context of the 

 2    report in which it was included and comes as part of a 

 3    section that, in fact, draws on those case studies.  In 

 4    those case studies that was, in part, the issue.

 5         Q.    You also believe that it is important for 

 6    statisticiants to highlight the pivotal assumptions in their 

 7    work as expert witnesses, isn't that right?

 8         A.    Yes.

 9         Q.    It is also your view that in their work as expert 

10    witnesses, statisticians often don't do that, isn't that 

11    right?

12         A.    Those same case studies presented examples of 

13    that, yes.

14         Q.    You testified yesterday about adjustment at some 

15    length, didn't you?

16         A.    Yes.

17         Q.    You testified about all the things that were 

18    entailed in adjustment and that worked well, including the 

19    smoothing process, isn't that right?

20         A.    That is correct.  

21               MR. SITCOV:  Your Honor, may I enter the well for 

22    a moment?  

23               THE COURT:  Yes.

24         Q.    I am going to put this up here.  Can you see it, 

25    Dr. Fienberg?
                                                              1607

 1         A.    Yes, I can.

 2         Q.    That is a reproduction, Dr. Fienberg, from 

 3    Defendant's Exhibit 66.  Do you recall that I asked you 

 4    about that document at your deposition?

 5         A.    Yes, I do.

 6         Q.    It is your opinion, isn't it, that it would be 

 7    extremely difficult for someone who is not an expert in 

 8    statistics to understand this document?  Isn't that right?

 9         A.    I think that is a reasonable description, yes.

10         Q.    The smoothing process uses a model, doesn't it?

11         A.    Yes, it does.

12         Q.    The model relies on stochastic assumptions, 

13    doesn't it?

14         A.    The specification of the model does rely on 

15    stochastic assumptions, yes.

16         Q.    You take those assumptions seriously, don't you?

17         A.    I take the implications of those assumptions 

18    seriously, yes.

19         Q.    Does that mean you don't take the assumptions 

20    seriously?

21         A.    As I indicated in my direct testimony, no 

22    assumptions are strictly correct.  All statistical models 

23    are approximations to real-world phenomena.  It is in the 

24    context in which the models are applied and the implications 

25    of the assumptions for the conclusions and for the analyses 
                                                              1608

 1    that one does, that is what one needs to look at.  And there 

 2    are a variety of ways in which statisticians examine them.

 3         Q.    When I took your deposition last month, you 

 4    weren't able to tell me whether or not this was, in fact, a 

 5    1990 production smoothing model, were you?

 6         A.    I believe I pointed out that the specification as 

 7    it was listed on that particular page was a form of linear 

 8    regression model -- actually, there is a pair of equations 

 9    put together -- and that without explicit detail, it was a 

10    set of general equations that looked like ones, but I 

11    couldn't be sure that it was exactly what was used, because 

12    X and beta were not completely specified and there were 

13    other aspects that were not included there.

14         Q.    You couldn't recognize it as the 1990 production 

15    smoothing model?

16         A.    I recognize it as a set of equations that look 

17    like other equations that I have seen in a variety of 

18    settings.  Those are linear equations.  Essentially 50 

19    percent of what one would find in a statistical journal 

20    would be equations that looked something like that. 

21         Q.    You know now that that is the 1990 production 

22    smoothing model, don't you?

23         A.    That combined with the model that produces the 

24    variances, yes.

25         Q.    You see four lines up from the bottom there is an 
                                                              1609

 1    equation that said big Y equals little y plus e; do you see 

 2    that?

 3         A.    That's correct.

 4         Q.    At your deposition I asked you to define the 

 5    terms in that equation, didn't I?

 6         A.    Yes.

 7         Q.    I asked you whether Y represents the true 

 8    adjustment factors for poststratum, and you told me that you 

 9    didn't even know what the words "true adjustment factor" 

10    mean, isn't that right?

11         A.    In the context of how the words were going, I may 

12    have said that, yes.

13         Q.    I asked you if e represented sampling error, and 

14    you said you wouldn't characterize it that way, isn't that 

15    right?

16         A.    That's correct.  In the specification of a 

17    regression equation written out in the form that it is in 

18    this particular way, e represents departures from the 

19    specification that capital Y equals little y; that is, e 

20    represents, as we usually label it, error; that one and 

21    perhaps major component of that error was in fact sampling 

22    variability; but that I couldn't, without looking at the 

23    context and very explicitly at the context in which it was 

24    being applied with data, attribute it solely to sampling 

25    error.
                                                              1610

 1         Q.    Were you here for Dr. Eugene Ericksen's 

 2    cross-examination?

 3         A.    For a part of it.

 4         Q.    Were you here when I cross-examined him about 

 5    this chart?

 6         A.    I'm not certain if I was.

 7         Q.    If Dr. Ericksen testified that little e 

 8    represents sampling error, would he be wrong? 

 9         A.    I am attempting to give a precise answer, and 

10    perhaps he was giving a more general one.  I think that the 

11    issue here is how one chooses to view that equation, whether 

12    you view it as a set of statistical equations, whether you 

13    attempt to explain the error itself, what meaning you would 

14    attach to it.  It is part of the process in going from 

15    formal statistical representations to the application of 

16    them in the real world.  People choose to use somewhat 

17    different words in describing it.  I wouldn't characterize 

18    it as wrong.

19         Q.    Can you look at the material immediately to the 

20    right of the equation that I have just been asking you 

21    about.  It reads, e with a squiggly line N paren zero V sub 

22    3 end paren.  Do you see that?

23         A.    Yes.

24         Q.    At your deposition I asked you if that assumption 

25    entailed that the dual system estimate is unbiased, and you 
                                                              1611

 1    told me that you didn't even know what the question meant, 

 2    isn't that right?

 3         A.    The problem is that the dual system estimate is 

 4    the product of something like 20 or 30 component processes.  

 5    This is an equation that, when used with the actual data and 

 6    the production of the post-enumeration survey, is one 

 7    component in the middle of that.  

 8               Indeed, this isn't even the end of the smoothing 

 9    process.  There is another step after the application of 

10    this where the raw factors are then weighted with the 

11    so-called smooth factors that come from these equations.  

12    The production dual systems estimate which comes out the 

13    other end is something which is known to have a variety of 

14    components of bias.  That is what the whole set of P studies 

15    and the evaluations were all about.  

16               To say that that assumption, therefore, makes the 

17    dual systems estimate unbiased I didn't think was an 

18    appropriate characterization of it.  Rather, I believe it 

19    says what it says, and you need to look at the implications 

20    of that carried through to the end.

21         Q.    My question to you is, when I asked you at your 

22    deposition if that assumption entailed a dual system 

23    estimate is unbiased, didn't you respond by saying that you 

24    don't know what that question means in the context of this 

25    model?
                                                              1612

 1         A.    Because the model is taken out of the full dual 

 2    system production process, and therefore this alone, even if 

 3    it were correct, would not imply that the dual system 

 4    estimator is unbiased.  That is what I just said.  

 5               (Continued on next page) 

 6    

 7    

 8    

 9    

10    

11    

12    

13    

14    

15    

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1613

 1    BY MR. SITCOV:

 2         Q.    If Dr. Ericksen testified that that equation 

 3    entails that that dual system estimator is unbiased, would 

 4    you have been able to understand his testimony?

 5         A.    I would have had to hear the testimony in the 

 6    context in which he was making the statement.

 7         Q.    The equation on the fourth line from the bottom 

 8    implies that the raw adjustment factors derived from the 

 9    dual system estimate are unbiased and that assumption is 

10    false, isn't it?

11         A.    It isn't exactly correct, that's correct.
                                                          

12         Q.    And the assumption e squiggle line and N (O v sub 

13    3), doesn't that entail that the standard error in the raw 

14    adjustment factors follows the normal distribution?

15         A.    Yes, it does.

16         Q.    And that assumption does not hold true in any 

17    direct and absolute way for the 1990 PES data, does it?

18         A.    I believe it does not.

19         Q.    Can you look at the next line on that page, the 

20    one that reads Y equals X Beta plus v where v as a normal 

21    distribution O sigma squared times identity?

22         A.    Yes.

23         Q.    Now, X is a matrix of explanatory variables, 

24    isn't that right?

25         A.    Yes, it is.
                                                              1614

 1         Q.    And Beta is a vector of unknown parameters, isn't 

 2    that right?

 3         A.    That's what it says in the line below.

 4         Q.    And v is a disturbance term?

 5         A.    An error term, yes.

 6         Q.    And the material to the right on the right side 

 7    of that equation, that is, v equals N sigma squared I is an 

 8    assumption, isn't it?

 9         A.    Yes, it is.

10         Q.    And it assumes that the components of v have mean 

11    zero, is that right?

12         A.    Yes, it does.

13         Q.    And assumes that the components of v --

14         A.    It does in large part, because, in fact, the mean 

15    becomes part of X times Beta and that's the standard way of 

16    writing such equations.

17         Q.    And it assumes that the --

18         A.    So it isn't an assumption, it's part of the 

19    specification and, in fact, the mean is in that equation 

20    somewhere else in the formula.

21         Q.    Now, that assumption also assumes that the 

22    components of v have constance variance equal to sigma 

23    squared, right?

24         A.    Yes, it does.

25         Q.    It also assumes that the components of v are 
                                                              1615

 1    independent, isn't that right?

 2         A.    Yes, it does.

 3         Q.    And that the components of v have normal 

 4    distribution?

 5         A.    Yes.

 6         Q.    And there is also an assumption that e and v are 

 7    independent, isn't that right?

 8         A.    Yes.

 9         Q.    And at the time I took your deposition last 

10    month, you weren't able to tell me whether the assumptions 

11    in this equation we have just been talking about were 

12    realistic or warranted, were you?

13         A.    No, I was not in the sense that I hadn't examined 

14    a number of the documents that related to what the Census 

15    Bureau had done in looking at the smoothing model and 

16    deciding on its adequacy.

17         Q.    And you don't know whether the production 

18    smoothing model we have just been discussing was ever field 

19    tested before it was used to produce adjusted data from the 

20    1990 PES, isn't that right?

21         A.    If you look at it in the form that it is 

22    specified here, I think it's reasonable to say that it, in 

23    fact, was tested in the St. Louis test dress rehearsal and 

24    some other variant form it was tested in the Los Angeles 

25    test of adjustment related operations, absent some other 
                                                              1616

 1    tests as well, but those are the two for which I've seen 

 2    some data.

 3         Q.    My question is, was this particular smoothing 

 4    model ever field tested before it was used in the 1990 

 5    production PES?

 6         A.    What does "this particular" mean, because those 

 7    are general linear equations for which there are X's and 

 8    Betas and v's and y's and e's and equations exactly like 

 9    those were used in those two other field tests.

10         Q.    Do you recall that at your deposition I asked you 

11    the following question.

12               "I am asking you if the Bureau has ever tested 

13    this model to see whether it works accurately before it was 

14    used in the production of the dual system estimate?" 

15               Your attorney objected to the form of the 

16    question.

17               And you said, "if you are referring to the exact 

18    equations, the exact coefficients, the same predictor 

19    variables and identical structure applied to a previous 

20    field test, then I have no knowledge of such a test." 

21               Do you recall giving that testimony on page 308?

22         A.    That sounds like something I probably would have 

23    said if we were talking about the specific coefficients and, 

24    in fact, it's always the case that if you fit a model, even 

25    if it's fully prespecified in terms of what variables are 
                                                              1617

 1    included and a variety of relationships not unlike the ones 

 2    that we have considered here and you applied them in real 

 3    world context and got specific coefficients, they wouldn't 

 4    be the same as the ones you got in some other test you 

 5    applied before, and that's what that statement says and it's 

 6    correct.

 7         Q.    It's the case, isn't it, that smoothing could 

 8    introduce bias into the undercount rates?

 9         A.    It is possible, yes.

10         Q.    It's also your opinion, isn't it, that any verbal 

11    explanation of a term in a mathematical formula tends to 

12    misrepresent what it may or may not be?

13         A.    I think that's often the case, yes.

14         Q.    In fact, you think that such labels or 

15    descriptions for a component in a mathematical equation used 

16    to develop a method of analysis have little direct meaning 

17    and interpretation and, therefore, you wouldn't use such 

18    terms, would you?

19         A.    In the context of working through a model and 

20    trying to isolate components, I try to avoid that wherever 

21    it's possible, but, like all other statisticians and others 

22    who use statistical models, I occasionally lapse and I may 

23    have in this discussion.

24         Q.    Variance presmoothing is not a commonly used 

25    standard procedure, is it?
                                                              1618

 1         A.    Variance presmoothing is the fitting of linear 

 2    models to variances.  That is something that lots of people 

 3    do, the Census Bureau itself has done in a variety of 

 4    settings.

 5               Indeed, all of the variances that the Census 

 6    Bureau fits, uses in every report it issues for every 

 7    national survey is a result of fitting linear models to 

 8    sampling variances calculated from complex formulas, and 

 9    usually there is an appendix at the back that describes what 

10    they call the generalized equations for variances, which 

11    come out of models that look like the presmoothing model.

12         Q.    At your deposition you couldn't give me a single 

13    example of presmoothing in the statistical literature, could 

14    you?

15         A.    You are referring explicitly to the use of the 

16    smoothing of variances in the context of this particular 

17    model in this particular kind of adjustment context.  I 

18    described to you the uses of smoothing of variances, which 

19    is common in a variety of places in the statistical 

20    literature, including at the Census Bureau.

21               This particular model combines the smoothing of 

22    variances with the smoothing of the adjustment factors and I 

23    believe that my deposition, although I would have to look at 

24    the details to be sure, I actually --  you asked me if I had 

25    any other example that was done in a large-scale sample 
                                                              1619

 1    survey or in a sample survey and I could not recall one of 

 2    those.

 3         Q.    Didn't I ask you if you could give me some 

 4    examples of presmoothing and you said explicitly no?  

 5               MS. BARRY:  Your Honor, I request that Dr. 

 6    Feinberg be able to see the transcript of his deposition and 

 7    to respond to this question.  

 8               THE COURT:  I can't hear you.  

 9               MS. BARRY:  I request that we permit Dr. Feinberg 

10    to look at his deposition, because I believe Mr. Sitcov has 

11    admitted part of his answer related to methodological 

12    literature which relates to dampening weights used when the 

13    weights reflect variances, and I believe that he should be 

14    able to see his transcript so that he can fairly answer this 

15    question, because I believe Mr. Sitcov is somewhat 

16    misrepresenting what went on at the deposition.  

17               THE COURT:  You wouldn't do that, would you?  

18               MR. SITCOV:  I would never do that.  I'm from the 

19    government.  I'm here to help, your Honor.  

20               THE COURT:  Do you have a copy to give the 

21    witness?  

22               MR. SITCOV:  Sure, I absolutely do.

23               (Handing to the witness)

24               (Pause) 

25         Q.    You have never used presmoothing --  
                                                              1620

 1               MS. BARRY:  The page reference is around 196 to 

 2    198.

 3         Q.    You haven't ever presmoothed yourself, have you?  

 4               MR. SITCOV:  I don't know if the syntax is right.

 5         Q.    You have never used presmoothing, have you?

 6         A.    Are we doing this in a technical or untechnical 

 7    context?

 8         Q.    Either way.

 9         A.    I have done something that might be roughly akin 

10    to that actually going back to work that was related to my 

11    doctorate dissertation, but it wasn't called presmoothing 

12    and the specifications were somewhat different.  That was in 

13    the context of smoothing death rates in the study of 

14    anesthetics across many hospitals and there were things that 

15    were not unlike this.

16               I'm sorry, I'm still looking for the passage that 

17    we were working on before.

18         Q.    Can you turn to page 196, please.

19               Do you recall that I asked you at your deposition 

20    on line 22 if you ever personally presmoothed variances in 

21    surveys you conducted, and you respond, "I haven't conducted 

22    my own surveys now for a number of years, and the ones I did 

23    didn't have use for this particular kind of methodology." 

24               Does that refresh your recollection that you 

25    testified at your deposition that you had never personally 
                                                              1621

 1    used variance presmoothing yourself?

 2         A.    That is not what I said.  What I said was that I 

 3    hadn't used them in the context of surveys that I myself had 

 4    conducted.

 5               I just described a large-scale study, the 

 6    National Halothane studies in which techniques not unlike 

 7    presmoothing were used in different ways.

 8               It wasn't a large-scale national survey, number 

 9    one, it was a different kind of study, and, number two, it 

10    wasn't one that I had headed myself, I was working on a 

11    piece of analysis with a large number of other statisticians 

12    and anesthesiologists.

13         Q.    So the answer to my question is, you have never 

14    used presmoothing myself?

15         A.    That isn't true.

16               What I have said repeatedly is that I haven't 

17    used it in this particular way in the context of analyzing 

18    my own surveys or in the context of analyzing large-scale 

19    surveys.

20               I have examined the use of models like this in 

21    other settings, I have read papers that describe them and --  

22               THE COURT:  The question is, have you used it? 

23               THE WITNESS:  This particular --  the question is 

24    this particular model on this --  

25               THE COURT:  Let's start with, have you ever used 
                                                              1622

 1    presmoothing? 

 2               THE WITNESS:  I have used something that one 

 3    might describe as presmoothing in that earlier setting.

 4         Q.    So was it correct when you said that, "I didn't 

 5    have use for that particular kind of methodology" when I 

 6    asked you if you had ever presmoothed variances?

 7         A.    Yes, it was correct, because that study was not a 

 8    large-scale national survey and it was not a survey that I 

 9    had carried out myself.  Those were the words that you just 

10    repeated to me that were part of the question and I was 

11    trying to answer the question that you asked.

12         Q.    As far as you know, no Census Bureau studies, 

13    other than the 1990 PES, used variance presmoothing as part 

14    of their standard preliminary reporting, isn't that right?

15         A.    As part of their standard preliminary reporting.

16               I think in that narrow context the answer is 

17    probably yes.

18         Q.    And you are familiar with the proposal that Dr. 

19    Ericksen and Joseph Kadane made to adjust the 1980 decennial 

20    census, aren't you?

21         A.    Yes.

22         Q.    And they didn't propose presmoothing variances in 

23    their model, did they?

24         A.    No, they did not.

25         Q.    Do you know how many P studies the Census Bureau 
                                                              1623

 1    did?

 2         A.    Oh, it's like another answer I think I gave.  It 

 3    depends how you count.

 4               They go up to P18, but there are A's and B's, so 

 5    I can't keep track of the full number.

 6               I think it's probably 20 or 21, 22, depending on 

 7    how you count, but the numbers go up to 18.

 8         Q.    None of the P studies considered the quality of 

 9    the variance presmoothing process, did they?

10         A.    In P16, which is the total error model, there is 

11    a component that relates to that.

12         Q.    But it didn't consider the quality of the 

13    presmoothing process, did it?

14         A.    It depends by what means, what one means by 

15    quality as to whether or not the answer to that is yes or 

16    no.

17                Yes, it included it.  It included an added 

18    component --  components of variance and it did it for each 

19    of the 13 evaluation post-strata and that was represented as 

20    part of the overall total error model.  In that sense it 

21    did.

22         Q.    The P16 didn't actually conduct studies of how 

23    well the variance presmoothing was done, did it?

24         A.    Explicitly as such, no, but it needed to go to 

25    such studies to gain some steps of those variance components 
                                                              1624

 1    that went into it.

 2         Q.    In fact, there is no P study that specifically 

 3    considered the quality of the presmoothing process itself, 

 4    isn't that right?

 5         A.    That's correct.  There are other documents that 

 6    describe that.

 7         Q.    And there are no P studies that considered the 

 8    quality of the smoothing process itself, are there?

 9         A.    No.  Again, there are other documents that 

10    attempt to describe that.

11         Q.    And there were no P studies that considered the 

12    reliability of the variance covariance matrix for the smooth 

13    adjustment factors, are there?

14         A.    No, but the other studies, again, attempted to 

15    look at the impact of the ensemble of assumptions in the 

16    form of the smoothing model.

17         Q.    And you are aware that the Bureau used a variant 

18    of a jackknife to estimate the raw variances from the survey 

19    data, isn't that right?

20         A.    I believe so, yes.

21         Q.    And no P study considered the quality of those 

22    estimates, did they?

23         A.    Not that I know.  That is, I don't recall any of 

24    that being part of the material in the P studies, that's 

25    correct.
                                                              1625

 1         Q.    Could you turn to Plaintiff's Exhibit 28, which 

 2    is one of the books that you were looking at yesterday.

 3               (Pause) 

 4               MR. SITCOV:  I am hearing the plaintiffs are not 

 5    going to make available to the witness the exhibit he was 

 6    given yesterday.

 7               I only have one.  I will be happy to step up 

 8    and --  

 9               THE COURT:  You must have misheard.  

10               MR. SITCOV:  Yes, I must have.

11               (Handing to the witness)

12               (Pause)

13         A.    I guess I did have it.  It had a different label 

14    on it in the version that you gave to me.

15         Q.    This is a press release that the Census Bureau 

16    issued on June 13, 1991, and you testified about this 

17    yesterday.  Do you recall?

18         A.    Yes.

19         Q.    Could you turn to table 1.  It's after the fifth 

20    page of proceeds.

21         A.    Yes.

22         Q.    There are a bunch of columns and a bunch of rows.

23               You see a column that says preliminary PES?

24         A.    Yes.

25         Q.    And the preliminary PES is the raw dual system 
                                                              1626

 1    estimate without smoothing or presmoothing, isn't that 

 2    right?

 3         A.    I would have to read the footnote.

 4               It says no smoothing of variance covariance 

 5    matrix, so it has no, no presmoothing I think is a 

 6    reasonable characterization.

 7         Q.    And then there is a column that says, "Selected 

 8    method." 

 9               Do you see that, and that is the method that was 

10    actually used, the one that includes smoothing and 

11    presmoothing, is that right?

12         A.    Yes.

13         Q.    Now, both of those methods give the same 

14    estimated net national undercount rate of 2.1 percent, isn't 

15    that right?

16         A.    Yes.

17         Q.    And that's because of benchmarking, isn't it?

18         A.    I can't tell you why, because I haven't looked at 

19    the calculations of those specific numbers in any way that I 

20    recall.

21               I'm sure that I may have, but I don't know at 

22    this moment what the distinction is.

23         Q.    You have in front of you, Dr. Fienberg, the 

24    document that has already been entered into evidence, it is 

25    Plaintiff's Exhibit 142.
                                                              1627

 1         A.    I'm sorry.  142?

 2         Q.    Yes.

 3               No, it's in the little pile.

 4               Do you see that?

 5         A.    The pile contains two items.  There is no 142.  

 6               MR. SITCOV:  May I approach the witness, your 

 7    Honor?  I think I put it up there.  

 8               THE COURT:  Yes.

 9         A.    There are only two documents up here.

10               Oh, I'm sorry, there are things in the back.

11               I got them.  They were stuck to the other 

12    document.

13         Q.    You have seen this letter before, haven't you?

14         A.    Yes.

15         Q.    On page 2, there is a table and that table shows 

16    the effect of presmoothing on the net undercount as 

17    determined by the PES, isn't that right?

18         A.    Yes.

19         Q.    And that table shows that without presmoothing, 

20    the PES shows a net national undercount of 1.15 percent, 

21    isn't that right?

22         A.    That's what that states.

23         Q.    But smoothing and presmoothing causes the net 

24    national undercount --

25         A.    I'm sorry.  The raw variances, it says --  oh, I 
                                                              1628

 1    guess that's what that means.

 2         Q.    Okay. 

 3               The table shows that without presmoothing, the 

 4    PES shows a net national undercount of 1.15 percent, but 

 5    with smoothing and presmoothing, the net national undercount 

 6    rate is 2.1 percent, isn't that right?

 7         A.    That's correct.

 8         Q.    Now, you don't know whether the.95 percent 

 9    difference is statistical significant, do you?

10         A.    I'm sorry, could you repeat the question?

11         Q.    Well, when I took your deposition, did you --

12         A.    I didn't hear the question.  I would just like 

13    you to repeat it so I can answer it.

14         Q.    And I'm happy to repeat it.  

15               When I took your deposition, didn't you a say you 

16    didn't know whether the .95 difference between the presmooth 

17    and smooth undercount or the smoothing without presmoothing, 

18    you didn't know whether that difference was statistical 

19    significant, did you?  

20               MS. BARRY:  Do you have a page number?  

21               MR. SITCOV:  Page 216.

22         A.    I said I didn't know.

23         Q.    And you don't know whether the .95 percent 

24    difference can be explained by sampling error alone, do you?

25         A.    I, I also said that, but I also said that I 
                                                              1629

 1    thought that it was irrelevant, because in essence the net 

 2    national undercount of 2.10 or whatever other number one 

 3    might shift it to disguises the distribution of that 

 4    undercount across the population, and in particular here I 

 5    would have been looking more directly at the differences and 

 6    asking whether they remain different from zero, that is, the 

 7    differences between the minority and the nonminority 

 8    evaluation post-strata.  They continue in these numbers, in 

 9    fact, to appear to be quite different with those for the 

10    minority areas, much higher than those for the nonminority 

11    areas.

12         Q.    Isn't it a fact that it is your opinion that 

13    there isn't enough information in that letter for an expert 

14    in statistics to know what presmoothing was doing?

15         A.    I think that an expert in statistics would need 

16    to go back and look at the evaluations that the Bureau did, 

17    to look at the justification for including or not including 

18    it, and if one did, which I have attempted to do, and traced 

19    it from the initial discussions of the St. Louis test dress 

20    rehearsal where the Census Bureau identified the possible 

21    need for introducing presmoothing and on which it carried 

22    out extensive tests and looked at its attempts, in fact, to 

23    develop this model in light of that information and the 

24    evaluations they did as they examined the data and looked at 

25    alternate specifications throughout the months of the spring 
                                                              1630

 1    of 1991, especially in May and June, that one could come to 

 2    that conclusion --  to a conclusion about the question you 

 3    asked me.

 4         Q.    Do you recall that on page 220 of your deposition 

 5    I asked you.

 6               "Q.    And you are an expert in statistics, is 

 7    that correct?

 8               "A.    That's correct.

 9               "Q.    And you can't tell from this, and we were 

10    talking about this table, whether the presmoothing was 

11    actually doing what they said it was doing, is that right?

12               "A.    From that paragraph alone, no." 

13               Does that refresh your recollection that you 

14    testified that you couldn't tell from that paragraph what 

15    presmoothing was doing?

16         A.    That's what I said, and I explained how I would 

17    tell, and I have since gone and looked at that.

18         Q.    The Secretary isn't an expert in statistics, is 

19    he?

20         A.    I don't think so.

21         Q.    After the census and the PES were concluded and 

22    production followup was finished, there were still people in 

23    both the E and P samples whose match or correct enumeration 

24    status remained unresolved, isn't that right?

25         A.    Yes.
                                                              1631

 1         Q.    And you have no idea, do you, whether the 

 2    unresolved in the PES are more like those who were resolved 

 3    in evaluation followup or whether they are more like those 

 4    who remained unresolved after evaluation followup, isn't 

 5    that right?

 6         A.    I think one would have to look at the detailed 

 7    numbers to try to make an assessment of that.

 8         Q.    Do you recall testifying at your deposition that 

 9    you didn't know the answer to that question?

10         A.    That's consistent with the answer that I just 

11    gave, that is, on the basis of that information alone, I 

12    would need to look at other information in order to make a 

13    conclusion about that.

14         Q.    You don't know how many E-sample unresolved cases 

15    there were, do you?

16         A.    I do not recall.  I've read whatever the number 

17    was in the reports, but I don't remember the numbers today.

18         Q.    And you don't know how many P-sample cases there 

19    were, do you?

20         A.    I do not recall that number, either.

21         Q.    Would it surprise you to learn that the total 

22    number of unresolved cases in the P and E samples is larger 

23    than the net undercount?

24         A.    If you were to stipulate that they were and were 

25    reading it out of a table, I would have no reason 
                                                              1632

 1    necessarily to dispute it, but like the information about 

 2    other rates, to make comparison with that on the order --  

 3    to say that they are on the order of magnitude of the net 

 4    national undercount is to be quite misleading.  The question 

 5    is how were they addressed and what do we know about them 

 6    and what is their impact.

 7         Q.    When I took your deposition last month, you 

 8    didn't know how the unresolved cases were resolved, did you?

 9         A.    I believe I probably said that, and I could not 

10    today walk you through the full process that was described 

11    in the P studies which has a number of components in it.  I 

12    know in some general sense.

13               We have the April census, April 1, the PES occurs 

14    roughly in June or July, then there is the linking of the 

15    two that occurs in the fall and the resolution that would 

16    have been part of the evaluation followup study occurs 

17    somewhere around February 1991, roughly a year after the 

18    original enumeration, and there are a number of steps in 

19    between that are detailed technical steps of what the Census 

20    Bureau did in attempting to address those.

21               So in that context I guess I can be slightly more 

22    specific.

23         Q.    But you didn't even know when I took your 

24    deposition that those unresolved cases were resolved on the 

25    basis of mathematical models, did you?
                                                              1633

 1         A.    I'm sorry, what was the page you were referring 

 2    me to?

 3         Q.    Oh, I wasn't, your attorney was.

 4         A.    I would have to go back and look at what question 

 5    you asked and how, but if you had asked me about explicit 

 6    mathematical models or if they were used, I probably would 

 7    not have recalled the details.

 8               There are a lot of details in that study, in the 

 9    study of that component of the P studies and even having 

10    read it more recently I still can't remember them all.

11               I have read it, I have gone through the 

12    components and I do not recall today explicitly what 

13    mathematical or statistical models were used.

14               As I have indicated, they are used in virtually 

15    all statistical calculations in some form or another, and so 

16    it would not surprise me to find them there, but I do not 

17    know.

18         Q.    You don't even know whether either of the models 

19    have ever been used in production before they were used in 

20    the 1990 PES, isn't that right?

21         A.    I'm not sure I understand the question.

22               You mean a specific model that I haven't been 

23    able to identify? 

24               No, I don't know if that model which I couldn't 

25    identify is, in fact, something that was used before.  It 
                                                              1634

 1    might well have, it might not.  I would have to be pointed 

 2    to the model and I'd have to look at it.

 3               The matching process is a very detailed process.  

 4    It has lots of components, all of which are designed to 

 5    produce as accurate information as possible.

 6               Whether or not the Census Bureau introduced 

 7    additional refinements in it since the last time that it had 

 8    used it in a full-scale production nationally, which would 

 9    have been 1980, I do not know.  I don't even recall for each 

10    detail whether it was used in the St. Louis test, the Los 

11    Angeles test or any other model.  

12               THE COURT:  Shall we take a mid-morning?  

13               MR. SITCOV:  I would be happy to, your Honor.  

14               THE COURT:  Let's resume at 11:15.

15               (Recess)

16               THE COURT:  Have we resolved the matter we began 

17    with early this morning with Dr. Cain and Mr. Segen?  

18               MR. SEGEN:  Your Honor, MAY I address that point?  

19               THE COURT:  Yes.  

20               MR. SEGEN:  I have reviewed the transcript of 

21    distribution testimony and the testimony indicates that 

22    there is a dilution of the vote of minorities and that it 

23    results in underrepresentation for minorities.

24               What it does not state is one component of a 

25    prima facia case under the Voting Rights Act, and that is 
                                                              1635

 1    that on some level, municipal, state, even board of 

 2    education, that a minority, an additional minority district 

 3    would be created.  The question and testimony stopped just 

 4    short of that.

 5               I have been sensitized by reading this transcript 

 6    as to the issues before the court and all of the manifold 

 7    problems and I would reduce and narrow my request in regard 

 8    to the court keeping open plaintiff's case to just that one 

 9    statement, preferably by Professor Cain, preferably by 

10    stipulation, although first indications from defense is that 

11    they are not inclined to stipulate to anything along those 

12    lines even if it was as broad as adjusted statistics were 

13    used that somewhere in this great nation of ours at some 

14    political level that is actionable under the Voting Rights 

15    Act that an additional minority district would be created.

16               That is all that I am looking for.  I am 

17    relatively certain that it can be done in a 175 different 

18    ways other than bringing Professor Cain back or having to 

19    bring in, you know, some other witness who can testify to 

20    what I consider to be palpably obvious.  

21               THE COURT:  I have no problem with your doing it 

22    in any one of those 175 ways, which means we don't have to 

23    call any witnesses.  If you want to do it in some other 

24    fashion, that is fine by me.  

25               MR. SEGEN:  Your Honor, I am open to suggestion 
                                                              1636

 1    as to how to do it.  

 2               THE COURT:  I don't conceive that to be my 

 3    obligation.  

 4               MR. SEGEN:  Well, if I can not find any other 

 5    recourse, would you allow, for instance, for a telephone 

 6    conference with Professor Cain with the court, plaintiff and 

 7    defendant present so that one question could be asked by me 

 8    or by plaintiffs' committee of Professor Cain and then 

 9    defendant could have cross-examination? 

10               I do not want to put anyone through any expense, 

11    I do not want to take up the time of the court, I'm trying 

12    to be as reasonable as my mind allows me to be.

13               I'm certain I may be missing some other possible 

14    permutations.  

15               THE COURT:  Let me hear from the government.

16               What about the telephone conference notion?  

17               MR. MILLET:  Your Honor, first of all, I don't 

18    understand how Hudson County has the standing to assert 

19    whether some minority district could be traded some level of 

20    government somewhere else in the government.

21               Putting that one to one side, it is rather late 

22    in the day to raise new issues, raise new evidentiary 

23    issues.  

24               I don't think the issue is necessarily as simple 

25    as Mr. Segen is proposing it, since it involves the actions 
                                                              1637

 1    of third parties who are not before the court as to what 

 2    legislatures or what other redistricting authorities are 

 3    going to do.

 4               If this was an issue that they wanted to raise, 

 5    they were required to give us notice of it, they were 

 6    required to tell us what their witnesses were, we would have 

 7    been allowed to have an opportunity to depose them.

 8               It is simply too late in the day to do this.  

 9               THE COURT:  What is wrong with asking Professor 

10    Cain the solitary question that he wants to put to him?  

11               MR. MILLET:  I'm not sure it's a solitary 

12    question, and we certainly have not been on notice that this 

13    was going to be an issue that was going to be litigated here 

14    and we have not had the opportunity to engage in any 

15    discovery of Professor Cain on that issue.

16               I mean, if Mr. Segen knew this was an issue he 

17    wanted to raise, it's rather late on the last day of 

18    plaintiff's case to decide they want to raise new issue, new 

19    testimony apparently because they haven't been that vigorous 

20    in asserting their rights so far in the case.

21               I do think at this point it is prejudicial to the 

22    defendants.  

23               THE COURT:  Mr. Zimroth, Mr. Rifkind, do you have 

24    any input here?  

25               MR. RIFKIND:  We will be happy to cooperate in 
                                                              1638

 1    any way the court would like us to cooperate.

 2               The only observation I have is I have imposed 

 3    rather heavily on Professor Cain, who has taken --  and I 

 4    don't know that I can procure him again --  

 5               THE COURT:  On the phone? 

 6               I have no notion in the world of calling him 

 7    back.  

 8               MR. RIFKIND:  I am prepared to try.  I have no 

 9    notion whether he has the answer to the question and no 

10    notion as to whether he is prepared to give the answer, but 

11    if I am asked to try to find him on the phone, I will 

12    certainly do that.  

13               THE COURT:  So it then boils down to the question 

14    of prejudice to the government and the only prejudice they 

15    can elicit is they haven't deposed Dr. Cain on this 

16    particular issue.  

17               MR. SEGEN:  May I address that point?  

18               THE COURT:  Yes.  

19               MR. SEGEN:  Two things:

20               The government argues that it has had some no 

21    notice.  I believe it has had notice.  Hudson County's 

22    complaint in intervention clearly stated the Voting Rights 

23    Act claim.

24               Furthermore, on May 12, when I came here, I very 

25    clearly stated that there were two claims that Hudson County 
                                                              1639

 1    wanted to preserve.  The second of those claims was the 

 2    Voting Rights Act claim.

 3               What I am asking for is one of the elements of 

 4    that Voting Rights Act claim.

 5               Furthermore, it was clear that my duty was to 

 6    monitor the proceedings in this court to make certain that 

 7    nothing slipped through the cracks.  I believe I have taken 

 8    that duty to heart, I have done everything that I could do 

 9    in that regard.  Professor Cain --  

10               THE COURT:  You might have showed up to hear 

11    Cain's testimony.  

12               MR. SEGEN:  Okay, your Honor, I stand corrected 

13    on that.

14               I must confess to you, your Honor, that I had no 

15    idea that Professor Cain would be testifying so quickly.  

16    That's what I meant earlier in the day when I said that 

17    things have gone very quickly.  You know, I should have been 

18    here.

19               However, your Honor, my intent was to have a 

20    meeting with the plaintiffs' committee and I have tried to 

21    have that meeting, although it has been difficult to arrange 

22    because of their schedules in regard to preparing for trial.

23               At that meeting I would have made clear the kind 

24    of testimony that I would have required that I needed from 

25    Professor Cain, although I think it's obvious.  
                                                              1640

 1               THE COURT:  Where is Professor Cain right now?  

 2               MR. RIFKIND:  Professor Cain is back in 

 3    California, where I understand he is busy with a number of 

 4    matters, but maybe I can simplify this thing and move it 

 5    down the road a little piece.

 6               I believe, and I'm only speculating but I think 

 7    it's an informed speculation, that Professor Cain will be 

 8    unable to answer the question that is proposed, that, in 

 9    fact, he does not have an opinion on the subject because the 

10    business of redistricting is, indeed, up to the legislature 

11    at least in the first instance.  It doesn't lend itself, I 

12    think, to an expert opinion as to what the legislature will, 

13    in fact, do.

14               That is my speculation, your Honor, but I believe 

15    it is an informed speculation.

16               Maybe we should find the answer to the question 

17    and then fuss about whether the government wants to depose 

18    him and all the other things.  

19               THE COURT:  Let's do that.  

20               MR. RIFKIND:  If the answer is as I suspect --  

21               THE COURT:  They haven't heard for days.  

22               MR. RIFKIND:  --  I think Mr. Segen is going to 

23    have to find another affiant or whatever your Honor will 

24    allow.  

25               THE COURT:  You will gets the answer by calling 
                                                              1641

 1    Cain?  

 2               MR. RIFKIND:  We will have someone call Dr. Cain.  

 3               MR. SEGEN:  Your Honor, I would ask to have the 

 4    opportunity to speak to Professor Cain.  

 5               MR. RIFKIND:  Be my guest.  

 6               THE COURT:  Go ahead.

 7               Put it on the back burner again.

 8               All right, let's resume.  

 9    BY MR. SITCOV:

10         Q.    You are aware, aren't you, Dr. Fienberg, that 

11    after the imputation models were used to assign match and 

12    correct enumeration probabilities to the unresolved in the 

13    PES, a subsample of that group was reinterview in the 

14    evaluation followup?

15         A.    That's correct.

16         Q.    When I took your deposition last month, you 

17    didn't remember if any such followup had occurred, did you?  

18               MS. BARRY:  Do you have a page cite, Mr. Sitcov?  

19               MR. SITCOV:  I guess I could.

20               I think it's on page 185.  

21               MS. BARRY:  Page 185.  Thanks.

22               (Pause)

23         Q.    Do you recall I asked you, "Are you aware that 

24    after the imputation models were used to assign match and 

25    correct enumeration probabilities to the unresolved people 
                                                              1642

 1    in the PES, a subsample of them were reinterview in the 

 2    evaluation followup?

 3               "A.    I don't remember any details of that." 

 4               Do you recall testifying to that effect?

 5         A.    That's what I said at the time, yes.  

 6               MR. SITCOV:  Your Honor, may I approach the 

 7    witness for a moment?  

 8               THE COURT:  Yes.           

 9               MR. SITCOV:  I am going to hand the witness a 

10    copy of the P3 report.

11               I opened the volume for your Honor and it's in 

12    front of the clerk.  

13               (Pause)

14         Q.    The P3 report shows the results of the Bureau's 

15    evaluation followup study of the imputation model used for 

16    the P-sample unresolved, doesn't it?

17         A.    Yes, it does.

18         Q.    Would you turn to page 10, please.

19               The purpose of this study was to determine how 

20    well the production imputation procedure worked, isn't that 

21    right?

22         A.    It was one component of that, yes.

23         Q.    And the study was done by the Census Bureau 

24    actually sending employees out and interviewing the 

25    unresolved people in the PES, isn't that right?
                                                              1643

 1         A.    Yes, it was, in the way I described roughly 

 2    before.

 3         Q.    You see table 3.1 on page 10?

 4         A.    Yes, I do.

 5         Q.    And that has a number of columns under the 

 6    heading "inputed match probability," is that right?

 7         A.    Yes.

 8         Q.    And the heading above the rows says, "Evaluation 

 9    match status," right?

10         A.    Yes, it does.

11         Q.    The columns represent the match probability for 

12    certain categories of resolved cases, isn't that right?

13         A.    Yes, they do.

14         Q.    And the rows identified the different results of 

15    evaluation followup reinterview, isn't that right?

16         A.    Yes, that's correct.

17         Q.    According to table 3.1, as the percent of inputed 

18    match probability increases, the number of unresolved cases 

19    increases, doesn't it?

20         A.    Yes, it does.

21         Q.    And, in fact, the table shows that for the 

22    unresolved P-sample cases where an inputed match probability 

23    of 75 to 100 percent after evaluation followup interviewed 

24    them, more than three-quarters of those cases were still 

25    unresolved, isn't that right?
                                                              1644

 1         A.    That's correct.

 2         Q.    And if the model was working correctly, the 

 3    number of unresolved cases in table 3.1 should have trended 

 4    down as the rate of inputed match probability increased, 

 5    isn't that right?

 6         A.    Not necessarily.

 7               Actually, the way to look at this table is to 

 8    look at the first two lines, the match and nonmatch, and you 

 9    see that the proportion of those cases actually matched goes 

10    up.

11               The difficulty in understanding unresolved is 

12    more complicated.  If you recall, the followup occurred 

13    almost a year after the original census, and if this was a 

14    case where the inputed match probability was high and they 

15    still weren't able to resolve it when they first went up to 

16    do the resolution, then, in fact, it may well be that for a 

17    variety of reasons, because of the passage of time and 

18    everything else, that it would become continuing to be 

19    difficult to find that information, and so looking only at 

20    the unresolved --  first of all, I don't know how I would 

21    have projected it would trend and it is what it is.  The 

22    proportions of matches in that table to nonmatches goes up 

23    and it goes up in a fairly consistent pattern.

24         Q.    So you say it's the ratio of matches to 

25    nonmatches that is important, right?
                                                              1645

 1         A.    Yes.

 2         Q.    So you are assuming, aren't you, that the 

 3    unresolved in the PES are like the resolved in evaluation 

 4    followup rather than the unresolved in evaluation followup?

 5         A.    What I am noting is that we have a great 

 6    difficulty getting the information and, therefore, an even 

 7    greater difficulty in understanding whether or not we could 

 8    expect unresolved cases to go down with increasing inputed 

 9    matches, and what I think you have to look at those that we 

10    were able to match.

11               If you want to generalize to some population, 

12    then we can talk about that activity.

13         Q.    If you are saying that it's the ratio of the 

14    match to unmatch that counts in that fourth column, that 

15    requires you to assume, doesn't it, that the unresolved in 

16    the PES are like the resolved in evaluation followup?

17         A.    It says we don't have information about those.  

18    At best we can assume that they are like those who are 

19    resolved in the followup, and those numbers are dramatically 

20    going in the direction that one would anticipate if 

21    imputation was, in fact, doing the job it set out to do.

22               You have a ratio of about 75 percent of matches 

23    to nonmatches in the probability 75 percent and above 

24    category, so it shows something roughly what one would hope 

25    to see.
                                                              1646

 1         Q.    But, Dr. Fienberg, you testified before the break 

 2    that you didn't know if the unresolved in the PES were more 

 3    like the resolved or the unresolved in the evaluation 

 4    followup, didn't you?

 5         A.    I'm sorry, the question went by too fast.  Could 

 6    you repeat it with care?  

 7               And I'm not sure whether that question was asked 

 8    in this context or you were asking me to recall a question 

 9    that you asked me in my deposition, and I'm now confused 

10    about that.

11         Q.    Didn't you testify this morning before the break 

12    that you didn't know if the unresolved in the PES were more 

13    like the resolved in evaluation followup or the unresolved 

14    in the evaluation followup?

15         A.    I don't recall my answer to that question from 

16    before the break.

17         Q.    It would be fair to say, wouldn't it, that you 

18    believed that the unadjusted 1990 decennial census is badly 

19    flawed, is that right?

20         A.    Yes.

21         Q.    But the fact that the enumeration counts are 

22    flawed isn't reason enough to adjust, is it?

23         A.    Not in and of itself, no.

24         Q.    And, in fact, you believe that the unadjusted 

25    1980 census counts are badly flawed, don't you?
                                                              1647

 1         A.    Yes, I do.

 2         Q.    But as of the date of your deposition, you hadn't 

 3    reached a conclusion even in retrospect, had you, that if 

 4    you had been in charge of the 1980 census, you would have 

 5    recommended an adjustment?

 6         A.    I think that the circumstances in 1980 were quite 

 7    different than those in 1990.  I think that that's what I 

 8    said at my deposition, and in that sense I think I would 

 9    still agree with that statement, that is, looking at the 

10    contents of the 1980 information in that setting where there 

11    were extremely high missing data rates, extremely high 

12    matching error rates, uncertainty about how to carry through 

13    the details of various kinds of procedures that it wasn't 

14    clear to me given the information what I would have 

15    concluded if I had been the director in 1980.

16         Q.    So you still don't know what you would have done?

17         A.    What I'm --  this is a convoluted question.

18               The question is, I think, as I understand --  let 

19    me see if I understand it.

20               You are asking me what I now think I would have 

21    done if I had been director in 1980 and had the information 

22    that the director had about adjustment, and I'm saying that 

23    if I had been the director in 1980 and in the context of the 

24    1980 census and the information that the director had at the 

25    time, as I look back on it now, I'm not sure whether I would 
                                                              1648

 1    have made a decision in favor of adjustment.  That was then 

 2    and this is now.

 3         Q.    The adjustment that plaintiffs propose for the 

 4    1990 decennial census relies on verified assumptions, 

 5    doesn't it?  

 6               MS. BARRY:  Objection.  That's not what the 

 7    plaintiffs' propose, it's what the Census Bureau proposes.  

 8               THE COURT:  Overruled.

 9               You may answer.

10         A.    I'm sorry, could I have the question again?

11         Q.    The adjustment that the plaintiffs propose for 

12    the 1990 decennial census relies on unverified assumptions, 

13    doesn't it?

14         A.    I've said that no assumptions in the real world 

15    are strictly correct and that the way to look at them is to 

16    examine their impact.

17               I don't believe that either the Census Bureau or 

18    those who have examined the Census Bureau's careful studies 

19    of the impact would conclude that departures from those 

20    assumptions understood in the context of how you take a 

21    census and how you take a PES would prevent one from 

22    concluding that the corrected census counts are clearly 

23    superior to the uncorrected counts.  That's what I said.

24         Q.    My question is, does the adjustment that the 

25    plaintiffs have proposed rely on unverified assumptions?
                                                              1649

 1         A.    And I've said yes in that context.

 2         Q.    And it is also your opinion that one doesn't 

 3    verify assumptions in statistical models from beginning to 

 4    end, isn't that right?

 5         A.    Yes, that's correct.

 6         Q.    Under that fat document by your right hand there 

 7    should be something marked as Defendant's Exhibit 901.

 8               Do you see it?

 9         A.    Yes.

10               This 901 is what we were on before?

11         Q.    I'm sorry, 902.  I made a mistake.

12         A.    Yes, I have 902.

13         Q.    That is an article you wrote, isn't it?

14         A.    Yes.  It's, I believe, an unedited copy of the 

15    galley from it.

16         Q.    And that is an article to be published this year, 

17    isn't it?

18         A.    That's correct.

19         Q.    And that discusses the 1980 decennial census and 

20    the 1990 decennial census, doesn't it?

21         A.    The article is focused primarily on 1980, but 

22    there are some comments about 1990.  I attempt, actually, to 

23    show the distinction between the two settings.  

24               MR. SITCOV:  Your Honor, the defendants offer the 

25    introduction into evidence of Defendant's Exhibit 902.  
                                                              1650

 1               MS. BARRY:  We have no objection.  

 2               THE COURT:  902 is admitted.

 3               (Defendant's Exhibit 902 marked for 

 4    identification was received in evidence.) 

 5    BY MR. SITCOV:

 6         Q.    Could you turn to page 4.  

 7               I would like you to look at the second full 

 8    paragraph, the third sentence that begins, "Of course," 

 9    toward the right-hand margin, and there you say, "Of course, 

10    the assumptions are not satisfied." 

11               Do you see that?

12         A.    Page 4, and what paragraph?

13         Q.    The second full paragraph, the third sentence, I 

14    believe.  It begins, "Of course."?

15         A.    Yes.

16         Q.    You said, "Of course, the assumptions are not 

17    satisfied." 

18               You were referring there to the 1980 adjustment, 

19    isn't that right?

20         A.    That's correct.

21         Q.    And it also applies to the 1990 proposed 

22    adjustment, doesn't it?

23         A.    No, it does not.

24         Q.    You recall that on page 123 of your deposition 

25    transcript I asked you the following question, series of 
                                                              1651

 1    questions.

 2               "Q.    Those are the assumptions for the 1980 

 3    adjustment, is that right?

 4               "A.    That's correct.

 5               "Q.    Okay.  That is true for 1990 as well, 

 6    isn't it?

 7               "A.    That is correct.  Assumptions and models 

 8    are never satisfied." 

 9               Do you recall that you testified at your 

10    deposition that the assumptions in the 1990 proposed 

11    adjustment weren't satisfied?

12         A.    I need to read.

13               (Pause)

14               I believe I had just been asked a series of 

15    questions about the assumptions in 1990 and I was indicating 

16    the difference between them and 1980, and I believe --  my 

17    interpretation of the question you just asked me was in that 

18    context, and I may have misheard the question.

19               Yes, there are assumptions in 1990, and as I said 

20    before, no, they have not all been directly verified, but I 

21    do think that 1990 is quite different than 1980, as I've 

22    said before.

23         Q.    You agree --

24         A.    There are different assumptions and the evidence 

25    that was looked at in 1990 is marketly different and much 
                                                              1652

 1    improved over the evidence that was looked at in connection 

 2    with the 1980 census.

 3         Q.    You agree, don't you, that the assumptions in the 

 4    1990 case often demand external data and other evidence to 

 5    buttress them, isn't that right?

 6         A.    That's correct.

 7         Q.    And you agree, don't you, that there is an 

 8    absence of evidence fully supportive of those assumptions, 

 9    isn't that right?

10         A.    Yes, because I believe that the way in which one 

11    should look at assumptions is not whether or not they are 

12    correct, since we know that they are not; it is the degree 

13    to which departures from their appropriateness actually 

14    influence the numbers that we wish to look at.

15         Q.    The last full paragraph on page 4 of Defendant's 

16    Exhibit 902 discusses adjustment of the 1990 decennial 

17    census, doesn't it?

18         A.    Yes.

19         Q.    In that paragraph you state that the key 

20    statistical methodologists at the Census Bureau and the 

21    Census Bureau directly recommended proceeding with an 

22    adjusted census, isn't that right?

23         A.    Yes.

24         Q.    And then you continue by saying that those people 

25    were overruled by the Secretary of Commerce, isn't that 
                                                              1653

 1    right?

 2         A.    I'm sorry, I was in the --  you are on paragraph 

 3      yes.

 4    

 5               (Continued on the next page) 

 6    

 7    

 8    

 9    

10    

11    

12    

13    

14    

15    

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1654

 1         Q.    There was not unanimity among the statisticians 

 2    at the census bureau on the issue of adjusting the 1990 

 3    decennial census, was there?

 4         A.    The majority of the Undercount Steering Committee 
                                                                   

 5    recommended that the bureau proceed with correction.

 6         Q.    That wasn't a unanimous recommendation, was it?

 7         A.    No, it was not.

 8         Q.    You know that both of the members of the 

 9    Undercount Steering Committee who voted against adjustment 

10    are statisticians, isn't that right?

11         A.    Yes, they are.

12         Q.    You also know that not all of the members of the 

13    Undercount Steering Committee who voted in favor of 

14    adjustment were statisticians, isn't that right?

15         A.    We would have to go through the list for me to 

16    respond explicitly.

17         Q.    You read the recommendation that the director of 

18    the census bureau made to the Secretary on the question of 

19    adjusting the 1990 decennial census, haven't you?

20         A.    Yes, I have.

21         Q.    I take it you agree with her recommendation?

22         A.    Yes, I agree that the corrected census counts are 

23    superior to the raw census enumeration and that the bureau 

24    should proceed with a correction.

25         Q.    I ask you to turn to Plaintiff's Exhibit 55, 
                                                              1655

 1    which is one of the exhibits that the plaintiffs asked you 

 2    about yesterday.  I see you don't have it.  It is sitting 

 3    here on plaintiffs' desk.  

 4               MR. SITCOV:  I will just show it to the witness, 

 5    if you don't mind, your Honor.  

 6               THE COURT:  Sure.

 7         Q.    Dr. Fienberg, I am directing your attention now 

 8    to page 3 of the director's recommendation which appears as 

 9    Plaintiffs' Exhibit 55.  It has a Bates stamp number from 

10    the so-called so-called administrative record of 1118.

11         A.    Yes.

12         Q.    I would like you to take a look at the paragraph 

13    that begins in the last two lines.  It says, "Adjustment is 

14    an issue about which reasonable men and the best 

15    statisticians and demographers can disagree."  Do you see 

16    that?

17         A.    Yes.

18         Q.    Do you agree with that?

19         A.    I have a hard time understanding the sentence.  

20    What I can say is that there are statisticians and they have 

21    disagreed.  How one characterizes people, men and women, as 

22    reasonable is not something that I would care to speculate 

23    about in this context.

24         Q.    So you have no opinion about whether reasonable 

25    statisticians can differ about the issue of adjusting the 
                                                              1656

 1    1990 decennial census, is that it?

 2         A.    What I believe is that if one looks at the 

 3    evidence and carefully weighs it and looks at evidence that 

 4    those who disagree put forward, I do not find those analyses 

 5    and that evidence reasonable in the context of judging the 

 6    quality of the post-enumeration survey and therefore whether 

 7    or not it should be used to correct a badly-flawed census.  

 8    That is what I mean.

 9         Q.    Using that definition of "reasonable," do you 

10    agree with the Secretary that the issue of adjustment is one 

11    about which reasonable statisticians can disagree?

12         A.    I have just said that I think that if someone 

13    looked at the evidence that was before the Secretary and 

14    looked at the reasons that were set forth to ignore the 

15    recommendation of the census bureau, that they were not 

16    valid reasons and therefore it was not reasonable for 

17    someone doing that analysis to reject that judgment.  

18               THE COURT:  What you are saying is that Dr. 

19    Bryant is not reasonable in making that statement?  You 

20    can't hide from it.  That is what you are saying?  

21               THE WITNESS:  I am saying that people do 

22    disagree, but that I think --

23               THE COURT:  But those who disagree here are 

24    unreasonable?  Say it.  

25               THE WITNESS:  Given the evidence that is in the 
                                                              1657

 1    record, yes.  

 2               THE COURT:  Let's not pussyfoot around.  

 3               THE WITNESS:  I'm sorry.  I didn't mean to 

 4    pussyfoot around, your Honor.

 5         Q.    You agree, do you not, that loss functions should 

 6    be thought of as organizational tools?  Isn't that right?

 7         A.    I may have used those phrases.  Is that my 

 8    phrase?

 9         Q.    I am just asking you if you agree with that.

10         A.    Loss functions are ways to summarize data and 

11    evidence.  In that sense they are organizational tools.

12         Q.    You agree, do you not, that other tools besides 

13    loss function analysis should be used in analyzing the data 

14    on which will be based the decision whether to adjust the 

15    1990 decennial census?  Isn't that right?

16         A.    That's correct.  I have indicated that, in fact, 

17    in reaching my assessment of the evidence, I looked at other 

18    numbers such as the information in the summary of P-16 and 

19    the information that it brought to bear on the differential 

20    undercount.

21         Q.    Have you finished your answer, sir?

22         A.    Yes.

23         Q.    You would agree, wouldn't you, that the 

24    production dual-system estimator makes the assumption that 

25    within each poststratum, capture in the census and recapture 
                                                              1658

 1    in the PES are independent?

 2         A.    Yes.

 3         Q.    And there is a further assumption that within 

 4    each poststratum the probability of capture in the census is 

 5    the same for all persons, isn't that right?

 6         A.    I'm sorry.  Could I have the first part?  I 

 7    thought you were doing the second part as the first.  Let's 

 8    do the first part again, because I think I disagree.

 9         Q.    You agree that there is a further assumption that 

10    within each poststratum, the probability of capture in the 

11    census is the same for all persons?

12         A.    No, the first part, not the second.  That's the 

13    second part.  The one before.

14         Q.    Do you agree that the production dual-system 

15    estimator makes the assumption that within each poststratum, 

16    capture in the census and recapture in the PES are 

17    independent?

18         A.    Oh, independence.  Yes.

19         Q.    There is a further assumption that within each 

20    poststratum the probability of capture in the census is the 

21    same for all persons, isn't that right?

22         A.    That's correct.

23         Q.    You also agree, don't you, that the estimation of 

24    undercoverage using the dual-system estimator is quite 

25    sensitive to these assumptions?
                                                              1659

 1         A.    I believe that the estimate when actually used in 

 2    practice understates the undercount as a consequence of 

 3    departures from these assumptions.

 4         Q.    My question to you is, do you agree that the 

 5    estimation of undercoverage using the dual-system estimator 

 6    is quite sensitive to these assumptions?

 7         A.    I wouldn't have used those words in describing 

 8    the dual systems estimator in this particular context.

 9         Q.    You testified yesterday, didn't you, that you 

10    were one of the members officio or ex officio of the Panel 

11    on Decennial Census Methodology?  Didn't you?

12         A.    Ex officio, yes.

13         Q.    Ex officio.

14         A.    And then officio after 1987.

15         Q.    I believe that Ms. Barry asked you yesterday 

16    about Plaintiffs' Exhibit 2, which is the report that that 

17    panel issued in 1985 called "The Bicentennial Census: New 

18    Directions for Methodology in 1990."  Do you recall giving 

19    testimony about that?

20         A.    Yes.

21         Q.    In there on page 162 --

22         A.    Could I see the report so that I could follow the 

23    context in which it is being -- 

24               MR. SITCOV:  Your Honor, could I ask the Court to 

25    perhaps allow plaintiffs' counsel to assist me in getting a 
                                                              1660

 1    copy for the witness?  They had numerous copies here 

 2    yesterday.  I wasn't expecting they would disappear 

 3    overnight.

 4         Q.    Would you turn to page 162, please.  Do you see 

 5    the section "Dual-System Estimation Independence and Equal 

 6    Capture Probabilities"?  That paragraph begins, the first 

 7    two sentences are, "The problems with the assumption of 

 8    independence for the inclusion of people in the census and 

 9    the evaluation study list and the assumption of equal 

10    capture probabilities are related to undercoverage problems 

11    discussed above.  The estimation of undercoverage when using 

12    dual-system estimation is believed to be quite sensitive to 

13    these assumptions."  

14               My question is, do you agree that the estimation 

15    of undercoverage using the dual-system estimator is quite 

16    sensitive to these assumptions?

17         A.    I must say I don't recall exactly what I had in 

18    mind when I was part of the group that wrote that.  What I 

19    do know is that while there is some sensitivity, all of the 

20    evidence suggests, both prior to the 1980 census and in the 

21    context of the 1980 census, that the nature of that 

22    sensitivity is such as to understate the undercount that 

23    would be the result of the use of this method.  

24               So it is sensitive in the sense that it goes in 

25    one direction, that is, departures -- the information we 
                                                              1661

 1    have is that departures go in one direction.  Therefore, 

 2    yes, in that sense it is still sensitive.

 3         Q.    You agree, don't you, that either the 

 4    independence assumption or the assumption of equal capture 

 5    probabilities or both request are very likely seriously in 

 6    error?  Don't you?

 7         A.    I think that they are not correct.

 8         Q.    No.  My question is, do you think that either/or 

 9    both are very likely seriously in error?

10         A.    I think that there is residual heterogeneity and 

11    I do not -- well, let's see what I believe about 

12    independence.  I actually believe today that if you take 

13    these two together and study the issue of dependence and 

14    heterogeneity simultaneously, that, in fact, it is 

15    impossible to separate the two so that you can't know if it 

16    is one assumption or the other that you are getting the 

17    departures from.  

18               But the answer is, either way, that sensitivity 

19    leads to an understatement of the undercount.

20         Q.    Would you turn to page 140, please, in 

21    Plaintiffs' Exhibit 2.  I would like you to look at the last 

22    full paragraph.  It says, "It has long been thought that 

23    either the independence assumption or the assumption of 

24    equal capture probabilities, or both, are very likely 

25    seriously in error."  
                                                              1662

 1               Do you agree with the statement in the book that 

 2    you were involved in the preparation of --

 3         A.    I'm sorry.  Where are we here?

 4         Q.    On page 140.

 5         A.    I'm sorry.  I was 141.  The last paragraph, got 

 6    you.

 7         Q.    It says, "It has long been thought that either 

 8    the independence assumption or the assumption of equal 

 9    capture probabilities, or both, are very likely seriously in 

10    error."  Is that statement incorrect?

11         A.    It depends on what one thinks of as serious.  It 

12    goes on to label that as correlation bias.  What I am saying 

13    is that the evidence we have, both then and much more so 

14    now, gives strong support to the conclusion that it is a 

15    positive bias.

16         Q.    My question is, are those two assumptions 

17    seriously in error?

18         A.    In the sense that they imply a positive 

19    correlation bias, net correlation bias, yes.

20         Q.    They are seriously in error?

21         A.    In the sense that they imply a positive 

22    correlation bias, yes.

23         Q.    You agree, don't you, that the production dual- 

24    system estimator for states and places is based on 

25    accumulating synthetic estimates for poststrata?  Isn't that 
                                                              1663

 1    right?

 2         A.    Yes.

 3         Q.    Could you turn to page 318 of Plaintiffs' Exhibit 

 4    2.  You agree, don't you, that Table 7.2 on page 318 

 5    discusses a probable undercount through the accumulation of 

 6    synthetic estimates, isn't that right?

 7         A.    I'm afraid there may be a confusion of terms 

 8    here.  It is my recollection, and I would have to go back 

 9    and read other parts of the volume, that as we wrote this in 

10    1984-85, the word "synthetic" was used to describe a method 

11    that would be applied by taking national totals and 

12    allocating them all the way down to the kinds of small 

13    levels illustrated here.  Another term has now been 

14    substituted for that one.  The term "synthetic" is, in fact, 

15    now being used currently by others in connection with 

16    discussions in 1990 to describe the dual-systems estimate.  

17               The two as we understood them in 1984-85 when 

18    this material was being written is different, and therefore 

19    I'm not sure that this point relates to the dual-systems 

20    estimate per se.  What it does say is it relates to the fact 

21    that its near optimality is not improved.  I think what is 

22    at issue here is not the near optimality.

23         Q.    Doesn't it say its near optimality is not 

24    necessarily preserved?

25         A.    That is correct.  That says it isn't the best way 
                                                              1664

 1    to do it.  That is what "optimality" means.

 2         Q.    It is your testimony, isn't it, that synthetic 

 3    estimation means something different today than it meant 

 4    seven years ago?

 5         A.    I believe that the term as it was used in 1984-85 

 6    did not refer to the dual-systems estimate, and that this 

 7    example may or may not be applicable.  I would have to read 

 8    other material in order to answer the question about whether 

 9    this was a set of paragraphs applicable to the dual-systems 

10    estimate.

11         Q.    So you can't tell as you are sitting here whether 

12    this book that you coauthored applies to dual-systems 

13    estimates?

14         A.    I can't take a page out of a volume that I 

15    contributed to seven or eight years ago -- and I have not 

16    read this page, as I recall, since -- and interpret the 

17    details of that without the context, is what I am saying.

18         Q.    You know who David Freedman is, don't you?

19         A.    Yes, I do.

20         Q.    Know who William Navidi is, don't you?

21         A.    Yes, I do.

22         Q.    You agree, don't you, that their work questions 

23    the validity of assumptions underlying the smoothing models 

24    and observes some lack of robustness with respect to 

25    departures from the assumptions as well as the choice of 
                                                              1665

 1    variables, isn't that right?

 2         A.    Excuse me.  That is a phrase that is mine or 

 3    theirs?

 4         Q.    I am just asking you if you agree with it.

 5         A.    Then you will have to take it a little more 

 6    slowly.  Could you read it again?

 7         Q.    Do you agree that Freedman and Navidi publication 

 8    question the validity of the assumptions underlying 

 9    smoothing models and observes some lack of robustness with 

10    respect to departures from these assumptions as well as the 

11    choice of variables? 

12         A.    Taken out of the context of exactly what studies 

13    and what context, it is hard to know whether I fully agree 

14    or not.  But I might say that about what they did in 

15    connection with the 1980 data.

16         Q.    Do you agree that their work emphasizes the need 

17    to validate the assumptions underlying any models made use 

18    of in adjustment?

19         A.    They certainly take that position, yes.

20         Q.    Would you turn to page 303 of Plaintiffs' Exhibit 

21    2.  I am going to direct your attention to the carryover 

22    paragraph on the top of 303, and to the last sentence there.  

23    Please feel free to read as much of the preceding material 

24    as you want.  

25               That sentence says, "Their," referring to 
                                                              1666

 1    Freedman and Navidi say, "As you can see from the preceding 

 2    sentence, their work emphasizes the need to validate the 

 3    assumptions underlying any models made use of in 

 4    adjustment."  Do you see that?

 5         A.    Yes.  That is what it says and that is what I 

 6    think they say.

 7         Q.    Do you agree with that statement?

 8         A.    No, I don't.  Do I agree with the fact that they 

 9    said it?  Yes, I agree with the fact that they said it.

10         Q.    No.  Do you agree with that statement as it 

11    appears in this report that was done by the panel that you 

12    were a member of ex officio, that the work of Freedman and 

13    Navidi emphasizes the need to validate the assumptions 

14    underlying any models made use of in adjustment.  

15               THE COURT:  The question is whether Navidi and 

16    Freedman emphasized that, not whether the underlying 

17    statement is true.  

18               MR. SITCOV:  No.  The question is -- 

19               THE COURT:  It is a compound question.  That is 

20    the problem.

21         Q.    Do you see that the sentence there says, "Their 

22    work," meaning Freedman and Navidi, "emphasizes the need" -- 

23               THE COURT:  Right.  Your question is do you agree 

24    with that?  

25               MR. SITCOV:  Yes.  
                                                              1667

 1               THE COURT:  Are you asking him whether he agrees 

 2    their work emphasizes this or are you asking him whether he 

 3    agrees that there is a need to validate the assumptions 

 4    underlying any models?  

 5               MR. SITCOV:  I am asking him does he agree that 

 6    their work shows that there is a need.  

 7               THE COURT:  Emphasizes?  

 8               MR. SITCOV:  Emphasizes that need.  

 9               THE COURT:  Is that --

10         A.    I agree with the sentence as it is right there.  

11    Yes, their work emphasizes.  

12               THE COURT:  With you their work is wrong?  

13               THE WITNESS:  I believe that it isn't 

14    appropriate.

15         Q.    So you don't need to validate the assumptions in 

16    the adjustment process, is that right?

17         A.    As I have indicated in response to, I guess, two 

18    or three versions of this already this morning, one doesn't 

19    go about validating assumptions because you know they are 

20    not correct.  In fact, one looks at reasonable departures 

21    that are reasonable in the sense of being rooted in the 

22    knowledge of the structure of the PES, of the taking of a 

23    census, and looks at the impact that such departures have on 

24    the actual outcome in order to decide if, in fact, those 

25    departures are ones that one should find troublesome.
                                                              1668

 1         Q.    Do you agree that it is almost axiomatic that the 

 2    greatest difficulties in measuring the undercount are 

 3    encountered in the same areas where the census itself 

 4    encounters great difficulties?

 5         A.    I wouldn't use the word "axiomatic" there, but 

 6    that is because of my training as a mathematician.  An axiom 

 7    has as very explicit meaning in mathematics.  

 8               On the other hand, places that are hard to count 

 9    are hard to count.  If that is what you are trying to 

10    suggest, it is hard to disagree with that in that sense.

11         Q.    Would you agree that the greatest difficulties in 

12    measuring undercount are encountered in the same areas where 

13    the census itself encounters great difficulties?

14         A.    No.  What I am saying is it is harder to count 

15    where it is harder to count.  That is, it is harder to get 

16    information on people where, in fact, the undercount is the 

17    greatest, and the question is what is the quality of that 

18    information and its impact on the estimate that you get at 

19    that stage.  

20               MR. SITCOV:  Your Honor, may I approach the 

21    witness?  

22               THE COURT:  Yes.  

23               MS. BARRY:  Do you have a number on that?  

24               MR. SITCOV:  Yes.  It is defendants' -- I'm not 

25    sure if I do, but I will let you see it in a second.  I 
                                                              1669

 1    think you have seen it before.

 2         Q.    Dr. Fienberg, you know who Kirk Wolter is, don't 

 3    you?

 4         A.    Yes, I do.

 5         Q.    In fact, you were here in the -- 

 6               MR. ZIMROTH:  Excuse me, your Honor.  Could we 

 7    wait until we get a copy of that so we can follow it at the 

 8    same time?  

 9               THE COURT:  All right.  

10               MS. BARRY:  Has it been marked as a defendants' 

11    exhibit now?  

12               MR. SITCOV:  No.  

13               MS. BARRY:  Mr. Sitcov, I don't think we have a 

14    copy of that.  

15               MR. SITCOV:  Then you can read it over my 

16    shoulder, if you like.  I am not going to enter it into 

17    evidence.  If you have Dr. Ericksen's deposition, it was an 

18    exhibit at Dr. Ericksen's deposition.  

19               MS. BARRY:  I don't have that we have Dr. 

20    Ericksen's deposition.  

21               MR. MILLET:  It is marked as Defendants' Exhibit 

22    67.  

23               MS. BARRY:  Thank you.  

24               MR. SITCOV:  Do you have it?  

25               MS. BARRY:  Yes.  
                                                              1670

 1    BY MR. SITCOV:

 2         Q.    Were you in the courtroom when Kirk Wolter 

 3    testified on behalf of the plaintiffs in this case?

 4         A.    Yes, I was.

 5         Q.    I am going to refer you to page 26 of Defendants' 

 6    Exhibit 67.  I am going to ask you to look at paragraph 9.  

 7    It reads, "For those of us who have actual experience with 

 8    the implementation of undercount studies" -- 

 9               MS. BARRY:  Excuse me.  We don't have a copy of 

10    page 26.  It is blank in our copy of Defendants' Exhibit 67.  

11               THE COURT:  This is the one with the alternate 

12    pages?  

13               MR. SITCOV:  Yes.  I will just read the one 

14    paragraph.  

15               THE COURT:  Go ahead.

16         Q.    "For those of us who have actual experience with 

17    the implimentation of undercount studies, it is almost 

18    axiomatic that the greatest difficulties in measuring the 

19    undercount are encountered in the same areas where the 

20    census itself encounters great difficulties."  

21               Do you agree with that statement of Dr. Wolter's?

22         A.    Yes, I do.  In fact, I noted that yesterday in my 

23    direct testimony when I made reference to the high 

24    imputation rates and the impact -- the higher imputation 

25    rates that occur in high undercount areas.  I noted that in 
                                                              1671

 1    1980 that had been a big problem, and in fact Dr. Wolter was 

 2    commenting on his experience in the 1980 census.  

 3               In 1990, those high rates that he was referring 

 4    to, in fact, led to very low errors of nonmatch, the 

 5    critical error.  We actually went through a calculation 

 6    which showed that the level of nonmatches due to imputation, 

 7    the big error, that is, the result of missing data, hard to 

 8    count errors, was in fact essentially the same 

 9    proportionately for minority and nonminority areas, and that 

10    it was an order of magnitude smaller than that about which 

11    Dr. Wolter was commenting on in the context of the 1980 

12    census in that passage in that paper.

13         Q.    Can you identify for me a sample survey in which 

14    the net impact of nonsampling error on the accuracy of the 

15    survey is below one half of one percentage point? 

16         A.    I'm sorry.  Let's get the precise words so that I 

17    can answer.

18         Q.    Can you identify for me the sampling a sampling 

19    survey in which the net impact of nonsampling error on the 

20    accuracy of the survey is below one half of a percentage 

21    point?

22         A.    The 1990 post-enumeration survey is the only one 

23    for which we have -- I'm sorry.  The impact or the 

24    percentage?  That is why I am missing.

25         Q.    Percentage.
                                                              1672

 1         A.    I'm sorry.  The precision of the words really 

 2    matters here.  Otherwise, my answer may not be at all 

 3    responsive.

 4         Q.    Can you identify a sample survey that you know of 

 5    where the net impact of nonsampling error on the accuracy of 

 6    the survey is below one half of one percentage point?

 7         A.    On the accuracy of the survey in terms of the 

 8    quantity being measured, we looked at a table, for example, 

 9    in P-16 which showed that the post-enumeration survey at the 

10    national level had a net bias which is the direct measure of 

11    impact that was about -- I seem to recall .4 associated with 

12    Table 17 of that.  That is bias.  

13               I don't know whether the statement you are 

14    referring to is a statement that tries to combine bias and 

15    variability.  That is why I am having trouble responding to 

16    it.

17         Q.    Can you identify for me a sample survey other 

18    than the 1990 PES in which the net impact of nonsampling 

19    error on the accuracy of the survey is below one half of one 

20    percentage point?

21         A.    Not immediately.  In fact, it is the reason why I 

22    believe that the 1990 PES in the context of what it was 

23    chosen to measure is of such high quality.  Usually, the 

24    sources of nonsampling error in regular ongoing large-scale 

25    sample surveys are quite substantial.
                                                              1673

 1         Q.    Do you recall testifying yesterday about your 

 2    expertise in evaluating nonsampling error on surveys?

 3         A.    Yes.

 4         Q.    When I took your deposition last month, I asked 

 5    you if you could identify any sample surveys where 

 6    nonsampling error was controlled to less than a tenth of a 

 7    percentage point.  Didn't you answer me by saying that you 

 8    could not even respond to my question unless I could find 

 9    for you the terms "survey" and "nonsampling error"?  

10               MS. BARRY:  Do you have a page for that?  

11               MR. SITCOV:  Page 6.

12         Q.    I will make it easier for you.  Do you recall 

13    that I asked you the following question?

14              "Q.    Can you identify for me any sample surveys 

15    you are aware of where nonsampling error has been controlled 

16    to less than a tenth of a percentage point?"  

17               Then your attorney, Ms. Barry, said, "Objection.  

18    What do you mean by the term 'nonsampling error'?  If you 

19    understand, you can go ahead and answer, Dr. Fienberg.  

20               "THE WITNESS:  I think the issue, I would have to 

21    go back and ask what you meant by 'survey' and 'nonsampling 

22    error' to be able to answer that."  

23               Do you recall that question and your answer?  

24               MS. BARRY:  I think, in fairness, Mr. Sitcov 

25    should read the rest of the questions and answers relating 
                                                              1674

 1    to that line of questioning to make clear Dr. Fienberg's 

 2    response.  

 3               MR. SITCOV:  Your Honor, the witness has the 

 4    transcript.  Ms. Barry is going to have an opportunity to -- 

 5               THE COURT:  He has the transcript in front of 

 6    him?  

 7               MR. SITCOV:  Yes, he does.  

 8               THE COURT:  You may continue.

 9         A.    That particular question and answer does exist in 

10    this.  The reason I had difficulty is much like the reason I 

11    guess I had difficulty when I finally got the gist of all 

12    the words put together a few moments ago.  

13               It is hard to recall the state of my mind as you 

14    asked me that question over a month ago.  But if I were to 

15    attempt to do so, I guess my reflection is that I 

16    interpreted that as the overall level of nonsampling error, 

17    which is one way that one might try to measure it.  In that 

18    sense I know of no survey.  

19               That is, for example, you might say missing data 

20    is represented in the post-enumeration survey at a 2 percent 

21    level, and therefore nonsampling error in the post- 

22    enumeration survey is at least 2 percent.  I don't think 

23    that is a meaningful way to do it.  

24               I tried to explain, as I recall, at the time that 

25    what one wants to do is look at the different components of 
                                                              1675

 1    nonsampling error, see their impact, and look at the 

 2    cumulative impact on the thing that is being measured.  I 

 3    would guess that if I was able to go through enough of the 

 4    pages here, that my statement there was consistent.  

 5               Nowhere do I know of any report that goes around 

 6    saying in my study nonsampling error was controlled to less 

 7    than a tenth of a percentage point.  The number is 

 8    phenomenally low.  The number you referred to me there was a 

 9    tenth of a percentage point.  Secondly, it was the context 

10    and the way in which it was being described.  So I gave as 

11    honest an answer as I could then and I gave as honest an 

12    answer as I could now.  I believe that it is the same 

13    answer.

14         Q.    You testified yesterday, didn't you, that you 

15    reviewed the bureau's post-stratification scheme?  Isn't 

16    that right?

17         A.    Yes, I did.

18         Q.    Do you have Plaintiffs' Exhibit 54 in front of 

19    you there.  If not, I think I can probably find it in the 

20    binders.

21         A.    This is the Undercount Steering Committee report 

22    that we are referring to?

23         Q.    Yes, it is.  Could you turn to Appendix 5 of that 

24    report.  I believe the Bates number is -- I am not sure what 

25    the Bates number is.  I think the Bates number of the report 
                                                              1676

 1    is 945.  Do you have it?

 2         A.    Yes, I to.

 3         Q.    Do you have Appendix 5 there?

 4         A.    Yes, I do.  It begins on 944, and there is a 

 5    table on 945.

 6         Q.    I would like you to take a look at evaluation 

 7    poststratum number 1.  Can you tell me what poststratum 

 8    group 200040 is?

 9         A.    You are referring to New York City?  200040?

10         Q.    Yes.  Can you tell me --

11         A.    It says New York City.

12         Q.    What is the race/origin-tenure group for that 

13    poststratum group?

14         A.    Asian and Pacific islander.

15         Q.    Do you know whether they live in central cities 

16    or not?

17         A.    There isn't a direct designation there, so I 

18    would assume that that is probably indicative of the fact 

19    that that is a poststratum that encompasses something beyond 

20    the central city, but I wouldn't be certain.

21         Q.    Doesn't place type say New York City?  Isn't that 

22    a central city?

23         A.    But as I say, the designation above talks about 

24    central cities in the New York City metropolitan statistical 

25    area.  Therefore, I assume that the designation below means 
                                                              1677

 1    something different, and I would have to go back and look at 

 2    what it is.  I don't recall at the moment.

 3         Q.    Do you happen to know how many poststrata there 

 4    are in this poststratum group?

 5         A.    No, I do not.

 6         Q.    I would like you to assume that there are 12 

 7    poststrata in this poststratum group.  There are two sexes 

 8    and six age groups.  

 9               Poststratum 20004012 would consist of Asian 

10    Pacific islander males aged 10 to 19 living in New York, 

11    isn't that right?

12         A.    I'm sorry.  I have to get the numbers carefully.  

13    2 --

14         Q.    -- 0004012.  Wouldn't that poststratum consist of 

15    Asian Pacific islander males age 10 to 19 living in New 

16    York?

17         A.    I'll have to take your word for it.  I don't have 

18    the labels that go on them here.

19         Q.    10 to 19 was one of the age groups that the 

20    bureau used, wasn't it?

21         A.    The answer is yes, and males was one of the 

22    groups that was used for sex.  There are two sexes.

23         Q.    So one of the poststrata in this poststratum 

24    group would be Asian Pacific islander males age 10 to 19 

25    living in New York, isn't that right?
                                                              1678

 1         A.    That's correct.

 2         Q.    Is this poststratum a logical grouping of people 

 3    for the purpose of adjustment?

 4         A.    You have to understand the nature of 

 5    post-stratification, what it is designed to achieve, and the 

 6    extent to which it can be done.

 7         Q.    I assume --

 8         A.    No.

 9         Q.    I thought you were finished.  Sorry.

10         A.    Without that explanation, I can't answer it, so I 

11    would like to give a brief explanation.  

12               THE COURT:  Maybe you would like to withdraw the 

13    question.  

14               MR. SITCOV:  No.  I would be happy to hear this 

15    one.  

16               THE COURT:  All right.  

17               MR. SITCOV:  He said it would be brief, your 

18    Honor.

19         A.    I'll try.  

20               The census bureau singled out a number of 

21    variables and tried to combine them in a consistent way in a 

22    cross-classification structure looking at different 

23    geographic regions.  There is a tension between increasing 

24    the number of poststrata and decreasing heterogeneity. 

25               Post-stratification is there to decrease 
                                                              1679

 1    heterogeneity.  If you keep on reducing things, everyone in 

 2    the PES becomes their own evaluation poststrata.  We can 

 3    introduce enough variables that you and I could define that 

 4    we could probably make everybody distinct.  They may not all 

 5    be in the census, but you would get down to that kind of 

 6    level.  

 7               The choices here are such to balance the need for 

 8    having a few people in each of the cells in the table that 

 9    they are going to look at, preferably more rather than less.  

10    Gaining that increased accuracy that comes with the reduced 

11    heterogeneity.  There is a trade-off.  It turns out that in 

12    this particular stratum, that trade-off is found by some 

13    kinds of combinations that you might not have chosen to do.  

14    For example, for nonblacks, non-Hispanics, non-Asians, there 

15    are more of them, and therefore you expect to get more of 

16    them in the parts of New York City.  Therefore, this is a 

17    consequence of that trade-off.  It is what it is.

18         Q.    Given all those concerns, I am just asking you 

19    if, with all those concerns in mind, you consider this to be 

20    a logical grouping of people for purposes of adjustment.

21         A.    Yes.

22         Q.    There are 1391 other poststrata, aren't there?

23         A.    Yes.

24         Q.    Is it your opinion that all are just as logical 

25    as the one we just discussed?
                                                              1680

 1         A.    I think that each one is a result of this process 

 2    that I just described for post-stratification.  The logic 

 3    behind that process, I think, is one that I have followed.  

 4    I have seen descriptions, and it seems one that I find 

 5    appropriate for the use in this context.  

 6               It weighs that trade-off.  It does some 

 7    aggregation where aggregation is needed in geographic areas.  

 8    That is exactly the kind of collapsing of areas that the 

 9    census bureau regularly does in all of its surveys and in 

10    all of its activities outside the PES.

11         Q.    Would you take a look at evaluation poststratum 

12    299950.  That is in the group of evaluation poststratum 

13    number 3 on that same page.

14         A.    Yes.

15         Q.    Would you tell me what that evaluation 

16    poststratum consists of?

17         A.    It is a grouping of five different groups, all of 

18    which are black and Hispanic in noncentral city-large 

19    central city places.  But the characteristic of those places 

20    is different and their location includes two different 

21    areas, mid-Atlantic and New England.

22         Q.    In fact, that poststratum group includes all 

23    blacks or Hispanics who live in the mid Atlantic division in 

24    a type 2 metropolitan area that is not a central city, is 

25    that right?
                                                              1681

 1         A.    Yes.

 2         Q.    It also includes those who live in the mid 

 3    Atlantic division in nonmetropolitan areas but in 

 4    incorporated places with a population of 10,000 or more, is 

 5    that right?

 6         A.    That's what it says.

 7         Q.    It also includes blacks or Hispanics who live in 

 8    mid Atlantic division in other nonmetropolitan areas, is 

 9    that right?

10         A.    Yes.

11         Q.    And blacks or Hispanics who live in the New 

12    England division in nonmetropolitan areas but in 

13    incorporated places with a population of 10,000 or more, is 

14    that right?

15         A.    Yes.

16         Q.    It also includes blacks and Hispanics who live in 

17    the New England division in other nonmetropolitan areas, 

18    isn't that right?

19         A.    Yes.

20         Q.    Again, there would be 12 poststrata in this 

21    evaluation poststratum, isn't that right?

22         A.    Yes, cross-classified by sex and age.

23         Q.    Are these poststrata logical groupings of people 

24    for the purpose of adjustment?

25         A.    Yes, in the sense that I described the process 
                                                              1682

 1    before.

 2         Q.    The census bureau statisticians created the 

 3    poststrata, didn't they?

 4         A.    Yes, they did.

 5         Q.    Their post-stratification is the only reasonable 

 6    post-stratification for the purpose of smoothing, is that 

 7    right?

 8         A.    It is the post-stratification that was designated 

 9    in advance and was the subject of considerable discussion 

10    and planning, and then it was the one that was used for 

11    carrying out evaluations, yes.

12         Q.    My question is, are there other reasonable 

13    post-stratifications?

14         A.    Could there have been a different way to plan the 

15    post-stratification in advance?

16         Q.    No.  Other reasonable --

17         A.    Of course there -- the answer is, in advance 

18    there were several ways that the groupings could have been 

19    done, that the boundaries between geographic areas could 

20    have been chosen, that the age groupings could have been 

21    chosen.  In fact, there was a lot of discussion about the 

22    need for the full spectrum of ages of the six categories 

23    that you described.  That was chosen on grounds of how it 

24    would be used in analysis.  

25               In advance, this was the most reasonable one that 
                                                              1683

 1    the census bureau had, and it designated it and said that on 

 2    its basis it would decide how to proceed.

 3         Q.    Were the other ones reasonable, the other ones 

 4    that the census bureau considered?

 5         A.    The census bureau decided that this one was 

 6    better than the other ones that it examined at the time.  

 7               THE COURT:  The question, though, is were the 

 8    other once reasonable?  

 9               THE WITNESS:  There were others that I assume 

10    were sufficiently close to this that you might have said 

11    they were reasonable in advance.  But because of the 

12    requirements to prespecify the post-stratification scheme, 

13    they had to pick one, and they picked the best one that they 

14    could come up with based on all of the evidence that they 

15    had before them at the time and the requirements that they 

16    were going to bring to bear on their use in the conduct of 

17    the post-enumeration survey and its evaluation.

18         Q.    Is it your opinion that all those other 

19    reasonable post-stratifications that had been considered by 

20    the census bureau would lead to the same state level 

21    adjustments?

22         A.    I have no idea what all of the other ones that 

23    they considered would lead to.  I would have to look at the 

24    details of the analyses and their implications as well as 

25    what it would produce for actual state estimates.
                                                              1684

 1         Q.    Do you recall yesterday you testified about the 

 2    P-12 study on heterogeneity?

 3         A.    Yes.  I did.

 4         Q.    You testified yesterday that the P-12 study 

 5    demonstrates a sufficient degree of homogeneity so that the 

 6    adjusted counts were more accurate than the unadjusted could 

 7    you please tell us, didn't you?

 8         A.    No.  I said that the evidence adduced by the 

 9    census bureau -- which included P-12 and included the 

10    knowledge that the lack of heterogeneity would, in fact, 

11    increase the undercount and the impact that that would have, 

12    and that undercount would be expected to be 

13    disproportionately distributed -- that, in fact, that 

14    allowed the census bureau to conclude that the numbers were 

15    sufficiently homogenius for purposes of proceeding with the 

16    corrected counts.

17         Q.    The P-12 study was part of that?

18         A.    The P-12 study was part that, yes.

19         Q.    Could you turn to Plaintiffs' Exhibit 33.  That 

20    is the P-12 study, isn't it?

21         A.    Yes, it is.

22         Q.    Would you turn to page 10, please.  Do you have 

23    page 10?  It should be Bates number 5855.

24         A.    Yes.

25         Q.    That is a table that shows the results of a study 
                                                              1685

 1    of state effects, is that right?

 2         A.    Yes.

 3         Q.    Do you see that there is a column for allocation 

 4    rate?

 5         A.    Yes.

 6         Q.    For the allocation rate for 95 percent of the 

 7    poststratum groups there was a significant state effect, was 

 8    there not?

 9         A.    I am trying to interpret the numbers at the 

10    moment.  Percentages, 95 percent is what it says.  This is 

11    number of poststratum groups, yes.

12         Q.    For the allocation rate for 95 percent of the 

13    poststratum groups, there was a significant state effect, 

14    wasn't there?

15         A.    Yes.

16         Q.    That is evidence of heterogeneity, isn't it?

17         A.    Yes.  But the census bureau, in carrying out 

18    those tests, also noted that given the large sample size, 

19    trivial differences would be significant, and therefore what 

20    you really have to then do is look at the magnitude of the 

21    differences in some of the poststratum groups.  That is not 

22    exhibited in this particular table.

23         Q.    Yesterday you didn't tell the Court about this 

24    evidence of heterogeneity, did you?

25         A.    These are surrogates for the actual quantities 
                                                              1686

 1    under discussion.  I want to go back to the comment I made 

 2    yesterday about face validity.  When you don't have the 

 3    numbers that you need, often you look for other touchstones.  

 4               What the census bureau did is it took a number of 

 5    other things that they thought could be related to 

 6    heterogeneity in this study and they said, let's look at the 

 7    variations in them and see what kind of differences you can 

 8    find.  This is one in which that kind of surrogate 

 9    information were looked at and where relatively small 

10    differences in those surrogates are being listed here as 

11    statistically significant.  Simply counting the number of 

12    units for which you find statistical significance doesn't 

13    convey the information about the extent to which 

14    heterogeneity or homogeneity has been achieved.

15         Q.    You didn't tell the Court yesterday about the 

16    evidence of heterogeneity that is shown in the allocation 

17    rate column on Table B on page 10, did you?

18         A.    Your Honor, I confess, I didn't tell you the 

19    numbers in that table in that column.

20         Q.    In fact, you didn't even testify about this 

21    table, did you?

22         A.    I didn't testify in detail about any of the 

23    things in this particular P study.  What I did testify was 

24    about the overall conclusions that the census bureau reached 

25    as a result of all of these tables, and that overall 
                                                              1687

 1    conclusion was in fact that these studies, including the one 

 2    that produced that table, were supportive of the fact that 

 3    the post-stratification scheme reduced heterogeneity 

 4    markedly and that the residual heterogeneity did not seem to 

 5    be a sufficiently large magnitude to prevent the use of the 

 6    dual systems approach to produce corrected counts.  That is 

 7    what I testified about.

 8         Q.    The next column is mail return rate.  Do you see 

 9    that?

10         A.    Yes, I do.

11         Q.    That table shows that for 93 percent of the 

12    poststratum groups, there was a significant state effect, 

13    wasn't there?

14         A.    That's correct.

15         Q.    That is evidence of heterogeneity, isn't it?

16         A.    Indirect, yes.

17         Q.    You didn't tell the Court about that evidence of 

18    heterogeneity yesterday, did you?

19         A.    No.  And the same caveat and explanation is 

20    applicable.

21         Q.    Now can you look at the next column, multiunit 

22    structure rate.

23         A.    Yes.

24         Q.    For 96 percent of the poststratum groups there 

25    was a significant state effect, wasn't there?
                                                              1688

 1         A.    Yes.

 2         Q.    That is evidence of heterogeneity, isn't it?

 3         A.    Yes.

 4         Q.    You didn't tell the Court about that evidence of 

 5    heterogeneity yesterday, did you?

 6         A.    As I just indicated, it is true for the entire 

 7    table and that those results don't demonstrate that there is 

 8    a large amount of residual heterogeneity.

 9         Q.    So for multiunit structure rate, male universe 

10    rate and substitution rate, there was a significant state 

11    effect, is that right?

12         A.    That's correct.

13         Q.    In each instance, that is evidence of 

14    heterogeneity, isn't it?

15         A.    Yes.  The census bureau looked at the magnitude 

16    of the differences in some of the poststratum groups, and 

17    they concluded that the magnitude was not sufficiently great 

18    to distort the results.  That is what it says on page 2.

19         Q.    Was the data analysis in P-12 done at the level 

20    of individual census blocks?

21         A.    I'm sorry?

22         Q.    Was the data analysis in P-12 done at the level 

23    of individual census blocks?

24         A.    I would have to go back to the front to recall 

25    the details.  There are some analyses in which they did 
                                                              1689

 1    applications involving block-level data in the third 

 2    approach, and looking at it in the second approach also.

 3         Q.    Could you look at the bottom of page 6 on P-12.  

 4    Do you see the equation that is displayed there?

 5         A.    Yes, I do.

 6         Q.    That is a measure of heterogeneity of rates 

 7    within a poststratum or poststratum group, isn't it?

 8         A.    What it is is an equation used to predict the 

 9    proportion in various kinds of categories like those 

10    surrogate proportions that we were just discussing.

11         Q.    My question is, is it a measure of heterogeneity 

12    of rates within a poststratum or poststratum group?

13         A.    I would characterize it in the way that I did.  I 

14    wouldn't characterize it directly in that fashion.  That 

15    requires an interpretation, and it is not what this 

16    particular equation is.  

17               It is an equation for predicting the log odds 

18    each of those characteristics.  That is, you take the 

19    proportion, you divide by 1 minus the proportion, and you 

20    take its logarithm.  That makes it linear in this equation.  

21               THE COURT:  Of course.

22         Q.    Doesn't the expression involve a variance?  

23               THE WITNESS:  That was the best explanation I 

24    could get.

25         A.    I'm sorry?
                                                              1690

 1         Q.    Does the expression involve a variance?  

 2               THE COURT:  Does the expression?

 3         Q.    Yes, in the equation involve a variance.

 4         A.    I assume that implicit is the variance 

 5    associated, that one would normal associate with this.

 6         Q.    You are looking at the one at the bottom, aren't 

 7    you, not at the one at the top?

 8         A.    I am still looking at the top.  Where would you 

 9    like me to look?

10         Q.    Then I made a mistake.  I was just asking you 

11    about the equation on the bottom or intending you to look at 

12    the equation on the bottom.

13         A.    We are going to to the bottom?

14         Q.    Yes.  My question is, the equation at the bottom 

15    is a measure of heterogeneity of rates within a poststratum 

16    or poststratum group, isn't it? 

17         A.    It is a measure of heterogeneity not of the 

18    quantity that we are interested in, which is the probability 

19    of being missed in the census or missed in the 

20    post-enumeration survey.  Rather, it is a measure of 

21    heterogeneity of all of these other surrogate variables that 

22    we were looking at.  I don't recall what is big, what is 

23    small when measured by it.

24         Q.    That measure was used to adjust the Wald 

25    statistics in the logistic regression manner, wasn't it?
                                                              1691

 1         A.    That's what it stays. 

 2         Q.    When the bureau found heterogeneity, they divided 

 3    it out of their test statistic for heterogeneity, is that 

 4    right?

 5         A.    I don't follow that question.  I'm sorry.

 6         Q.    So you don't know how to understand the question?

 7         A.    You used a bunch of words, and you said divided 

 8    out.  I don't follow what you meant by that.

 9         Q.    You said that the equation on page 6 was a 

10    measure of heterogeneity of rates and that the measure was 

11    used to adjust the Wald statistics in the logistic 

12    regression model.

13         A.    Yes.

14         Q.    So my question is, when the bureau found 

15    heterogeneity, it divided it out of the statistic for 

16    heterogeneity, didn't it?

17         A.    I don't know what it meant -- I don't know what 

18    you mean by "divided it out."

19         Q.    All right.

20         A.    I'm sorry.

21         Q.    Could you turn to page 1 of P-12.

22         A.    Of?

23         Q.    P-12.  

24               THE COURT:  Same exhibit.

25         Q.    The 1990 production smoothing model was based on 
                                                              1692

 1    a regression model, wasn't it?

 2         A.    Yes, it was.  That was used as part of the -- one 

 3    of the many components that yielded the results.

 4         Q.    Would you take a look at the fifth block of type, 

 5    the one that says, "Our third approach."  Would you read 

 6    that to yourself and just let me know what you are done.

 7         A.    Yes.

 8         Q.    The approach that is described in that paragraph 

 9    that was used to demonstrate homogeneity involved a 

10    comparison between smoothed adjustment factors and predicted 

11    adjustment factors, is that right?

12         A.    I'm sorry.  The smoothed adjustment factor is 

13    called the predicted factor in step 3.  Then they compared 

14    that -- for the blocks.  Okay.

15         Q.    What that approach was doing was demonstrating 

16    homogeneity by comparing smoothed adjustment factors with 

17    predicted adjustment factors, is that right?

18         A.    Yes.

19         Q.    The predicted adjustment factors were predicted 

20    from the regression model that was part of the smoothing 

21    model, weren't they?

22         A.    For the poststratum, yes.

23         Q.    So the bureau demonstrated that outputs from 

24    different parts of the same smoothing model were closer to 

25    each other than they were to the census, is that what it 
                                                              1693

 1    comes down to?

 2         A.    No, not at all.  

 3               THE COURT:  Shall we break for lunch?  

 4               MR. SITCOV:  Yes.  I won't have much more when we 

 5    come back.  

 6               THE COURT:  How much more will you have?  

 7               MR. SITCOV:  I would guess 10, 15 minutes at the 

 8    most.  

 9               THE COURT:  Do you want to finish it?  

10               MR. SITCOV:  If it turns out I'm a liar, I would 

11    just as soon be a liar after lunch.  

12               THE COURT:  All right.  We will resume at 2:15.  

13               (Luncheon recess) 

14    

15    

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1694

 1               A F T E R N O O N    S E S S I O N

 2                                            2:15 o'clock p.m.

 3    STEVEN ELLIOT FIENBERG,           resumed. 

 4    CROSS-EXAMINATION (Continued 

 5    BY MR. SITCOV:

 6         Q.    Dr. Feinberg, there are upwards of four million 

 7    inhabited blocks in the United States, right?

 8         A.    I don't remember the number, but if that's what 

 9    it is, I would accept that.

10         Q.    Well, you don't know how many inhabited blocks 

11    are around the country?

12         A.    I don't remember the number is what I said.

13         Q.    Does four million sound about right?

14         A.    Not unreasonable as an order of magnitude.

15         Q.    You are familiar with the current population 

16    survey, aren't you?

17         A.    Yes, I am.
                           

18         Q.    The current population survey doesn't estimate 

19    unemployment rates for every single block in the country, 

20    does it?

21         A.    No, it does not.

22         Q.    You are familiar with the crime survey run by the 

23    Census Bureau, aren't you?

24         A.    Yes, I am.

25         Q.    The crime survey doesn't estimate victimization 
                                                              1695

 1    rates for every block in the United States, does it?

 2         A.    No, it does not.

 3         Q.    Does the PES estimates separate undercount rates 

 4    for each of the four million blocks in the United States, 

 5    doesn't it?

 6         A.    Yes, for the purposes of aggregating back up.

 7         Q.    And that estimate is based on data from 12,000 

 8    blocks in the PES, isn't that right?

 9         A.    Yes.

10         Q.    And to make those estimates, the Bureau relies on 

11    the homogeneity assumption, doesn't it?

12         A.    I wouldn't say that it relies on the homogeneity 

13    assumption.  It uses it and that produces a lower bound 

14    about the extent of undercount to which, which is carried 

15    down to those blocks.

16         Q.    And that is the assumption that is tested in the 

17    P12 study, isn't it?

18         A.    The P12 study looks at the issue of homogeneity 

19    and hetrogeneity in three different ways, yes.

20         Q.    It's your opinion that virtually every component 

21    of the PES relies on assumptions, isn't that right?

22         A.    That's correct.

23         Q.    Now, suppose that a statistician testified that 

24    with respect to the census adjustment, statistical 

25    procedures don't have assumptions, that statistical 
                                                              1696

 1    procedures, instead, have circumstances where they worked 

 2    better and circumstances where they don't work as well, you 

 3    wouldn't agree with that testimony, would you?

 4         A.    That person would be characterizing the 

 5    procedures in a different kind of way than I've chosen to 

 6    characterize them.

 7         Q.    Well, would you agree if someone said that you 

 8    said there weren't assumptions in the PES?

 9         A.    I would need to see the circumstances in which 

10    that statement was made in order to know how to interpret 

11    them, but as I've said, I believe that there are some 

12    assumptions that are made in the models that are used.  I 

13    characterized them as models and someone could characterize 

14    that in a different way.

15         Q.    Would you turn to Plaintiff's Exhibit 13, please.

16               That is a book that was written by Mitroff, Mmson 

17    and Barabba, isn't that right?

18         A.    That's correct.

19         Q.    And Mr. Barabba was the director of the Census 

20    Bureau, wasn't he?

21         A.    In the 1980 census, yes.

22         Q.    And you testified about some things in this book 

23    yesterday.

24               Do you recall that?

25         A.    I pointed out that, in fact, he had made the 
                                                              1697

 1    statement in this book that the census is an estimate.

 2         Q.    Could you turn to page 9 of Plaintiff's Exhibit 

 3    13.

 4               I would like you to take a look at the first full 

 5    paragraph on that page.

 6               Do you see it?  It says, "The battle of the 

 7    undercount".

 8               Do you see that, Dr. Fienberg?

 9         A.    Yes.

10         Q.    Do you agree that the battle of the undercount is 

11    the 20th Century equivalent of the 17th Century battle 

12    between rationalism and empiricism?

13         A.    I wouldn't have characterized it in quite that 

14    way.

15         Q.    Do you agree that it is a war between completing 

16    philosophical world views and not just technical methods?

17         A.    This is a description in particular about the 

18    1980 census, and I think that is a characterization of the 

19    discussions that surrounded the buildup to the decision on 

20    whether or not to adjust in 1980.  There were those, in 

21    fact, who were taking a philosophical position, for example, 

22    that there was no estimation in the census when, as I've 

23    indicated, there was.

24               So as a description of the 1980 census, I think 

25    it's --
                                                              1698

 1         Q.    Do you agree that with respect to the 1990 census 

 2    it's a war between competing philosophical world views and 

 3    not just technical methods?

 4         A.    I think in 1990 that given --  let me begin 

 5    again.

 6               No, I don't, that is, I believe in 1990 that 

 7    there is a context in which the issue of correction 

 8    adjustment can be said where it becomes something other than 

 9    simply a war between competing positions.  I think that 

10    there is clear evidence here about what to do and why to do 

11    it.

12         Q.    Do you agree that with respect to 1990 the issue 

13    of the undercount can't be settled either by pure reason or 

14    observations alone because it is the status of reason versus 

15    observation that is in dispute?

16         A.    If by that you mean that we haven't seen the true 

17    count, it would be very hard for me to dispute it.  That is, 

18    we are doing it on the basis of what we can observe, but the 

19    issue is, are we making things better, and ultimately there 

20    is only one person that can tell you that and I'm not the 

21    one.

22         Q.    Would you turn to page 35, please.

23               Would you look at the second full paragraph.

24               Do you agree that by a long and arduous series of 

25    complex and technical considerations, we have been steadily 
                                                              1699

 1    inching toward the inevitable conclusion that the Census 

 2    Bureau has demonstrated that with consummate skill it is 

 3    still -- it is impossible to achieve an estimate of the 

 4    undercount in the United States population that is not 

 5    subject to substantial challenge?

 6         A.    I believe that's a reasonable characterization of 

 7    the 1980 census about which its being written.

 8         Q.    I'm asking you if you agree with that as to the 

 9    1990 census.

10         A.    No, I don't.

11         Q.    Would you turn to page 47, please.

12               I will direct your attention to the first 

13    paragraph on that page under, "Conclusion." 

14               Do you agree that the problems of establishing 

15    the basic existence of an undercounting, finding a procedure 

16    for accounting or adjusting for it and assessing the impact 

17    of allocations are complex almost beyond believe?

18         A.    No, I don't.

19         Q.    Do you agree that the more one examines the 

20    adjustment issue, the more complex it becomes?

21         A.    No.

22         Q.    Could you turn to page 77, please.

23               I would like you to look at the first full 

24    paragraph.

25               Do you agree with the statement that the notion 
                                                              1700

 1    of the undercount is only a hypothetical construct?

 2         A.    I'm sorry, I can't --  there, it is.  I've got 

 3    it.  I've got it.

 4               (Pause)

 5               No.  

 6               MR. SITCOV:  No further questions.  

 7               THE COURT:  Thank you.

 8               Redirect.  

 9               MS. BARRY:  No redirect, your Honor.  

10               THE COURT:  Pleasure to say goodbye to you.

11               (Laughter)  

12               MR. RIFKIND:  I think subject to the reservations 

13    that were made this morning, your Honor, with respect to 

14    depositions and documents, we rest at this time.  

15               THE COURT:  What about our friend from New 

16    Jersey, is he here?  

17               MR. RIFKIND:  The last I saw --  
                                                  

18               THE COURT:  Is he resting comfortably in the 

19    park? 

20               (Laugher) 

21               MR. RIFKIND:  My understanding is that he is now 

22    at my office going through my collection of exhibits and 

23    transcripts, which I welcomed him to visit with, and in the 

24    meantime a search is on for Professor Cain.  He had not been 

25    found this morning, except I had a report that he was 
                                                              1701

 1    engaged in some hearings in California, but we will report 

 2    back when we hear from him.

 3               I have nothing further to report at this time.  

 4               THE COURT:  Subject to all of that, plaintiffs 

 5    rest.  

 6               MR. MILLET:  Your Honor, at this time I would 

 7    like to rise to move under Rule 52 (c) for a judgment as a 

 8    matter of law.

 9               I want to highlight, if I may briefly, a couple 

10    of points.

11               First of all, as a point which we have raised in 

12    our pleadings and is an issue which, quite frankly, has not 

13    been addressed by the plaintiffs in their testimony, and I'm 

14    going to, for purposes of my motion, your Honor, I am going 

15    to assume, since everyone is making assumptions here, I'm 

16    going to make my assumptions, and my model is going to 

17    assume all the transcripts offered by the plaintiffs are 

18    going to be receive in evidence and I'm going to make my 

19    argument on that basis.

20               Your Honor, we previously moved to dismiss many 

21    of the plaintiffs in this case for lack of standing.  

22    Standing, as the court no doubt knows, involves injury in f 

23    act.

24               It is incumbent at this point for the plaintiffs 

25    to have demonstrated, for each plaintiff to have 
                                                              1702

 1    demonstrated an injury in fact.

 2               Now, I will at the outset, your Honor, concede 

 3    that if a adjustment were ordered, that the states of 

 4    California and Arizona would gain one congressional seat 

 5    according to the PES as presented, the results of the PES as 

 6    presented to the Secretary.

 7               However, it is equally clear on the evidence 

 8    presented by the plaintiffs themselves, that a number of 

 9    plaintiffs will lose population share if an adjustment were 

10    to be put into effect.

11               We argued in our motion to dismiss, your Honor, 

12    as to many of these plaintiffs and at that point the court 

13    indicated, if I can paraphrase the court's ruling, there 

14    probably would be some plaintiffs that would have standing 

15    and at that preliminary stage there really wasn't a reason 

16    to try to sort it out.

17               Well, this, your Honor, we submit is the stage to 

18    sort it out.

19               Let me --  

20               THE COURT:  What is to be gained by parsing out 

21    which plaintiffs have standing and which don't?  

22               MR. MILLET:  Your Honor, I don't think it is an 

23    esoteric point, I don't.  It determines who is entitled to 

24    proceed with the remainder of this case, it determines who 

25    is bound on the merits by any judgment in this case, it 
                                                              1703

 1    determines who may appeal from any judgment on the merits in 

 2    this case.

 3               It's not a technicality, it is a serious point.

 4               Frankly, your Honor, we can't understand why at 

 5    this point some of the plaintiffs aren't on our side of the 

 6    courtroom, and I take as an example the states of New York 

 7    and New Jersey.

 8               Under the evidence presented, and I point your 

 9    Honor to the June 13 press release which breaks down the 

10    population shares by states, the states of New York and New 

11    Jersey will lose population share relative to the remainder 

12    of the country if this adjustment were put into effect.  

13    They will be injured by the relief they seek.

14               I don't know if there is a converse of injury in 

15    fact, but those parties certainly demonstrate it.

16               More to the point, your Honor, as to the 

17    remaining parties, and, again, putting California and 

18    Arizona aside, as to the remaining parties, there is no 

19    evidence that they will gain in proportional representation 

20    in the House of Representatives.  

21               There is no evidence to show that any plaintiff 

22    will gain any additional federal funding based upon the 

23    evidence that the plaintiffs have submitted to you, and on 

24    that basis all plaintiffs, except California and Arizona, 

25    really must be dismissed at this juncture.
                                                              1704

 1               The second point I want to make, your Honor, goes 

 2    to the merits of the plaintiff's claim as to whether the 

 3    Secretary's decision is arbitrary and capricious.

 4               I think what we have seen played out in many 

 5    respects is a microcosm of what occurred before the 

 6    Secretary.

 7               The plaintiffs have showed that they can bring up 

 8    some experts who disagree with the Secretary's decision.  

 9    That's really hardly surprising.  But what is most telling, 

10    your Honor, is that these experts can't even agree amongst 

11    themselves.  They are very good at discussing generalities, 

12    but they are not very good at explaining data.

13               Mr. Rifkind, in his opening argument, invoked the 

14    name of Director Bryant as the head of the expert agency 

15    whose judgment ought to be given defference by this court.

16               Well, your Honor, as we pointed out time and 

17    again, Director Bryant said to the Secretary that reasonable 

18    men and women, statisticians and demographers, can differ.  

19    That's the conclusion of the head of the expert agency to 

20    whom the plaintiffs have argued this court should defer.

21               Your Honor, I don't think we need to go any 

22    further than that.  Once it's shown that there are basis for 

23    reasonable disagreements, clearly the Secretary decision is 

24    not arbitrary and capicious.

25               Dr. Bailar, who testified a couple of days ago, 
                                                              1705

 1    went so far as to say everything, just about everything the 

 2    Census Bureau does is controversial in the field.

 3               Dr. Bailar, again, is re-emphasizing Dr. Bryant's 

 4    conclusion that the decision whether or not to adjust is one 

 5    upon which reasonable persons can differ.

 6               Your Honor, even Dr. Wolter, Dr. Kirk Wolter, 

 7    who, if I may say, was certainly not lacking in confidence 

 8    in his own judgment, said on cross-examination that there is 

 9    an element of professional judgment that must be brought to 

10    bear by experts in the field in deciding this issue.

11               On other issues, your Honor, the plaintiffs' 

12    witnesses couldn't even agree with themselves.

13               Dr. Fisher and Dr. Rolph relied on loss function 

14    analysis for their opinions.

15               Dr. Tukey reached his opinion without any 

16    reference to loss function, as well as Dr. Ericksen.  In 

17    fact, Dr. Tukey didn't use any recognizable statistical test 

18    for coming up with his decision.  

19               THE COURT:  He seemed to be beyond tests.  

20               MR. MILLET:  He was definitely beyond tests, 

21    definitely beyond tests, and I mean no disrespect to Dr. 

22    Tukey.  

23               THE COURT:  Nor do I.  He teaches the way I do.

24               (LaugheR)

25               MR. MILLET:  My point is, however, that there is 
                                                              1706

 1    no unanimity even among plaintiffs' experts as to how this 

 2    court should decide this issue.  That further underlies the 

 3    lack of unanimity in the field which was expressed before 

 4    the Secretary at the time he made his decision.

 5               Dr. Fisher said that the variance of the smooth 

 6    adjustment factors are robust to a factor of eight, while 

 7    Dr. Wolter and Dr. Tukey both said if the variances were 

 8    doubled, they want to think about their recommendation 

 9    again.

10               Dr. Feinberg says that there are assumptions 

11    involving underlying PES, while Dr. Tukey says there is no 

12    such thing as assumptions, a line of questions which Mr. 

13    Sitcov just questioned Dr. Feinberg.

14               They can't even agree among themselves as to how 

15    accuracy should be measured.  Dr. Ericksen says that the 

16    objective is to prove distributive accuracy, while Dr. 

17    Bailar in her testimony emphasized numeric accuracy, and Dr. 

18    Wolter in his testimony said that his main objective was 

19    ameliorating the differential undercount.

20               Dr. Feinberg said that all statistical procedures 

21    involve a homogeneity assumption, while Dr. Tukey agreed 

22    that the homogeneity assumption wasn't even an issue in 

23    1990.

24               Your Honor, they can't even agree amongst 

25    themselves.  And this is the group that is to persuade you 
                                                              1707

 1    that there is a unanimity of opinion that the Secretary was 

 2    irrational.  How can that be when they can't even agree 

 3    amongst themselves as to how this should be measured, what 

 4    are the relevant principles, what are the relevant 

 5    considerations? 

 6               The last point I wanted to mention, your Honor, 

 7    is plaintiffs' claim of political manipulation and racial 

 8    bias, and, frankly, at this point I think it's hardly worth 

 9    mentioning.  Certainly there has been no mention of it 

10    during the trial so far.

11               Plaintiffs would have you believe, your Honor, 

12    that two different Secretarys of Commerce at two different 

13    times along with numerous subordinates and other officials 

14    and who knows who else engaged in some grand conspiracy over 

15    the span of five years to deprive plaintiffs of their 

16    constitutional rights based on nothing else than the fact 

17    that those Secretarys were not in favor of adjustment.

18               Your Honor, at the beginning of this trial in his 

19    opening statement Mr. Rifkind stood up and said in so many 

20    words he was going to prove and his colleagues were going to 

21    prove that the defendants were racists and that they 

22    manipulated this entire process for political gain.

23               I would submit, your Honor, that that rhetoric 

24    has not been supported by credible evidence and that this 

25    aspect of their claim must also be dismissed.  It rests upon 
                                                              1708

 1    conjecture and suspicion and not evidence.

 2               The allegations of bad faith and manipulation are 

 3    the same allegations that were presented before Magsitrate 

 4    Ross, they are the same allegations that have been presented 

 5    to you in pretrial proceedings.  They is no evidence, the 

 6    evidence is no different, there has been nothing new in this 

 7    trial to support those; the results should be no different 

 8    than it was before Magsitrate Ross or before your Honor in 

 9    the pretrial proceedings.

10               I can discuss a few of the examples that they 

11    pointed to in their pretrial brief as evidence of bad faith 

12    and political motivation.

13               For example, they argued that somehow the 

14    Secretary ginned up this notion of null hypothesis testing 

15    and sprung it on the Bureau at the last minute.

16               In fact, that's not true.  There is record 

17    evidence in the Jones deposition that demonstrates that that 

18    is wrong, that the null hypothesis testing was considered 

19    from the very beginning of the guidelines.

20               They have argued that Dr. Plant's drafting of an 

21    anti-adjustment decision before July 15, but the lack of 

22    drafting of pro-adjustment decision is somehow indicative of 

23    bad faith.

24               Well, that argument was presented before Judge 

25    Ross and she dismissed it as ignoring the fact that the 
                                                              1709

 1    stipulation and order required the Secretary to articulate 

 2    the reasons behind an anti-adjustment decision by July 15, 

 3    but there was no deadline for doing anything for a 

 4    pro-adjustment decision other than releasing the tapes.  

 5               If there was to be a pro-adjustment decision, 

 6    those tapes were ready, they could have been released.  The 

 7    Secretary certainly would not have been written a decision, 

 8    but they didn't have to have a decision ready by July 15 and 

 9    Judge Ross realized that in the discovery rulings and the 

10    evidence is no different now.

11               I guess the last point I want to mention is 

12    another point that Mr. Rifkind adverted to in his opening 

13    argument, and these were these somehow nefarious contacts 

14    between Mr. Willkie and the White House, and Mr. Rifkind 

15    said that there obviously had to be bad faith going on, some 

16    hanky-panky had been going on, why else would Wendell 

17    Willkie go to the White House and indicate that they didn't 

18    want to hear from the White House? 

19               Your Honor, the only thing I want to say on that 

20    score is if that is enough for the plaintiffs to win, then 

21    the lesson that this court would be sending to me and my 

22    colleagues when we go back to advice policymakers is never 

23    to advise them to abide by the law, because when you do that 

24    it would be taken as evidence that you are breaking the law.  

25    That kind of logic is backwards.
                                                              1710

 1               There is no evidence to support the bad faith 

 2    claim on the merits, your Honor; even their own experts are 

 3    contradicting each other; and very clearly on the standing 

 4    issue the courtroom ought to get a little smaller if we go 

 5    forward.  

 6               THE COURT:  Thank you.  

 7               MR. RIFKIND:  Your Honor, I am not prepared to 

 8    recapitulate the evidence and I think that's probably the 

 9    last thing you wish to hear at this hour.

10               On the standing proposition, I would have thought 

11    it sufficient to observe that every state that is a party to 

12    this action will lose money on a basis that the Court of 

13    Appeals has already taken judicial notice of.  There are 

14    programs in the statute books that distribute money on the 

15    basis of absolute population and other factors.

16               Given that, I didn't think it was worth burdening 

17    the court with a blow by beow run-through on each of those 

18    statutes, although we identified a fair number of them the 

19    last time the standing argument came up in a brief we 

20    submitted at the time.

21               Each of these states gains in population.  It may 

22    not gain relatively, but it gains absolutely, and there are 

23    funding --  

24               THE COURT:  What does that mean?  

25               MR. RIFKIND:  Whether or not a particular state 
                                                              1711

 1    gains in its aggregate share of the nation, it gains in 

 2    absolute numbers, that is, its population goes up by, let's 

 3    say, 100,000, although its percentage of the nation might, 

 4    in fact, go down because other places are going up even 

 5    more.

 6               But as long as its absolute numbers are going up, 

 7    there are programs which we identified to you in a brief 

 8    some months ago under which they lose money, and that's, 

 9    frankly, the end of it, and since the case has to go 

10    forward, I don't really see that there is much utility in 

11    burdening the court with that at this time, although we 

12    could if you wanted us to.

13               There is also the fact that each of these states 

14    things it important that its congressional delegation be 

15    fairly allocated within the state and that the state 

16    legislatures be fairly allocated within the state, and 

17    although there is a ruling in the Sixth Circuit to the 

18    contrary, I believe that is sufficient for standing.  But, 

19    again, I don't think one has to resolve that issue finally 

20    at this hour.

21               Of course there are disagreements on minutia 

22    among the array of experts on every side.  Of course each of 

23    them brings his own point of view, his own way of 

24    characterizing statistical phenomena and so on to the table.

25               But there isn't the slightest disagreement among 
                                                              1712

 1    them that a robust sturdy mechanism was designed to assess 

 2    the undercount, it was assessed, the professional agency 

 3    charged with doing that came to a conclusion based on 

 4    substantial evidence which viewed from a great many 

 5    different perspective, both internal and external, seemed 

 6    very robust, and the grounds on which the Secretary rejected 

 7    them are --  what shall I say?  --  changing the rules after 

 8    the game is played or selectively examining pieces of 

 9    evidence as to which he was advised by the people who are 

10    giving him the evidence that it was not reliable.

11               The loss function analysis was built into this 

12    piece of machinery from the very day it began.  For the 

13    Secretary to say in his decision that no, no, that's the 

14    wrong way to do it is changing the rules of the game.

15               Having changed the rules of the game, there was 

16    evidence that even under the standard he adopted, the 

17    adjustment was overwhelmingly preferred over nonadjustment.

18               I never suggested, I never said --  maybe I 

19    suggested --  I never said and I didn't mean to say that any 

20    particular individual was engaged in a racist cabal, I said 

21    that there are well-established known shortcomings of the 

22    enumeration which have a known impact, that impact is on 

23    racial minorities, and it is incumbent upon the government 

24    when it knows that and it has an alternative to avail itself 

25    of that alternative, and I think that's quite clear.
                                                              1713

 1               To say that there is no evidence in this case of 

 2    political involvement or political considerations, although 

 3    as I said when I started I don't think I need to prove that, 

 4    I think is quite evident.  

 5               The fact is the undercount is in Democratic 

 6    neighborhoods very heavily.  It's also in minority 

 7    neighborhoods very heavily.  A series of exhibits and 

 8    studies and witnesses demonstrated that point.  I think at 

 9    one point your Honor said it was so obvious that you 

10    wondered why we were going on about it.

11               When you add to that what the depositions do add 

12    to that and that Mr. Sununu was calling the Commerce 

13    Department and urging them in a particular direction, I 

14    think one is entitled to draw appropriate inferences.

15               (Continued on the next page) 

16    

17    

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1714

 1               I do not regard that as indispensible to my case.  

 2    What I regard as indispensible to my case is that the short- 

 3    comings of the census enumeration are well known and well 

 4    documented and the mechanism for improving on that is also 

 5    well established.  

 6               Thank you.  

 7               THE COURT:  Thank you.  
                                         

 8               The government's motion under Rule 52 to dismiss 

 9    on grounds that some of the plaintiffs lack standing is 

10    denied for the reasons I have outlined in the prior opinion.  

11    The motion to dismiss on the other two grounds, i.e., that 

12    the plaintiffs have proven only that experts can disagree on 

13    this and that plaintiffs have failed to prove any political 

14    manipulation or racial bias in the decision, is reserved.  

15               MR. MILLET:  Thank you, your Honor.  

16               THE COURT:  We are ready to go.  

17               MR. MILLET:  Your Honor, our first order of 

18    business is to offer into evidence Defendants' Exhibit 1, a 

19    certified copy of the administrative record, which we have 

20    piled in its glory over here.  

21               THE COURT:  Hearing no objection -- 

22               MR. RIFKIND:  I do object.  

23               THE COURT:  Oh?  

24               MR. RIFKIND:  I guess I am not clear what it is 

25    offered for.  
                                                              1715

 1               MR. MILLET:  It is being offered as evidence of 

 2    the administrative record.  It is a self-authenticating 

 3    document.  

 4               MR. RIFKIND:  Your Honor, in that cart, if it is 

 5    the same -- if it is the same -- as was given to me in 

 6    August, although I didn't get it with the seals flying, I 

 7    take it there are some 18,000 pieces of paper, and, by my 

 8    last count, some 12,000 documents.  

 9               I put aside for this discussion the question of 

10    authenticity, which I don't wish to address.  I am not 

11    waiving it, but I put it aside.  

12               If I might pick up one of these precious volumes.  

13    There is an affidavit from Mr. Mark W. Plant on the face of 

14    it with a red seal.  He says this is the administrative 

15    record or this is 500 pages of the administrative record.  

16               That is a central issue in this lawsuit, as your 

17    Honor recognized in the decision on the shape of the trial.  

18    They are undoubtedly entitled to put in competent evidence 

19    on what is the administrative record, but Mr. Mark Plant's 

20    sealed piece of paper saying it is the administrative 

21    evidence is not competent.  It is not competent because Mr. 

22    Plant has testified that he doesn't know what was in the 

23    administrative record, had no basis for knowing what was in 

24    the administrative record, didn't assemble the 

25    administrative record himself.  
                                                              1716

 1               Indeed, defendant's counsel, while Mr. Plant was 

 2    testifying, admonished us when we asked him whether he knew 

 3    what was in the administrative record, and said, "Dr. Plant 

 4    is not going to be able to tell you because that would be 

 5    his speculation as to what the Secretary directly and 

 6    indirectly considered."  Plant deposition at 34.  

 7               I could go on at some length reading portions of 

 8    the deposition of Mr. Plant to you, but my basic submission, 

 9    to make it very simple and short, is the evidence of his own 

10    testimony is that he didn't know and that therefore the 

11    representation that this is the administrative record is not 

12    to be credited.  

13               Indeed, the evidence is that the administrative 

14    record was assembled by a variety of people, that is, what 

15    is called the administrative record, no one of whom asserts 

16    that they knew what was directly or indirectly considered by 

17    the Secretary, that it was finally edited, compiled, and 

18    assembled with final decisions being made not by Mr. Plant 

19    and not by a subordinate of Mr. Plant, but by counsel in 

20    this case.  

21               Counsel have represented to your Honor that the 

22    record was assembled with an eye toward the defense of this 

23    litigation.  Indeed, the record also shows and the 

24    depositions also show that the counsel also participated in 

25    the drafting of the decision.  So it is a very odd, very 
                                                              1717

 1    strange administrative record that they are presenting us 

 2    with.  

 3               I don't know whether there isn't some basis to do 

 4    something with this document, but it is certainly not to be 

 5    admitted for the proposition Dr. Plant asserts, even with a 

 6    red seal on it.  I urge it not be received on that basis.  

 7               THE COURT:  Mr. Millet?  

 8               MR. MILLET:  First as to whether any one person 

 9    put this all together.  Your Honor, no one person could have 

10    put all that together.  It's impossible.  

11               As to whether counsel was involved in advising 

12    the defendants as to a matter while litigation was pending, 

13    your Honor, I certainly hope so.  That is our job.  

14               As to the admissibility of this document, your 

15    Honor, it is in compliance with Rule 902.  It is a domestic 

16    document under seal, and it is self-authenticating.  We 

17    offer it as the certified copy of the administrative record.  

18               If the Court would like case authority on the 

19    point of the admissibility, I am happy to cite it to you.  

20    Otherwise, I will be quiet.  

21               THE COURT:  The objection is overruled.  The 

22    exhibit is received as Defendants' Exhibit 1.  

23               (Defendants' Exhibit 1 for identification was 

24    received in evidence) 

25               MR. SUBAR:  Your Honor, defendants call at this 
                                                              1718

 1    time Peter Bounpane to the stand.  We might need just a 

 2    moment to move a couple of easels and some charts.  I don't 

 3    know if we need as much time as it would take to take a 

 4    break.  Why don't we do this just quickly.  

 5               THE COURT:  Just do it.  Come on up, Mr. 

 6    Bounpane.  

 7    PETER BOUNPANE, 
                      

 8         called as a witness by the defendants, having

 9         been duly sworn, testified as follows:

10               THE COURT:  State your full name and spell your 

11    last name for the record.  

12               THE WITNESS:  Peter Bounpane, B O U N P A N E. 

13    DIRECT EXAMINATION 

14    BY MR. SUBAR:

15         Q.    Mr. Bounpane, by whom are you employed?

16         A.    By the U.S. census bureau.

17         Q.    Since when you have been employed by the census 

18    bureau?

19         A.    Since 1965.

20         Q.    In what capacity are you currently employed 

21    there?

22         A.    I am the assistant director for decennial 

23    censuses at the current time.

24         Q.    Where does that fall within the hierarchy of the 

25    census bureau?
                                                              1719

 1         A.    The census bureau is divided into about five key 

 2    areas.  One of them has to do with the decennial census, and 

 3    this would be the top level of management at the census 

 4    bureau for that particular topic.

 5         Q.    What is the senior executive service in the 

 6    federal government?

 7         A.    The senior executive service is a class of civil 

 8    servants within the entire government, the more senior set 

 9    of government managers that used to be the grades 15, 16, 

10    17, or super grades, that were reinstituted into what was 

11    called the senior executive service.

12         Q.    Are you a member of the SES?

13         A.    Yes, I am.

14         Q.    How long have you been in the SES?

15         A.    Since 1981.

16         Q.    What other jobs other than the one that you 

17    currently hold have you held at the census bureau?  

18               THE COURT:  Are you going to put in a CV?  

19               MR. SUBAR:  Yes, your Honor.  

20               THE COURT:  It is helpful if I have it in front 

21    of me.  

22               MR. SUBAR:  You can find that at the tab marked 

23    Exhibit 2.  

24               THE COURT:  Go ahead.

25         A.    I began in 1965 as a junior statistician.  I was 
                                                              1720

 1    only a grade 5.  Dr. Bailar was a grade 67, I think I heard 

 2    her say, when she started.  Basically, I worked in the 

 3    statistical research area of the census bureau in my first 

 4    years.  That was the more theoretical aspects of statistics 

 5    that I was working on at that point in time.  

 6               I then changed to go to work for what was called 

 7    the statistical methods division, which is more applied 

 8    aspects of what the census bureau does, including surveys 

 9    and censuses.  My first job there was to work on the 1970 

10    census, various functions related to it -- in particular, 

11    how to put a weight on the records from the sample.  

12               In the census, there are really two 

13    questionnaires, a short form and a sample questionnaire that 

14    goes to only some people that has extra questions.  You have 

15    to weight that up to make it look like what would have 

16    happened if you asked those questions of everyone.  I worked 

17    on that, variance calculations, and evaluations.  

18               After the 1970 census, I went to work on what is 

19    called current surveys in the census bureau.  These are the 

20    kinds of surveys that other people witnesses here have 

21    talked about, the monthly unemployment survey, the survey 

22    that produces the consumer price index.  

23               Following that, I began work on the 1980 census, 

24    and eventually held two positions in that regard.  I was 

25    senior demographic adviser to the person who had 
                                                              1721

 1    responsibility for planning the 1980 census.  I was also 

 2    chief of the decennial census division, which handles the 

 3    nuts and bolts of taking the census.  So my role in 1970 was 

 4    no statistical in nature.  My role in 1980 was more of the 

 5    day-to-day management.  

 6               Also in 1980 I was one of the senior census 

 7    officials involved in the decision about whether or not the 

 8    1908 census should have been adjusted for undercount.  After 

 9    the 1980 census I began my work on the 1990 census.  I was 

10    one of the key architects for planning the structure of how 

11    the 1990 census would be put together.  

12               I continued my role of senior policy review of 

13    the undercount through the 1990 census and also later in the 

14    decade I had prime responsibility for developing the 

15    outreach and promotion campaign that we actually used in the 

16    1990 census.

17         Q.    Specifically with regard to the evaluation of the 

18    1990 census, do you currently have any responsibilities?

19         A.    Yes.  I continue to have this policy-level review 

20    of the undercount activities going on at the census bureau 

21    that have continued even since the Secretary's decision.

22         Q.    With regard to the 1990 post-enumeration survey, 

23    the PES, did you have any responsibilities?

24         A.    I certainly did.  I was a member of the 

25    Undercount Steering Committee throughout the decade of the 
                                                              1722

 1    eighties that examined a post-enumeration survey, came to 

 2    the conclusion in 1987, and then after the stipulation again 

 3    looked at things, eventually leading to a choice in the 

 4    spring of 1991.

 5         Q.    In the course of your work, have you had occasion 

 6    to work with statistical methodologies of one sort or 

 7    another?

 8         A.    Yes.

 9         Q.    Can you tell the Court about that?

10         A.    They would be primarily in the area of sample 

11    selection, sample design, survey methodology, survey 

12    research, and obviously evaluations of censuses, in 

13    particular coverage.

14         Q.    Have you done any other work in your professional 

15    career besides your work at the census bureau?

16         A.    Yes, I have.  For three years in the evening I 

17    taught statistics at Georgetown University.

18         Q.    What title did you have at Georgetown?

19         A.    I was just an adjunct professor there.

20         Q.    Have you received any honors and awards in the 

21    course of your professional career?

22         A.    Yes.  I have received what the census bureau has 

23    as its highest award, which is only a bronze medal.  Also I 

24    received the Department of Commerce's silver medal, as well 

25    as various and sundry performance awards, including several 
                                                              1723

 1    SES performance awards.

 2         Q.    Are you a member of any professional societies?

 3         A.    Yes, I am.  I am a member and officeholder of the 

 4    American Statistical Association.  I am also a member of the 

 5    Population Association of America, and also a member of the 

 6    Association of Official Statistics, which is a subsection of 

 7    the International Statistics Institute.

 8         Q.    Mr. Bounpane, could you take a look, please, at 

 9    Defendants' Exhibit 2, which is in those notebooks to your 

10    right.  

11               MR. SUBAR:  I believe the Court is looking at 

12    that.

13         A.    Yes.

14         Q.    What is that document?

15         A.    That is my curriculum vitae.

16         Q.    Is it current and up to date?

17         A.    Yes, I think so.  

18               MR. SUBAR:  Your Honor, we offer Defendants' 

19    Exhibit 2 into evidence.  

20               MR. COHEN:  No objection.  

21               THE COURT:  Defendants' Exhibit 2 is admitted. 

22               (Defendants' Exhibit 2 for identification was 

23    received in evidence).

24         Q.    There has been some testimony in this case 

25    concerning the Nationality Academy of Sciences Panel on 
                                                              1724

 1    Decennial Census Methodology.  Did you have any 

 2    responsibilities with regard to that panel?

 3         A.    Yes, I did.

 4         Q.    Can you tell the Court what those 

 5    responsibilities were? 

 6         A.    Early in the eighties, when I was putting 

 7    together the plans for the 1990 census, there was a question 

 8    of whether we should actually make a contract with the 

 9    National Academy of Sciences, and, if we did, what would be 

10    the scope of it and how to obtain the budget to support it.  

11    Since that was my function at that time, the direct 

12    decisions for those questions came to me.  I certainly 

13    agreed with the decision to have the panel and to the scope 

14    it had.  I arranged for the budgeting to support it.

15         Q.    There has been some testimony in this case 

16    concerning the undercount research staff that was formed at 

17    the bureau in the early 1980s.  Did you play any role with 

18    regard to the creation of that staff?

19         A.    Yes, I did.

20         Q.    What was that role?

21         A.    The role there was very similar to what I just 

22    mentioned with regard to the National Academy panel.  Again 

23    there was a choice of should we have such a staff.  That 

24    decision came to me, and I said yes and agreed with that, 

25    also obtaining funding for it.  
                                                              1725

 1               Then a key decision of where it should be placed 

 2    within the census bureau.  I chose to assign it to the 

 3    associate director for statistical standards, who at that 

 4    time was Dr. Bailar.

 5         Q.    Mr. Bounpane, was there an undercount in the 1990 

 6    decennial census?

 7         A.    I think so, yes.

 8         Q.    Was it differential by race?

 9         A.    I think so, yes.

10         Q.    Do you know the true population of the United 

11    States, Mr. Bounpane?

12         A.    No, I don't.

13         Q.    Mr. Bounpane, could you take a look at 

14    Defendants' Exhibit 6, please.  

15               MR. SUBAR:  Which, your Honor, we have blown up 

16    as a chart.

17         Q.    What is Defendants' Exhibit 6?

18         A.    Briefly, that is a flow chart of the key steps in 

19    how the census is taken when you take a mail census.

20         Q.    Were you here in the courtroom when Professor 

21    Ericksen was testifying last week?

22         A.    Yes, I was.

23         Q.    Did you hear his description of the 1990 census 

24    operations?

25         A.    Yes, I did.
                                                              1726

 1         Q.    Did his description match the components of the 

 2    census as shown on this chart?

 3         A.    In general, yes, it did.

 4         Q.    Were there any exceptions?

 5         A.    Yes, I think there were.

 6         Q.    Could you turn, please, to Defendants' Exhibit 7.  

 7               MR. SUBER:  Perhaps that can be placed on the 

 8    easel alongside Defendants' Exhibit 6.  

 9               THE COURT:  You want both of them visible at the 

10    same time?  

11               MR. SUBAR:  Yes, please.

12         Q.    What is Defendants' Exhibit 7?

13         A.    That exhibit or chart up there is a list of some 

14    of the key exceptions to the mail out-mail back census.  My 

15    initial flow chart and Dr. Ericksen's description of the 

16    mail out-mail back census really agree in many ways, and I 

17    don't care to go through those again.  The Court has heard 

18    those.  

19               THE COURT:  I appreciate your consideration.

20         A.    There are a couple of things about it which 

21    didn't come out clearly, and that is we do not use mail 

22    out-mail back.  There are several key exceptions to that, 

23    for the reasons many witnesses have already said, that the 

24    mail would not necessarily work.  They are listed there.  

25               The first three tend to make sense unto 
                                                              1727

 1    themselves.  In institutions like prisons or on military 

 2    bases, group living arrangements like dormitories, it would 

 3    be very difficult to mail them out and ask people to mail 

 4    them back, etc.  So you actually send an interviewer there 

 5    and do the interviews directly in person.  

 6               The fourth line there, street and shelter 

 7    inhabitants, was probably one of the more publicized events 

 8    of the 1990 census.  In one night we hired a crew of 

 9    interviewers around the whole country to go out and visit 

10    every shelter that every local area had identified to us 

11    where people without homes might be staying.  Following that 

12    visit, we actually canvassed the areas of the city where you 

13    might find those people:  Park benches, in subway tunnels, 

14    and things like that.  Once again, using the mail census 

15    there would make no sense.  

16               The fifth one down -- by the way, that street and 

17    shelter one was new in 1990, as was the next one on there, 

18    selected public housing developments in major urban areas.  

19    Most major urban areas are done by mail out-mail back.  

20    However, there are large public housing projects in most of 

21    our major urban areas, and we were quite concerned that that 

22    mail out-mail back process might not be effective in those 

23    kinds of developments.  

24               In effect, we decided to not use mail out-mail 

25    back in certain public housing projects, but in fact have 
                                                              1728

 1    someone do the interviewing personally, someone from the 

 2    public housing project, to get around some of the problems, 

 3    once again, I think a couple of the other witnesses have 

 4    stated here to the Court.  

 5               I might point out that in New York City itself, 

 6    at the request of the city, in writing, we did not do that 

 7    here, because they felt the addressing structure in their 

 8    public housing projects was so sufficient that it would not 

 9    be necessary to do that.  

10               MR. COHEN:  Objection.  I move to strike that as 

11    hearsay, your Honor.  

12               MR. SUBAR:  That is an admission against a party, 

13    your Honor.  

14               THE COURT:  Overruled.  It will stand.

15         A.    The last four on there I think probably speak for 

16    themselves.  American Indian reservations; the low density 

17    areas, which are really the Rocky Mountain states, resort 

18    areas, Cape Cod, the ski areas of Colorado, etc., and the 

19    remote areas of Alaska, where once again it would not be 

20    reasonable to try to conduct the census by mail.

21         Q.    Mr. Bounpane, could you go back to Defendants' 

22    Exhibit 6 for just a moment and take the Court on a very 

23    quick spin around that chart.

24         A.    Basically, the first step is to create an address 

25    list, and once again I think Dr. Ericksen did point out how 
                                                              1729

 1    we did that, although we did some checks to make sure that 

 2    some of the things he pointed out were, in fact, corrected.  

 3               Then the massive job of actually affixing an 

 4    address label to 100 million questionnaires and delivering 

 5    those, asking people to mail them back, at this point in 

 6    time checking every one you got back against the list that 

 7    you had.  Now you have a list of addresses for which you 

 8    have not received a questionnaire, and you hire the staff of 

 9    temporary census interviewers to go out and conduct those 

10    interviewers in person.

11               MR. SUBAR:  Perhaps we could have these two 

12    charts now replaced by Defendants' Exhibits 8-B and 8-C.  

13    Your Honor, in the defendants' exhibit notebook, those are 

14    simply the second and third pages of Defendants' Exhibit 8.

15         Q.    Mr. Bounpane, during his testimony Dr. Ericksen 

16    mentioned some coverage improvement activities.  Did you 

17    hear his testimony in that regard?

18         A.    Yes, I did.

19         Q.    Was his list of coverage inprovement activities 

20    complete?

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

22         Q.    What are Defendants' Exhibits 8-B and 8-C?

23         A.    Those are a list grouped into four categories of 

24    a set of the coverage enhancement programs, I call them, 

25    rather than coverage improvement programs, that we actually 
                                                              1730

 1    used in the 1990 census.

 2         Q.    Could you describe them to the Court, please?

 3         A.    Sure.  Briefly, the first bullet there, the 

 4    primary housing unit checks and rechecks, are the kinds of 

 5    checks that have been described somewhat about making sure 

 6    we had a complete mailing list before we started.  I won't, 

 7    again, repeat those three because they have been described 

 8    here, except to say that they were purposely redundant.  

 9    That is, they intended to do the same thing, somewhat 

10    wasteful monetarily, but we felt that was necessary to make 

11    sure we had as adequate a list as possible, because that is 

12    crucial to the mail out-mail back census.  

13               The second set of enhancement programs there I 

14    put under the heading of second mailings.  We did feel it 

15    was very important to get as many cases back by mail as 

16    possible.

17         Q.    Why was that?

18         A.    There are really two reasons.  One is, of course, 

19    the cost of doing it in personal visit relative to doing it 

20    by mail.  We wanted to reduce that to the maximum extent 

21    possible.  

22               More important, the quality of the information 

23    you get on a mail return questionnaire, from our experience 

24    at the census bureau, is generally higher than that which 

25    you get in personal visits.  So that is the second reason.  
                                                              1731

 1               You can see here the kinds of things we did:  

 2    Send a reminder card.  I only want to point out in a little 

 3    more detail the second one there.  All of the 

 4    nondeliverables that the Postal Service returned to us we 

 5    gave to a set of census interviewers and asked them to try 

 6    to deliver them.  

 7               We might have made as simple a mistake as a 

 8    transposition error in the ZIP code.  That would make it 

 9    nondeliverable for the Postal Service, but one of our 

10    employees went out there, they would then be able to deliver 

11    that particular questionnaire.  

12               So we made every attempt that we knew to get a 

13    questionnaire in someone's hands 

14               The last one there is a special activity that 

15    only occurred here in New York City.  That is, in a couple 

16    of district offices that had extremely low mail return 

17    rates, we actually remailed questionnaires.  We didn't do 

18    that anywhere else in the country.

19         Q.    Could you continue on with Defendants' Exhibit 

20    8-C, please.

21         A.    Yes.  The third block, the way I have grouped 

22    them, is where you could get information from another source 

23    to check against the census and then try to do something 

24    about it.  The first one up there, called precensus local 

25    review, was also new in 1990 over previous censuses. 
                                                              1732

 1               Basically, what we did is produce a count of 

 2    housing units by block and turn them over to local officials 

 3    and ask them to check them against their records, whether 

 4    they be tax records, utility hookups or something like that, 

 5    and indicate for us the blocks where they thought we might 

 6    be deficient.  When they sent that information back to us, 

 7    we actually sent interviewers to the field to recheck those 

 8    blocks and add any missing units as appropriate.  

 9               The questionnaire responses are an attempt to get 

10    at some of the people that might be called not part of the 

11    nuclear family or something like that.  When people have to 

12    list who they should count on a census questionnaire, that 

13    is often difficult along the residence rules, so we included 

14    several check questions to make sure, to jog people's 

15    memories, like did you leave anyone off because you weren't 

16    sure if you should have put them down, or something like 

17    that.  

18               Any time someone answered that, a census 

19    interviewer went back to the House and conducted a coverage 

20    follow-up in person.

21         Q.    What is the next bullet?

22         A.    The next one is people who called in and said, I 

23    didn't get a questionnaire.  We got many more phone calls to 

24    this 800 phone line than we had anticipated.

25         Q.    Could you tell the Court about that 800 phone 
                                                              1733

 1    line?

 2         A.    This is the first time we had an 800 toll-free 

 3    help assistance line in conjunction with the census.  We 

 4    printed it out questionnaire, but, more important, we had it 

 5    advertised in various and sundry ways.  We also advertised 

 6    the fact that a questionnaire should have arrived at a 

 7    certain point in time.  Obviously, some people did not 

 8    recognize it, and they called this number.  

 9               When they did, we took down the address they gave 

10    us so that we could then go back and check our address list.  

11    If 3 Green Street, the Jones family at 3 Green Street called 

12    us up and said, I never got my census questionnaire, we 

13    checked to see that 3 Green Street was on our mailing list 

14    -- and if it was not there, we would have added it -- and 

15    then sent an interviewer there to conduct the interview in 

16    person.  

17               We did one more thing with that, and that is 

18    tally up the phone calls by geographic area.  So if we had a 

19    certain part of the city where we got lots of calls on our 

20    800 line with people saying they did not get a 

21    questionnaire, we actually outlined that on a map, sent 

22    interviewers back, and asked them to check the whole address 

23    list there and add any units that might have been missing.  

24    It might have been a new housing development that was not on 

25    the list or something like that.
                                                              1734

 1         Q.    What is the next item?

 2         A.    Finally, I will just briefly say that we also 

 3    built in our own self-checks against our own information 

 4    with a management information system to identify any 

 5    potential problems like the Pop 1 check that was described 

 6    earlier to the Court, some other checks similar in vein.  

 7    Where we had any indication there might be something wrong, 

 8    we sent someone back to check it.

 9         Q.    The final item there listed as final safety nets, 

10    could you tell the court about those?

11         A.    Yes.  After all of these, we still wanted to be 

12    sure we had as many people as possible.  Here are three of 

13    those final checks.  The parolee check that has been 

14    mentioned before, I should probably say a few words about 

15    that.  

16               In 1980 we had a process whereby we took lists 

17    independent of the census, matched them to the census, and 

18    then tried to follow up anybody who was on that 

19    administrative list that was not originally on the census 

20    questionnaire and add them to the census as appropriate.  It 

21    was called the nonhousehold sources program.  

22               The program proved to be not very effective, even 

23    though we thought it would be, because when you knocked on 

24    the door -- let's say you had John Jones on the driver's 

25    license list matched against the same address on the census 
                                                              1735

 1    and you knock on the door and say, does John Jones live 

 2    here?  Very often the person said never heard of John Jones.  

 3    So we were not able to add very many people to the census 

 4    through that.  So we did not do this in 1990, except in one 

 5    instance, which was parolees.  

 6               We did that because we thought those were people 

 7    who had very high probability of being missed in the census.  

 8    So we went state by state asking them for their parolee list 

 9    and, something more important, we asked them whether or not 

10    they relied on their list as an accurate way to find that 

11    parolee at that address.

12         Q.    When you say "they," who are you referring to?

13         A.    They, the state.  If the state said yes, then we 

14    checked the parolee against the census questionnaire.  If it 

15    was not on the census questionnaire, we added them to the 

16    census.  If the state said the list was not reliable, we 

17    would only add the parolee after a field check.

18         Q.    What is the next one?

19         A.    The last two I will just mention.  There was a 

20    Were You Counted campaign, which basically was a publication 

21    saying, with forms and newspapers and things like that.  One 

22    last chance, if someone thought they were missed, they 

23    simply had to fill this on it, send it in to us, we would 

24    check it.  

25               The post-census local review is the complement of 
                                                              1736

 1    what we discussed earlier as the precensus local review, 

 2    except it happened after the census, same operation.  

 3               Many cities turned in a number of blocks that 

 4    they thought we had done incorrectly in the post-census 

 5    review, in particular many big cities.  So the census bureau 

 6    sent its senior staff out to actually examine this in all of 

 7    the major cities around the country to find out what was 

 8    really going on with those complaints.

 9         Q.    Were you one of the senior bureau staff members 

10    who did some of the post-census local review field work?

11         A.    Yes, I was.

12         Q.    Where did you do that work?

13         A.    I did that in New York City.

14         Q.    What did your work consist of?

15         A.    It consisted of taking the block that the city 

16    challenged, taking the census address register -- list of 

17    addresses is what that is -- going out and seeing who was 

18    correct, the city or the census.

19         Q.    What were your findings in that regard?

20         A.    The blocks that I observed, the census was 

21    basically correct.  I want to be sure that what I say here 

22    is -- it is not so much that anything bad was going on.  It 

23    is that city lists are often not in the same format as 

24    census lists, they don't come down to census blocks the same 

25    way.  The addresses might have been across the street or 
                                                              1737

 1    something like that.  It is a complicated process.

 2         Q.    Again, perhaps we can change the charts here to 

 3    9-B and 9-C, which your Honor in your notebook are the 

 4    second and third pages of Defendants' Exhibit 9. 

 5               Earlier, Mr. Bounpane, I believe you said that 

 6    you were involved in promotion and advertising efforts with 

 7    regard to the 1990 census.  Do these two charts, 9-B and 

 8    9-C, address the promotion activities that the bureau was 

 9    involved in?

10         A.    Yes.  They are an outline of some of the key 

11    activities in the promotion outreach campaign.

12         Q.    Could you explain them to the Court, please.

13         A.    Briefly, the first bullet indicates the fact that 

14    we had an advertising campaign.  The only thing I want to 

15    point out there is that in 1990, for the first time, we 

16    hired four separate advertising agencies, not only a general 

17    agency, but also one for the black population, one for the 

18    Hispanic population, and one for the Asian population.  As 

19    you can see there, we also had special efforts for the 

20    American Indian population.  

21               Supplementing advertising, however, we had many 

22    other things.  The collaterals there, while it is nice to 

23    have some of those T-shirts and caps, the most important 

24    thing, I think, were the specialized brochures, a brochure 

25    that said, why should you as an African-American be counted 
                                                              1738

 1    in the 1990 census, or why should as a Thai-American be 

 2    counted in the 1990 census.  That, of course, was in Thai.  

 3               The question was, how do you get these kinds of 

 4    brochures to the people who should get them?  The mechanisms 

 5    are listed in some of the next bullets.

 6         Q.    What were some of those mechanisms?

 7         A.    You can see here the third bullet that I have 

 8    listed are additional minority efforts.  The promotion and 

 9    outreach campaign for 1990 was specially geared towards the 

10    minority communities.  The first line there is really the 

11    only one I think I might want to talk about a little about, 

12    our community awareness program.  

13               Two years in advance of the sense we hired 300 

14    community specialists.  They went to the major urban areas 

15    around the country and started to build a network with the 

16    local community organizations, so at the time the census 

17    actually came around, they were known, had some rapport.  

18    When se said, okay, it's time for all of us to get together 

19    to get people counted, they actually could give the 

20    brochures to the local organizations to hand out, could go 

21    to the meetings sponsored by the local organizations and 

22    hand them out, and things like that.

23         Q.    Could you turn to the next chart, please.

24         A.    The next page there lists some of the cooperative 

25    ventures that we did, asking local governments and in fact 
                                                              1739

 1    in some instances state governments to join with us on that 

 2    and to develop their own supplemental advertising and 

 3    promotion campaigns.  I think the only thing I want to say 

 4    here is that the cooperation we got was generally 

 5    outstanding, and some of the efforts in a lot of those 

 6    cities and states were really quite good.  

 7               The last section highlights five special types of 

 8    programs.  You can see just from the range of them, from 

 9    school to public housing to Headstart to migrant farm 

10    workers, the kinds of things we attempted to the.  

11               Once again, these were avenues to actually again 

12    hand out some of the key census collaterals.  On the 

13    religious projects we even went so far as to write sermons 

14    that religious leaders to give on April 1st, because it 

15    happened to be a Sunday in 1990.

16         Q.    With regard to all of these promotion activities, 

17    what population did they focus on?

18         A.    They really had two focuses.  Obviously, there 

19    was a general campaign.  But the bulk of these were aimed at 

20    the minority populations.

21         Q.    Did you hear Dr. Ericksen talk the other day 

22    about the difficulty of the process of completing census 

23    forms?

24         A.    Yes, I did.

25         Q.    Did the bureau provide assistance in that regard?
                                                              1740

 1         A.    Yes, we did.

 2         Q.    Again perhaps we can flip to the next couple of 

 3    charts, which would be Defendants' Exhibits 10-B and 10-C, 

 4    which are the second and third pages of Defendants' Exhibit 

 5    10 in the notebook.  

 6               What are these charts, Mr. Bounpane?

 7         A.    This is an outline of the key steps that we used 

 8    to help assist people in filling out their census 

 9    questionnaires.

10         Q.    Could you explain the steps to the judge?

11         A.    The first group are things we mailed out to 

12    people in advance of the census.  The one I would like to 

13    talk about is the multilingual motivational message.  

14               We isolated areas of the country that had large 

15    concentrations of people that spoke languages other than 

16    English, Spanish and several Asian languages.  Then, in 

17    advance of the census mailing itself, we mailed what we 

18    called a motivational message.  It was in English, Spanish, 

19    and these Asian languages, that basically said the census is 

20    coming, it's important to be counted, the questionnaire will 

21    be in English, if you need help in your language, here is 

22    how you get it.  Then the 800 phone number was actually 

23    listed there.

24         Q.    With regard to questionnaire assistance, what did 

25    the bureau do?
                                                              1741

 1         A.    Here again, new for 1990 as we instituted toll 

 2    free phone numbers in eight languages, English, Spanish, and 

 3    six Asian languages.  So if someone either got that 

 4    motivational message or if they saw the 800 number somewhere 

 5    else, for example, both Hispanic TV networks ran the 800 

 6    number as a banner during their TV programming around census 

 7    time, they could call up, answer questions, get help in 

 8    filling out their form.  

 9               As many people are aware, we got overloaded with 

10    phone calls, far more than our 800 line assistants could 

11    really handle.  One of the purposes of calling in on the 

12    Spanish 800 line was to say I got my questionnaire in 

13    English, but I really would like to have it in Spanish.  As 

14    Dr. Ericksen said, there was a small little message on the 

15    questionnaire itself, but we tried to get other ways for 

16    people to know that they could have their questionnaire in 

17    Spanish.  

18               When we saw that we could not satisfy all of 

19    those calls through the 800 line, we instituted the last 

20    thing there, which were questionnaire exchange centers, 

21    which basically said on one Saturday, which we advertised, 

22    if they brought their English question near in, we would 

23    hand them a Spanish questionnaire.  This was particularly 

24    used in the southern Texas area.

25         Q.    Going over to the next chart, what does that 
                                                              1742

 1    show?

 2         A.    This shows our language other than English 

 3    assistance.  The questionnaire itself was produced in both 

 4    English and Spanish.  But we realized that people had other 

 5    language difficulties, so we produced language guides in 32 

 6    languages.

 7         Q.    Can you give the Court some examples of some of 

 8    those languages?

 9         A.    Yes.  They were Laotian, Cambodian, Chinese, 

10    perhaps Polish, German, Russian, Hebrew, Arabic.  They 

11    spanned the whole range of immigrants in this country.  

12               Here in New York, the third line there, the 

13    regional office itself, even though headquarters only 

14    produced the questionnaire in English and Spanish, produced 

15    the questionnaire in Chinese and used it here in the 

16    enumeration of Chinatown and duplicated it and sent it to 

17    both Houston and San Francisco for use there.  

18               The last section there has to do with the 

19    interviewer, him or herself.  Obviously, the interviewer is 

20    a way to assist the person in filling out the questionnaire.  

21    I only want to talk about one thing there, which is the word 

22    "indigenous hiring."  

23               Our goal in 1990 was to give an interviewer a 

24    cluster of blocks to work in.  We very much wanted somebody 

25    who lived in that cluster of blocks to actually do the 
                                                              1743

 1    enumeration for that cluster of blocks.  That was our 

 2    primary hiring tool.  We used the word "indigenous" to refer 

 3    to that.  Obviously, if they spoke a language in that 

 4    particular area, then the interviewer needed to be able to 

 5    speak that language as well as English.

 6         Q.    How many census interviewers were hired in 

 7    connection with the 1990 census?

 8         A.    About 300,000.

 9         Q.    How many people in toto were hired as temporary 

10    workers by the bureau in connection with the 1990 census?

11         A.    About 500,000.

12         Q.    Could you flip in your book, please, to 

13    Defendants' Exhibit 3.  

14               THE COURT:  Why don't we take a break at this 

15    point.  We will come back and resume at Defendants' Exhibit 

16    ten to four.  

17               (Recess) 

18               THE COURT:  Where is Mr. Bounpane?  

19               MR. SUBAR:  While he is coming out, at Mr. 

20    Rifkind's request, we have moved the easels so they are 

21    repositioned so people don't have to move around the 

22    courtroom.

23         Q.    Mr. Bounpane, do you have Defendants' Exhibit 3 

24    there?

25         A.    Yes, I do.
                                                              1744

 1         Q.    What is that document?

 2         A.    This is a document that explains many of the 

 3    planned improvements for the 1990 census, some of which I 

 4    outlined here briefly in the courtroom and some others as 

 5    well.

 6         Q.    Did you have any role in the preparation of this 

 7    document?

 8         A.    Yes.  My office prepared this document.

 9         Q.    Were the planned improvements that are mentioned 

10    in the document in fact adopted in the 1990 census?

11         A.    Yes, I think almost all of them are.

12         Q.    Were all of the improvements listed in the 

13    document mentioned in your testimony here today?

14         A.    No.

15         Q.    There were some that were in there that you 

16    didn't go through in court?

17         A.    That is correct.  

18               MR. SUBAR:  Your Honor, we offer Defendants' 

19    Exhibit 3 into evidence.  

20               MR. COHEN:  No objection.  

21               THE COURT:  3 is admitted. 

22               (Defendants' Exhibit 3 for identification was 

23    received in evidence)

24         Q.    Could you flip, Mr. Bounpane, to Defendants' 

25    Exhibit 4 in the binder.  What is that document?
                                                              1745

 1         A.    This document is a brief description of the plans 

 2    and operations for actually conducting the 1990 census.  At 

 3    the back of it, the second piece of it, is a brief 

 4    description of the post-enumeration survey.

 5         Q.    Is this document an accurate description of the 

 6    census and the PES?

 7         A.    Yes, I think it is.  

 8               MR. SUBAR:  Your Honor, this document is 

 9    excerpted from the administrative record which is in 

10    evidence and which has been moved over to that corner of the 

11    courtroom.  

12               THE COURT:  Are you offering it?  

13               MR. SUBAR:  Since it is part of an exhibit that 

14    is in evidence, I see no need to.  If you Court wants to 

15    receive it as a separate exhibit, that is all right too.  

16               THE COURT:  No, whatever you want.  

17               MR. SUBAR:  Thank you.

18         Q.    Mr. Bounpane, we have heard about the PES over 

19    the course of this trial.  What is the purpose of the PES?

20         A.    The purpose of a PES is to take a survey 

21    independent of the census, then estimate total population 

22    from that survey, compare that estimate of total population 

23    with the census count itself.  Then the difference is an 

24    estimate of potential coverage error in the census.

25         Q.    Are there other methods that could be used to 
                                                              1746

 1    estimate coverage error in the census?

 2         A.    Yes, there are.

 3         Q.    Could you give the Court some examples of those?

 4         A.    Unlike the dual-system estimate that is used in 

 5    the post-enumeration survey for 1990, you could do a 

 6    post-enumeration survey that is a single-system estimate, 

 7    just a simple sample survey without the capture/recapture 

 8    methodology.  You could use economic analysis which has been 

 9    discussed.  You could use tracing samples, which mainly 

10    means you try to identify some people in advance of the 

11    census, trace them through the census to see if they 

12    actually get counted.

13         Q.    Could you take a look at the Defendants' Exhibit 

14    12, which is the chart that is being put up on the easel 

15    now.  Could you tell the Court what that is?

16         A.    That is the beginning of a flow chart to explain 

17    how the post-enumeration survey using a P and E sample 

18    works.

19         Q.    How does the PES work?  First of all, what is the 

20    P sample?

21         A.    When we first started doing post-enumeration 

22    surveys, we simply selected a land area, went back and 

23    reinterviewed it, and compared that reinterview against the 

24    census to make some estimate of coverage error.  

25               There are a couple of problems with that 
                                                              1747

 1    approach.  One is that if you only match within the 

 2    specified land area, like a block, the person could still be 

 3    in the census but you wouldn't know it.  Also, the census 

 4    itself makes some errors.  

 5               Therefore, in order to try to measure both, the 

 6    process is split into two pieces, a P sample and an E 

 7    sample.  Each of them is aimed at a different purpose, the P 

 8    trying to get an estimate of total population, the E trying 

 9    to get an estimate of erroneous enumerations in the census.

10         Q.    Mr. Bounpane, just one moment.  

11               MR. SUBAR:  Your Honor, I notice now at Mr. 

12    Rifkind's request they have moved the charts, they are right 

13    in front of the fan which makes the PES sort of shaky.  

14    Perhaps we could have the fan turned off.  

15               THE COURT:  The shake is not bothering anybody, 

16    is it?  

17               MR. SUBAR:  It was bothering me a trifle.  

18               THE COURT:  Go ahead, turn the fan off.  

19               MR. SUBAR:  Thank you.

20         Q.    Okay, Mr. Bounpane.  I think you were describing 

21    the two samples in the PES.

22         A.    Yes.  If I could start with the P sample.  Again, 

23    I will be relatively brief, since I think this has been 

24    explained to the Court before.  

25               A sample of blocks gets selected.  You then ask 
                                                              1748

 1    someone to list all the housing units in the blocks.  That 

 2    is a second chart there.  Then you send interviewers out to 

 3    actually interview all the people listed in those Houseing 

 4    units.  

 5               The thing that is important at this step in the 

 6    process is that when you knock on the door you ask, please 

 7    give me the name of everyone who is living here.  You then 

 8    have to also ask, where is your residence, was your 

 9    residence, on April 1st, because in order to make that 

10    coordinate with the census you need to know where that 

11    person was on April 1st.  

12               You then begin the step the third box down on the 

13    left there called matching, which has been discussed.  

14               Again, if I could go back to one of the earlier 

15    examples, if you have a Mr. Jones at 10 Oak Street in the 

16    PES and you have a Mr. Jones at 10 Oak Street in the census, 

17    those are relatively easy, and a lot of them are just like 

18    that.  The computer match did those very easily.

19         Q.    Mr. Bounpane, what is the geographic scope of 

20    matching?

21         A.    In the PES that we actually used, we called it a 

22    ring, which is basically the block that the PES was in as 

23    well as all the blocks surrounding that particular block.  

24    That was the ring in an urban area.  The ring was slightly 

25    different in other parts of the country, but that was the 
                                                              1749

 1    primary way that the ring was developed.  

 2               The matching occurred within the ring.  The 

 3    reason for that is if the house was simply across the street 

 4    on the other side in the census, you didn't want to go 

 5    through the problem of saying you couldn't find it in this 

 6    particular block when it was simply on the other side of the 

 7    street by mistake.  That was the intent of using the ring 

 8    rather than the block itself.

 9         Q.    With regard to matching, do you agree with 

10    testimony that was given earlier today by Dr. Fienberg that 

11    the matching process is a very detailed process?

12         A.    Yes, I think so.

13         Q.    Have you ever done this sort of matching 

14    yourself?

15         A.    Yes, I have.

16         Q.    Can you tell the Court about your experience?

17         A.    In conjunction with the court case in 1980, the 

18    court asked the census bureau to match a list of names 

19    against the census files and report back to the court how 

20    many of those were actually included in the census.  I was 

21    responsible for managing that match and, in fact, spent two 

22    weeks in our processing office matching records myself.  

23               The matching there was slightly different than 

24    here in the 1990 PES.  That is, it wasn't clustered by block 

25    as the 1990 PES is.  But still the same problems tend to 
                                                              1750

 1    occur.  The easy ones go easy, and then you get down to a 

 2    set of ones that are not so easy.  

 3               Are Linda Martin and Linda Martinez the same 

 4    person?  Is J., initial J, Robinson and John Robinson the 

 5    same person?  Where Hispanics use a maiden middle name, 

 6    sometimes use that as the last name, is that the same 

 7    person?  In some of the Oriental name where the family name 

 8    in fact comes first sometimes and sometimes not, is that the 

 9    same person?  

10               Anyhow, when you get to cases like that, more 

11    judgment is required for saying whether or not those really 

12    are the same people.

13         Q.    That was your experience in 1980, you are saying?

14         A.    It certainly was, while I sat there and tried to 

15    match those cases.

16         Q.    In 1990, what sorts of cases would go beyond the 

17    computer matching stage to clerical matching?

18         A.    I think the latter step that I just answered to 

19    you are the kinds of things that the computer itself could 

20    not be sure was a match and so therefore would have not been 

21    able to do it, and a clerical staff would have been required 

22    to look at them.

23         Q.    If there is no match found in the PES process at 

24    that stage, what happens next?

25         A.    The next thing that happens is if we think we 
                                                              1751

 1    could get more information from the field, we reassign the 

 2    case to an interviewer and ask them to go back and get 

 3    additional information.  That is the step there that is 

 4    called "Additional Field Interviews."  They could find out 

 5    whether J. Robinson and John Robinson was in fact the same 

 6    person by asking that.  So those went back later, almost 

 7    fall of 1991.

 8         Q.    What is the next step in the process?

 9         A.    When the answers from those questionnaires come 

10    back, you take the information you get from these additional 

11    field interviews and you are able to match more at that 

12    particular point.  You finally do get left with a residual 

13    set of cases for which you cannot make up your mind one way 

14    or another.  Those are the ones we have heard described as 

15    match status had to be imputed or something like that.  I 

16    included that in the last box on the left there with the 

17    various computer edits that had to be done on the P sample 

18    before we went into dual-system estimation.

19         Q.    Is there any difference in terms of matching 

20    between P sample matching and E sample matching?

21         A.    Yes, there is.  First of all, as the arrow points 

22    out here, we did use the exact same sample for both the P 

23    sample and the E sample.  As Dr. Ericksen did testify, you 

24    could really view it as you have the two piles and to match 

25    the P against the E, you do that, and then you also just 
                                                              1752

 1    reverse it and match the E against the P.  

 2               But what you have to do one additional step, and 

 3    that is you have to match the E against itself.  That is the 

 4    only way you are going to get a measure of duplicates in the 

 5    census, and it is the necessary piece to get in order to get 

 6    an estimate of erroneous enumerations in the census.  

 7               MR. SUBAR:  Perhaps we could have Defendants' 

 8    Exhibit 13 put up on the easels now.

 9         Q.    What is Defendants' Exhibit 13?  What does that 

10    show?

11         A.    That is really a continuation of the chart I was 

12    just looking about.  The little box at the bottom there on 

13    the righthand one, which says dual-system estimation, is 

14    really the first box on the next chart.  So after completing 

15    all the field operations for both the P and the E sample and 

16    with the allocations and computer editing that had to be 

17    done, we then go to the next step, which is by poststratum 

18    produce a dual-system estimate of total population.

19         Q.    How does that process work?  Is there another 

20    chart that might help you explain it?

21         A.    Yes.  I have a chart that might explain it.  

22               THE WITNESS:  Your Honor, is it okay if I walk up 

23    to and talk from that one?  

24               THE COURT:  Sure.  Go right ahead.

25         Q.    Perhaps you could help us find the right chart.  
                                                              1753

 1    You are referring now to Defendants' Exhibit 15-A, correct?

 2         A.    I am referring, yes.  

 3               Basically, this is how the dual-system estimate 

 4    works.  You take the census, which is the first system, and 

 5    you count people as to whether they are in or out of the 

 6    census.  You cross-classify that by the PES, that is the 

 7    second system, about whether or not people are in and out of 

 8    the PES.  

 9               In doing that, so you have the cases here, you 

10    really can observe only three cells out of this table.  You 

11    can observe cases that match, you can observe cases that do 

12    not match, and you can observe the total number of people 

13    that the census counted.  

14               What we are most interested in, however, is an 

15    estimate of total population.  That is what we would like to 

16    have.  So from this table of which we can fill in this cell, 

17    this cell, and this cell, we would like to get this cell.  

18               Before I explain how we do that, I just would 

19    want to make a couple of more comments here.  Every nonmatch 

20    is not necessarily a miss in the census.  It is just that 

21    you couldn't match it.  Sometimes it really is a miss in the 

22    census.  Sometimes it's because you made a mistake in 

23    matching.  That is why you try to measure matching error.  

24               Sometimes it is because of what we call a 

25    geocoding error.  That is, the case is really in the census 
                                                              1754

 1    but it is so far away from the ring, by mistake, that you 

 2    don't find it in the ring, so you call it a nonmatch but it 

 3    really is in the census.  

 4               The complement happens down here in the census 

 5    count.  The total census count is as it is, but it does have 

 6    some errors in it, like duplicate people, like the ones who 

 7    were born after April 1st but still counted, and things like 

 8    that.

 9         Q.    Mr. Bounpane, you were just pointing to the box 

10    labeled 1 plus 3, correct?

11         A.    Yes, I was, I'm sorry, total census.  

12               So even though these names go on here, there are 

13    some problems with accurately putting all of those numbers 

14    in the cell.  That is one problem.  But the biggest one is 

15    how do I now get this when I don't have all the information 

16    to add it up.  

17               (Continued on next page) 

18    

19    

20    

21    

22    

23    

24    

25    
                                                              1755

 1         Q.    And "this" is total population?

 2         A.    This is, excuse me, total population.

 3               If I can pull this overlay over, I will explain 

 4    how we went around about doing that.

 5               MR. SUBAR:  The overlay is marked, your Honor, as 

 6    Defendant's Exhibit 15 B.

 7         A.    In order to try to estimate total population from 

 8    the dual system estimate, you have to make an assumption, 

 9    and it's one of the many assumptions in the process but one 

10    of the major ones, and it's called the independence 

11    assumption.

12               Really what it says is if I can assume that the 
                                                                 

13    ratio of this, that is matches, cell 1, to total PES, which 

14    is cell 1 plus 2, is the same as the ratio of the total 

15    census count is to the total population, if I can make this 

16    assumption, which I have written here, then it's simple 

17    algebra to solve for the total population.

18               And it's often written this way that the capture 

19    probability in the PES, the probability that somebody is 

20    really in the post-enumeration survey has to equal the 

21    probability that somebody is in the census.  That's often 

22    the way people explain this assumption.

23               Now, nobody really things that assumption is 

24    fully true and as long as it is almost correct, that's 

25    really not a problem here.
                                                              1756

 1               When you make this assumption, you then can 

 2    algebraically solve here and it says the total population is 

 3    simply the total PES times the total census divided by the 

 4    matches.

 5               And then we actually said our dual system 

 6    estimate will use this computational form to be an estimate 

 7    of the total population.

 8               Now, when you have done that, I now know total 

 9    population, I now know matches, PES nonmatches and total 

10    census, now it's simple subtraction to fill in all the other 

11    cells on the table, even though we really don't do that, 

12    because we didn't care about the other cells, we really only 

13    care about total population.

14               I might point out, however, that though we are 

15    only interested in the total population, this cell is of 

16    importance.

17         Q.    "That" is cell 4?

18         A.    This is cell 4 and often called the fourth cell.

19               You can see it is classification of people not in 

20    the PES and not in the census, people missed in both.  It's 

21    been what has been described as correlation bias in the 

22    courtroom, your Honor.

23               Obviously you cannot directly observe people 

24    missed in both the PES and the census.  But once you have 

25    made your estimate of total population, by subtraction and 
                                                              1757

 1    all the other cells, you can put a number in here.

 2               So this estimate of total population using this 

 3    assumption does make some account for people missed in both 

 4    the PES and the census.

 5               Basically, this operation has to happen 1,392 

 6    times, once for each of the post-strata.

 7         Q.    What happens after that operation happens 1,392 

 8    times?

 9         A.    Then I now have the total population for each 

10    post-stratum --  excuse me --  I now have the estimate of 

11    total population for each of the 1,392 strata, that is the 

12    dual system estimate, I also have the census count for each 

13    of those post-strata, and the ratio of them or a calculation 

14    from the ratio of them is the adjustment factor, that is, 

15    the factor you would have to multiply the census by in order 

16    to come up with this estimate of total population.

17         Q.    At this point do we switch back to Defendant's 

18    Exhibit --

19         A.    What I just described here is the second box in 

20    this flow chart.

21         Q.    That is Defendant's Exhibit 13?

22         A.    Yes, Defendant's Exhibit 13.

23               By post-stratum we produced a dual system total 

24    population, produced a dual system estimate of total 

25    population and then by comparing that estimate of total 
                                                              1758

 1    population to the census count we get an estimated 

 2    adjustment factor.

 3         Q.    There are four more boxes on that chart that have 

 4    to be gone through, correct?

 5         A.    Yes, they are.

 6         Q.    What are they?

 7         A.    Well, the next box is now that I have calculated 

 8    these 1,392, what we call raw adjustment factors, we then go 

 9    through the operation of smoothing those 1,392 factors which 

10    I am not going to describe here today.

11         Q.    Perhaps we can have the next chart put up in 

12    place of the one overlay, which is Defendant's Exhibit 16.

13               (Pause)

14               What does this chart show?

15         A.    This is a chart that refers back to the fourth 

16    box in the chart, it says now by block part with 

17    post-stratum, apply adjustment factor to census counts.

18               Your Honor, the best way I know how to describe 

19    that is if we can envision a table that has 1,392 columns 

20    and about 5 million rows.  This is one row for each occupied 

21    block, which is about, I think there are 4.7 million or so 

22    occupied blocks in the census.

23               If you take each block, they have people in 

24    various of the post-strata.  This block has some owners and 

25    some renters and some males and some females, it has some 
                                                              1759

 1    people 5 to 15 and has some people 30 to 45, so you can 

 2    split each block up into the count that fit in each 

 3    post-stratum.

 4               We happen to call each of those little pieces a 

 5    block part.

 6               Then the adjustment factor for each post-stratum 

 7    is multiplied down for every single block part within that 

 8    post-stratum.

 9               This is where the homogeneity assumption comes 

10    in.

11               The assumption here is that every block part 

12    within every post-stratum has to have, you are assuming has 

13    the same undercount rate because you are going to multiply 

14    that same factor all the way down this whole table for every 

15    single time a block part appears in that column.

16               Now, usually what happens is you get a number 

17    like 1.03 multiplied by ten and it doesn't come out even, it 

18    comes out to 10.3, and there is a random rounding process 

19    that will round it either to 10 or 11.  It will not 

20    necessarily go to 10, as one might thing, although it 

21    generally does, in order to keep the thing additive all the 

22    way around, it's just a mathematical quirk of how the 

23    rounding works.

24         Q.    What is done after that process is completed?

25         A.    When that is completed, I now have adjusted block 
                                                              1760

 1    part counts.  I can simply add them back up and I now have 

 2    an adjusted count for the block.  

 3               So now I have adjusted block counts for each of 

 4    the approximately 5 million occupied blocks in the United 

 5    States, and once I have that, I can move --  I can decide 

 6    how to get those people into the census.  And that's the 

 7    fifth box on this chart.

 8         Q.    That is, again, on Defendant's Exhibit 13?

 9         A.    That's Defendant's Exhibit 13, yes.

10               It says if I had a block that used to have 100 

11    people and the adjusted count to that block is now 102 

12    people, I have to, as the thing says here, create adjustment 

13    people in that case.

14         Q.    Are adjustment people real?

15         A.    No.  What you would do is take two of the people 

16    already in that block and duplicate their record.

17         Q.    That is, with regard to so-called adjustment 

18    people who are created, is that correct?

19         A.    Yes.  The two additional people.  My estimate was 

20    102, the census count was 100, I had to create two people.  

21    It's taking two of the hundred that is already there and 

22    doubling their weight, in effect, recreating them.

23         Q.    I notice that the block says create/subtract 

24    adjustment people.

25               Are the adjusted people subtracted by the census 
                                                              1761

 1    real?

 2         A.    Yes.  The reverse would happen if we had a block 

 3    of count of 100 in the census and a PES estimate for that 

 4    block was 98.  Now two people have to be removed in reverse 

 5    of the other situation, so two people who were actually 

 6    enumerated from that block would be deleted.

 7         Q.    So Mr. Bounpane, is there a chance that someone 

 8    who is in this courtroom right now was subtracted from the 

 9    census when the PES base dual system estimates were 

10    prepared?

11         A.    Yes, there is.

12         Q.    What is the next step in the process?

13         A.    The final step in the process now, and I have 

14    adjusted counts for each of the blocks in the United States, 

15    is you can add them up to any level.  Add up the blocks to a 

16    county, add up the blocks to a city, add up the blocks to a 

17    state, add up the blocks to the total U.S. population.

18         Q.    Mr. Bounpane, turn to Exhibit 5.

19               What is Defendant's Exhibit 5?

20         A.    Defendant's Exhibit 5 is a paper by Howard Hogan 

21    which describes in a lot more detail than I just did the 

22    post-enumeration survey process.

23         Q.    Is it an accurate description of that process?

24         A.    Yes, I think so.  

25               MR. SUBAR:  Your Honor, again, that's part of the 
                                                              1762

 1    administrative record.  

 2    BY MR. SUBAR:

 3         Q.    You mentioned earlier, Mr. Bounpane, that the 

 4    census used a mailout/mailback procedure.

 5               Did the PES use that kind of procedure?

 6         A.    No, it didn't.

 7         Q.    Which types of interviews were done closer in 

 8    time to census day, nonresponse followup interviews done in 

 9    the context of the census or PES interviews?

10         A.    Nonresponse follow-ups in conjunction with the 

11    census.

12         Q.    You mentioned that language assistance guides 

13    were available in 32 languages for the census.

14               Did the same hold true for the PES?

15         A.    No, it didn't.

16         Q.    You mentioned that the Bureau tried to use 

17    indigenous interviewers for the census.  Were indigenous 

18    interviewers used for the PES?

19         A.    No.

20         Q.    Mr. Bounpane, were you here the other day when 

21    Dr. Bailar testified?

22         A.    Part of the time, yes.

23         Q.    Did you hear her testify as follows, and I am 

24    reading from page 1,332 of the trial transcript:

25               "Q.    When you left the Bureau, was work 
                                                              1763

 1    on-going concerning the possibility of an adjustment in 

 2    1990?  

 3               "A.    No.

 4               "Q.    Why not?

 5               "A.    Because the decision had been made not to 

 6    adjust the census.

 7               "Q.    Was there a point at which work on 

 8    adjustment resumed at the Bureau?

 9               "A.    Yes, after the stipulation and order.

10               "Q.    That was in 1989?

11               "A.    Yes." 

12               Did you hear that testimony?

13         A.    Yes.

14         Q.    Do you know when Dr. Bailar left the Bureau?

15         A.    Yes.  It was January 1988.

16         Q.    Between the time that she left the Bureau and the 

17    time that the stipulation and order was entered in this 

18    case, did work continue at the Bureau with regard to the 

19    PES?

20         A.    Yes, it did.

21         Q.    In that same time period, did the undercount 

22    research staff continue to exist?

23         A.    Yes, it did.

24         Q.    During that time period, did the Bureau intend to 

25    conduct a PES with regard to the 1990 census?
                                                              1764

 1         A.    Yes, it did.

 2         Q.    Could you please take a look at the charts which 

 3    are labeled Defendant's Exhibits 12 and 13, and we will pull 

 4    12 out for just one second, the one leaning against the wall 

 5    with all the filled in blue boxes.

 6               (Pause)

 7               Could you take a look at those two charts, 

 8    please, Mr. Bounpane, and tell the judge which steps in the 

 9    PES process the Bureau still intended to perform after the 

10    decision was made not to adjust, but before the entry of a 

11    stipulation and order in this case?

12         A.    All of the steps on the chart that is to my right 

13    hand here -- I don't know if that is  -- 

14         Q.    That is Exhibit 12.

15         A.    Exhibit 12.

16               --  were continuing, and the chart to my left? 

17               Is that Exhibit 13?

18         Q.    Yes.

19         A.    The first, second and third box would also 

20    continue.

21               Where you would stop, where we stopped was 

22    somewhere about the fourth box on the chart, because if you 

23    are not going to adjust, it made no sense to carry it all 

24    the way down to the block level.  

25               Basically, all of the PES activities, including 
                                                              1765

 1    interviewing, computer matching, all of those, the dual 

 2    system estimate and smoothing continued.

 3         Q.    Did the Census Bureau conduct address rehearsal 

 4    of the 1990 census?

 5         A.    Yes.

 6         Q.    Where and when?

 7         A.    It was conducted in March-April 1988 in three 

 8    places; St. Louis, Missouri, some rural counties in Missouri 

 9    and some rural counties in Washington state.

10         Q.    Were any PES procedures included in that dress 

11    rehearsal?

12         A.    Yes, they were.

13         Q.    Which ones?

14         A.    For sure in St. Louis and rural Missouri, I can't 

15    remember precisely whether or not we also did those PES 

16    activities in Washington state, and it was the same category 

17    that I just mentioned to you, that is, we did the PES, the 

18    matching, all that interviewing, the dual system estimating 

19    and smoothing for the dress rehearsals.

20         Q.    Was that after Dr. Bailar left the Bureau, but 

21    before the entry of a stipulation and order in this case?

22         A.    Yes, it was.

23         Q.    Mr. Bounpane, let me read to you some testimony 

24    that Dr. Ericksen gave in this case last week on page 438 of 

25    the trial transcript.
                                                              1766

 1               "Q.    So using the range from page 12 of P6, 

 2    there would be between 2,000 and 6,000 fabrications in the 

 3    PES, is that right?

 4               "A.    No, it's not right, because I explained to 

 5    you before in the current population surveys there is no 

 6    procedure for detecting fabrications and redoing the 

 7    interviewers' work.  In the post-interview work there was a 

 8    procedure detecting fabrications and when they were found 

 9    the interviewers' work was redone." 

10               Mr. Bounpane, are you familiar with the current 

11    population survey?

12         A.    Generally, yes.

13         Q.    Is there a procedure for detecting fabrications 

14    and redoing interviewers' work in the current population 

15    survey?

16         A.    Yes, there is.

17         Q.    You mentioned earlier that you were a member of 

18    the undercount steering committee.

19               Did that committee make a recommendation with 

20    regard to adjustment?

21         A.    No, it didn't.

22         Q.    Did that committee make a recommendation with 

23    regard to the accuracy of the adjusted counts?

24         A.    Yes, it did.

25         Q.    What is the difference between those two, that 
                                                              1767

 1    is, what is the difference between a recommendation with 

 2    regard to adjustment and a recommendation with regard to the 

 3    accuracy of the adjusted counts?

 4         A.    Basically, the Census Bureau is a statistical 

 5    agency and our role ought to be to analyze statistical 

 6    issues and make recommendations on those.

 7               The issue about whether or not to adjust the 1990 

 8    census obviously is heavily based on technical issues, but 

 9    also includes other issues as well; legal policy and so on.

10               So at the Census Bureau we concentrated on 

11    answering the technical questions and, therefore, our goal 

12    was to answer the question did we, as a group, feel that 

13    adjusted counts would be more accurate than the census 

14    counts themselves.  That would then feed into an eventual 

15    choice about whether or not to adjust.

16         Q.    When was the undercount steering committee 

17    recommendation made?

18         A.    About June 1991.

19         Q.    And what was that recommendation?

20         A.    The committee concluded that on average, they 

21    felt that adjusted counts were more accurate than census 

22    counts.

23         Q.    Did you concur in that recommendation?

24         A.    No, I didn't.

25         Q.    Did you have occasion to provide your views on 
                                                              1768

 1    adjustment directly to Secretary Mosbacher?

 2         A.    Yes, I did.

 3         Q.    And under what circumstances was that?

 4         A.    There was a meeting at the request of the 

 5    Secretary where various members of the undercount steering 

 6    committee were asked to come and discuss their views on the 

 7    accuracy of adjusted counts so that he could listen to them 

 8    and question and hear in person the kinds of things that we 

 9    had written in our report.  

10               THE COURT:  Have you fixed a time when this 

11    occurred?  

12               THE WITNESS:  It was early July 1991.  I don't 

13    know the precise date.  

14               THE COURT:  Early July.  

15    BY MR. SUBAR:

16         Q.    It was before his decision?

17         A.    It was before his decision, yes.

18         Q.    Was it after the undercount steering committee 

19    recommendation was given?

20         A.    Yes, it was.

21         Q.    Could you take a look, please, at Defendant's 

22    Exhibit 75, which I think you will find in the second 

23    binder.

24               There might not be a tab in front of the exhibit, 

25    but it should be at the very end:?
                                                              1769

 1         A.    I can only see a 74.  I don't have a 75.  Sorry.  

 2               MR. SUBAR:  Your Honor, I happen to have an extra 

 3    copy which, if I may approach the witness, I can hand to 

 4    him.  

 5               THE COURT:  Sure.

 6               (Handing to the witness)

 7               (Pause)

 8               MR. SUBAR:  Your Honor, Defendant's Exhibit 75 is 

 9    a copy of some material from the administrative record in 

10    evidence.  

11               THE COURT:  All right.  

12    BY MR. SUBAR:

13         Q.    What is Defendant's Exhibit 75?

14         A.    It's a summary of the meeting that was held with 

15    the Secretary of Commerce.  It apparently was held July 8, 

16    1991.

17         Q.    Have you had occasion to read that document?

18         A.    Yes, I did.

19         Q.    Is it an accurate account of what happened at the 

20    meeting?

21         A.    Generally, yes, I think.

22         Q.    Were there any members of the undercount research 

23    committee who supported the proposed adjustment or who 

24    recommended that the adjusted counts were more accurate than 

25    unadjusted counts were present at that meetings?
                                                              1770

 1         A.    Yes, there were.

 2         Q.    Who were they?

 3         A.    Bob Fay and Bob Gross.

 4         Q.    Were they given an opportunity to make their 

 5    views known?

 6         A.    Yes, they were.

 7         Q.    Did the Secretary engage them in a discussion of 

 8    the technical aspects of the adjustment issue?

 9         A.    Yes, he did.

10         Q.    Could you please flip to the last page of this 

11    document and just read aloud the very last sentence.

12         A.    "On the Secretary's behalf, Soll thanked the 

13    group for its enlightening debate and requested any further 

14    analysis or reflections be provided to the Secretary as soon 

15    as possible."

16         Q.    Did you take Mr. Soll up on his offer?

17         A.    Yes, I did.  

18               THE COURT:  Who is Soll? 

19               THE WITNESS:  It was Bruce Soll, and he was one 

20    of the people who worked for the Secretary of Commerce.

21         Q.    Backing up just a moment, Mr. Bounpane, could you 

22    just briefly recap what happened at that meeting?

23         A.    Well, basically we thought we had approximately 

24    an hour, the meeting turned out to be a little over two 

25    hours long.
                                                              1771

 1               The primary purpose of the meeting was that the 

 2    Secretary wanted to hear from us at the Census Bureau.  He 

 3    had read our document, he had made a list of questions from 

 4    reading that document and basically he would pose the 

 5    questions and then those of us who were there at the meeting 

 6    would talk about our views on those particular questions.

 7               At the end of the meeting, a relatively short 

 8    amount of time,Dr. Darby, from the Commerce Department, also 

 9    expressed his views.

10         Q.    Again, at the end of the meeting, Mr. Soll 

11    invited further comments with respect to the adjustment 

12    issue?

13         A.    Yes, he did, because the meeting had gone on 

14    considerably longer than anticipated and it was still being 

15    productive and I think that's why he asked for more 

16    comments.

17         Q.    Was the focus of the meeting primarily technical 

18    or otherwise?

19         A.    It was almost entirely technical.

20         Q.    And for how much of the meeting was the Secretary 

21    in attendance?

22         A.    He left with about 15 minutes to go, so he was 

23    there for the first hour and 45 minutes, something like 

24    that.

25         Q.    And in what respects did you take Mr. Soll up on 
                                                              1772

 1    his offer to provide further news?

 2         A.    I wrote a memorandum through our director to the 

 3    Secretary of Commerce expressing some of my views why I felt 

 4    there was enough concern about the accuracy of adjusted 

 5    counts --  

 6               THE COURT:  I'm sorry, I missed that.

 7         A.    There was concern about the accuracy of adjusted 

 8    counts.

 9         Q.    Could you take a look, please, at Defendant's 

10    Exhibit 76.

11         A.    Okay.

12         Q.    Is that the memorandum that you just referred to?

13         A.    Yes, it is.  

14               MR. SUBAR:  Your Honor, this document is in 

15    evidence in two places.  Again, it is part of the 

16    administrative record and it's also Plaintiff's Exhibit 673 

17    in evidence.  

18    BY MR. SUBAR:

19         Q.    Could you briefly summarize this exhibit?

20         A.    Okay.

21               Well, I had personal concern that we still did 

22    not have enough evidence to show one way or another whether 

23    adjusted counts were more accurate than the census counts, 

24    and so I summarized them into eight points and within those 

25    points I listed various and sundry concerns and sent them to 
                                                              1773

 1    the Secretary, you probably all tired of hearing the 

 2    statement at this time, but prefaced by an issue upon which 

 3    reasonable people can disagree, and I disagreed with my 

 4    colleagues at the Census Bureau and I said hopefully 

 5    reasonably, and I wanted him to understand my views.

 6         Q.    In this memorandum you discuss the concept of 

 7    face validity, among a number of other issues?

 8         A.    Yes, I did.

 9         Q.    Were you here yesterday for Dr. Fienberg's 

10    testimony during which he discussed your discussion of face 

11    validity?

12         A.    Yes, I was.

13         Q.    Can you flip to page 7 of the memorandum, or I 

14    should say page 7 of the attachment to the memorandum.

15               That is your discussion of face validity that 

16    begins at the bottom of that memorandum, correct?

17         A.    Point eight, yes.

18         Q.    That is the eighth point out of a number of 

19    points?

20         A.    Yes.

21         Q.    And are the various points that you address in 

22    this memorandum technical statistical points?

23         A.    Generally, yes.

24         Q.    You mention the state of New Mexico in the course 

25    of your discussion of face validity and, in fact, you use 
                                                              1774

 1    New Mexico as an example.  

 2               The document says a few examples will illustrate 

 3    the point, that is, the point you were making about face 

 4    validity.

 5               What was the point you were making about face 

 6    validity?

 7         A.    Well, I'll answer with regard new Mexico and then 

 8    with regard to that generally.

 9               Certainly, I don't know what the undercount rate 

10    of New Mexico is any better than anybody else, but my 

11    expectation would be that it should be --  should not be so 

12    high.  I say that for the following:

13               Of course it has a very high percentage of 

14    Hispanics, but the Hispanics in New Mexico have been there 

15    for a long time, they are generally second and third 

16    generation Hispanics.  New immigrants to the United States 

17    don't tend to go there so fast, they usually start in Texas 

18    and California.

19               Also New Mexico has no large cities with the 

20    kinds of problems we identify with crime areas and things 

21    like that in our bigger cities, so I would not have expected 

22    New Mexico to have that high an undercount rate.

23               However, as was mentioned in the courtroom 

24    yesterday, I also then concluded with an understatement, 

25    which is talk about this as the only criteria is not 
                                                              1775

 1    reasonable.

 2               I don't think because I think New Mexico --  

 3               THE COURT:  Talk about what? 

 4               THE WITNESS:  My low expectation for New Mexico 

 5    and the fact that the PES came in high.

 6         A.    That's no way to decide whether or not you should 

 7    adjust, that's just my expectation compared to a sample 

 8    estimate.

 9               The only thing I was trying to say is you should 

10    see how many times that happens, and if it happens a fair 

11    number of times, then perhaps you need to do a much more 

12    stringent statistical test, but a face validity argument is 

13    not a statistical test, just to understand what is really 

14    going on with the data.

15               That was all I was trying to say.

16         Q.    Could you flip to the second page of the 

17    document, which is the first page of your eight page 

18    attachment to the cover memo.

19         A.    I have that.

20         Q.    You make a statement at the beginning, I guess --  

21    I'm sorry --  in the third paragraph down you refer to 5 

22    million people having been added to the dual system estimate 

23    from the fourth cell.

24               Do you see that statement?

25         A.    Yes, I do.
                                                              1776

 1         Q.    Is that number 5 million correct?

 2         A.    No, I don't think it is.

 3         Q.    When did you first become aware that it might be 

 4    not be correct?

 5         A.    Some colleagues at the Census Bureau told me that 

 6    my number was wrong about six months ago, something like 

 7    that.

 8         Q.    Had you been informed that the number was wrong 

 9    before you wrote this memorandum, would your advice to the 

10    Secretary contained in the memorandum been the same or 

11    different?

12         A.    I certainly would not have made the error I made, 

13    but the rest of my advice would have remained the same.

14         Q.    Why is that?

15         A.    This is only one point out of the eight major 

16    points and --  

17               THE COURT:  What is the correct number? 

18               THE WITNESS:  The correct number, I think, your 

19    Honor, is somewhere around the order of 1.5 million.

20         Q.    And were you finished with your answer as to --

21         A.    Well, basically, I had several concerns about the 

22    adequacy of the PES numbers.  Still the potential error in 

23    the dual system estimator, I still think is a concern.  This 

24    would not have changed my view of that.

25         Q.    Are you familiar with an entity at the Census 
                                                              1777

 1    Bureau known as CAPE?

 2         A.    Yes, I am.

 3         Q.    What is CAPE?

 4         A.    CAPE is an acronym standing for Committee on the 

 5    Adjustment of Population Estimates.

 6         Q.    Are you a member of CAPE?

 7         A.    Yes, I am.

 8         Q.    What other sorts of people are on CAPE?  Who is 

 9    on CAPE?

10         A.    They tend to be senior statisticians at the 

11    Census Bureau, a number of people who are also on the 

12    undercount steering committee supplemented by some people 

13    who are actually familiar with the program that the Census 

14    Bureau called population estimates.

15         Q.    How long has CAPE been in existence?

16         A.    With that name, since about September of 1991.

17         Q.    Has it been around under other names before that?

18         A.    Yes.  The undercount steering committee continued 

19    to meet even after the July decision, we continued to meet 

20    under the name Undercount Steering Committee and then 

21    changed our name in September of 1991 because our function 

22    changed.

23         Q.    Have you been a member of the committee since its 

24    inception?

25         A.    Yes, I have.
                                                              1778

 1         Q.    And what is the function of CAPE?

 2         A.    The Census Bureau produces things called 

 3    population estimates which are just that.  Between censuses 

 4    we make estimates of state populations and of city 

 5    populations and we turn them over to various and sundry 

 6    agencies.  

 7               About a third of the federal grant programs are 

 8    actually allocated on the basis of these intercensal 

 9    estimates, that's the word used, because they are between 

10    censuses, rather than using the same number for ten years.

11               The purpose of CAPE is to analyze the PES further 

12    than we were able to do last July 1991.  If we were able to 

13    solve some of the problems that existed there, that we would 

14    then try to adjust on these intercensal estimates.

15         Q.    Does the committee anticipate coming to a 

16    conclusion with regard to the adjustment of these 

17    intercensal estimates?

18         A.    Yes, it does.

19         Q.    When is the decision anticipated?

20         A.    About July of this year.

21         Q.    Specifically, what work has CAPE been involved 

22    in?

23         A.    Well, there are many, many research projects they 

24    are looking at, but they tend to fall in about five key 

25    areas.  
                                                              1779

 1               MR. COHEN:  I am going to object to this 

 2    testimony, your Honor, as irrelevant to the issues in this 

 3    case.  We are talking about research that is being conducted 

 4    after July 15 and I don't see the bearing on the issues 

 5    here.  

 6               MR. SUBAR:  Your Honor, that was an issue that 

 7    was addressed in the motion in limine filed before the trial 

 8    of this case.  The court ruled in our favor with regard to 

 9    the motion in limine.  

10               MR. COHEN:  Your Honor, throughout this trial 

11    defendants have been objecting to the introduction of 

12    post-July 15 documents.  As a matter of fact, as I 

13    understand this morning, in response to defendants' motion, 

14    Mr. Zimroth withdrew some of those documents that we did 

15    offer.  

16               What is good for the goose is good for the gander 

17    and I don't think that testimony irrelevant to the period 

18    before July 15 ought to be admitted.  On that basis I 

19    object.  

20               MR. SUBAR:  Your Honor, there is a great deal of 

21    difference between what the goose is trying to do here and 

22    what the gander is trying to do.

23               The plaintiffs are trying to say that on the 

24    basis of information developed after July 15, a decision 

25    made on or before July 15 should be vacated because, they 
                                                              1780

 1    say, what happened after July 15 should have been known 

 2    before July 15.  What they want is for this court to order 

 3    an adjustment.

 4               What we are saying the purpose for which we are 

 5    introducing this evidence is to show that the adjustment 

 6    that was contemplated last July 15 wouldn't make any sense 

 7    at all based on information that we now know.  That is a 

 8    totally different purpose than the purpose for which the 

 9    plaintiffs are trying to introduce post-July 15 evidence.  

10               THE COURT:  Let me pose a hypothetical to you:

11               Supposing that the Secretary was arbitrary on 

12    July 15, but subsequently developments fortuitously proved 

13    he was right, what do I do?  

14               MR. SUBAR:  What you should do under those 

15    circumstances, your Honor, is certainly not order the 

16    adjustment that was contemplated last July 15.  The numbers, 

17    as everyone would agree today based on this post-July 15 

18    evidence, or as everyone should agree today, the July 15 

19    numbers are totally wrong.  

20               MR. COHEN:  Your Honor, our view is that should 

21    you determine that what the Secretary did on July 15, 1991 

22    was arbitrary, then you can take up in due course the 

23    question of relief.

24               But the issues as to what new research the Census 

25    Bureau has developed in order in the post-litigation 
                                                              1781

 1    context, that evidence ought not be before the court on the 

 2    issue of whether the Secretary acted arbitrarily and 

 3    capiciously on July 15.

 4               (Pause) 

 5               THE COURT:  The objection is overruled.  

 6               MR. SUBAR:  Thank you, your Honor.  

 7    BY MR. SUBAR:

 8         Q.    Mr. Bounpane, I believe you were describing some 

 9    of the significant research projects that CAPE has been 

10    involved in?

11         A.    There are various projects.  What I would just 

12    like to do is summarize them as the five key areas or so 

13    that are being examined.

14               The first have to do with the issue of the 

15    smoothing model, the assumptions underneath it, the error 

16    that is associated with it, whether it is robust or not and 

17    things like that.

18               The second has to do with a set of studies to try 

19    to get more information about whether or not this 

20    homogeneity assumption really does work and does it hold, 

21    because the information we had previously was minimal.

22               The third had to do with differences between the 

23    post-enumeration survey results and the demographic analysis 

24    results; what kind of information could we get to see why 

25    that is occurring, which one might be correct.
                                                              1782

 1               The fourth has to do with the total error model.  

 2    Do we have all the error in the total error model, and once 

 3    we do, are are we using that figure properly in estimating 

 4    those target numbers that get used in the loss function 

 5    analysis? 

 6               And finally, the fifth area has to do with 

 7    removing the bias from the PES estimate.

 8               The July 15 PES estimate was an estimate of about 

 9    2.08 percent, but we also estimated at a negative .7 bias on 

10    it, so really if you would like to adjust, one of the best 

11    things to do is to remove that minus .7 bias before you did 

12    it.

13               Unfortunately, no one knows how it distributes by 

14    state or by city, so we are trying to see if we can do 

15    anything about lessening that number, that is, removing it 

16    from certain places where it exists and getting it down 

17    smaller.

18         Q.    Can you take a look, please, at Defendant's 

19    Exhibit 22, which is in the first binder.

20         A.    Okay, I have it.

21         Q.    What is that document?

22         A.    This is a document that was written by Howard 

23    Hogan at the Census Bureau which basically describes an 

24    operation which in this memorandum is called outlier cluster 

25    review which has subsequently been renamed selective cluster 
                                                              1783

 1    review.

 2               Basically, this regards that last of the five 

 3    points I just mentioned.  In order to try to remove the bias 

 4    from the PES estimator, we selected about 100 or so blocks 

 5    that we thought would be major contributors to that and we 

 6    went back and reworked them trying to remove the bias 

 7    wherever we could find it within those 100 blocks.  

 8               MR. COHEN:  Your Honor, I am going to object to 

 9    this line of testimony.  There has been no foundation 

10    established for it yet.  

11               THE COURT:  What sort of foundation do you think 

12    is necessary?  

13               MR. COHEN:  Personal knowledge might be one 

14    basis.

15         Q.    Mr. Bounpane, do you have personal knowledge with 

16    respect to the outlier cluster review?

17         A.    Yes, I do.

18         Q.    Can you tell the court what the basis of that 

19    personal knowledge is?

20         A.    The outlier cluster review or selective cluster 

21    review gets discussed often at the meetings of the CAPE 

22    group and I sit on that group so we talk about it often.  

23               MR. COHEN:  I will also object on the grounds of 

24    hearsay, your Honor.  

25               THE COURT:  Overruled.
                                                              1784

 1               You may continue. 

 2               MR. SUBAR:  At this time, your Honor, we offer 

 3    Defendant's Exhibit 22 into evidence.  

 4               MR. COHEN:  One more objection, your Honor.  

 5    Again, this is a hearsay document and it is my understanding 

 6    from the last paragraph that it is not a final report, and I 

 7    believe that it does not come in within the government 

 8    documents exceptions under those circumstances.  

 9               MR. SUBAR:  Your Honor, it certainly does.  It 

10    false smack dab within the he will realm of Rule 803 (8) (a) 

11    of the Federal Rules of Evidence.  

12               THE COURT:  It can't be 803 --  it could be, I 

13    guess.  

14               MR. SUBAR:  Whatever it is.  The subsection is 

15    (8) (a).  

16               THE COURT:  This is a record, report or 

17    statement, et cetera, setting forth the activities of whom, 

18    the committee on adjusting the population estimates?  

19               MR. SUBAR:  That, in part, and more generally the 

20    Census Bureau.  

21               THE COURT:  Let me just skim this document.

22               (Pause) 

23               THE COURT:  It seems to fit.  

24               MR. COHEN:  Your Honor, may I just be heard for a 

25    moment?  
                                                              1785

 1               THE COURT:  Yes.  

 2               MR. COHEN:  It is a nonfinal report.

 3               I believe under United States Airlines 

 4    against Austin Travel in this circuit, the court has held 

 5    that incomplete interim documents are not admissible under 

 6    803 (8) 

 7               THE COURT:  What case?  

 8               MR. COHEN:  United States Airlines against Austin 

 9    Travel, 867 Fed 837.  

10               THE COURT:  You don't happen to have that there, 

11    do you?  

12               MR. COHEN:  No.  

13               THE COURT:  It doesn't ring a bell.  It doesn't 

14    even sound even remotely correct.

15               Was it a report conducted after a finding of 

16    fact, a hearing of some sort as in the Beech Aircraft case?  

17               MR. COHEN:  I don't know the particular --  

18               THE COURT:  The finality requirement is normally 

19    limited to subsection (b) and (c), not to (a).  He is 

20    offering it under (a).  I suspect that case is a (b) or (c) 

21    case.  

22               MR. COHEN:  In addition, your Honor, I would like 

23    to point out these are all post-litigation documents and as 

24    the court has earlier observed at the very least, some 

25    tighter scrutiny ought to be given to them for the 
                                                              1786

 1    trustworthiness.  

 2               THE COURT:  I will certainly bear that in mind.

 3               I will admit it and give it a tighter scrutiny 

 4    than usual.

 5               Defendant's Exhibit 22 admitted with tight 

 6    scrutiny.

 7               (Laugher)

 8               (Defendant's Exhibit 22 marked for identification 

 9    was received in evidence.) 

10    BY MR. SUBAR:

11         Q.    Has the outlier cluster review or the selective 

12    cluster review, which I think you said are one and the same, 

13    been conducted?

14         A.    Yes, it has.

15         Q.    Were there any amonalies found in the course of 

16    that review?

17         A.    Yes.  While reviewing these 100 or so clusters to 

18    try to remove the bias, we found a computer coding error in 

19    the program that had been used to produce the 

20    post-enumeration survey estimates last July.

21         Q.    What is the nature of that computer coding error?

22         A.    It had to do with the category of erroneous 

23    enumerations.  Basically, it was a set of erroneous 

24    enumerations that had been identified one way clerically, 

25    but somehow through the computer process that code had not 
                                                              1787

 1    been carried through properly.  Some of those made the 

 2    estimated undercount go up and some of those made the 

 3    estimated undercount go down.

 4         Q.    What was the magnitude of that error?

 5         A.    It was about four/tenths of one percent on the 

 6    estimated undercount rate in a negative sense.

 7         Q.    What would that translate into with regard 

 8    to the --

 9         A.    That says that the estimated undercount rate of 

10    last June 1991 of 2.1 percent, had we not made this computer 

11    error, and that's all it was, would have come out as 1.7 

12    percent.

13         Q.    Could you take a look at Defendant's Exhibit 21, 

14    please, and would you tell the court what that is?

15         A.    This is a memorandum from John Thompson as the 

16    Census Bureau to his supervisor, Charles Jones, informing 

17    him that during the outlier cluster review they had 

18    recognized this problem, they were going to go research it 

19    more and they gave an estimate of what they thought the 

20    effect of it would be.

21         Q.    Exhibit 21.

22         A.    Oh, I'm sorry, I was looking at the wrong number.  

23    I was looking at 20.

24         Q.    All right.  What is 21?

25         A.    Exhibit 21 is a memorandum from the same John 
                                                              1788

 1    Thompson of the Census Bureau to his supervisor Charles 

 2    Jones.

 3               After correcting the computer error and rerunning 

 4    the entire post-enumeration survey dual system estimate 

 5    process that I just described, this is the set of the 

 6    results you got and these are the numbers we would have had 

 7    available in early 1991 during our deliberations had we not 

 8    made this mistake.

 9         Q.    Do you know how many people were accounted for by 

10    this error that you mentioned?

11         A.    No, I think I only know it in percentage terms.

12         Q.    Are there some attachments attached to 

13    Defendant's Exhibit 21?

14         A.    Yes, there are.

15         Q.    What are those attachments?

16         A.    They show this result not only at the U.S. level, 

17    but they show it by cities and by counties.

18               You can turn to attachment --  it is table 8, the 

19    last page of the attachments.  It's a summary at the 

20    national level by race and sex.

21               The column that is labeled PES and under that the 

22    letters UCRT, that is undercount, those are the numbers we 

23    had available to us in early 1991 when we were doing our 

24    deliberations.

25         Q.    When you say early 1991, what do you mean?
                                                              1789

 1         A.    May-June for the undercount steering committee.  

 2    I'm sorry.

 3               The ones in the column January 1992 also under 

 4    the colume UCRT for the same categories, the rerunning 

 5    correcting for the computer coding error and the estimated 

 6    undercount rate at this time.

 7         Q.    Does this chart show anything with regard to the 

 8    differential undercount?

 9         A.    Yes.  I think you could get that by looking at 

10    the black undercount in June 1991 and the block undercount 

11    in January 1992 and the nonblock for those two same periods 

12    and then subtracting it.

13         Q.    Do you know whether the differential undercount 

14    goes up or down as a result of the fixing of this mistake?

15         A.    I think the differential undercount goes down.

16         Q.    That is, if the mistake is corrected, the 

17    differential undercount goes down?

18         A.    That's correct.

19         Q.    Is this document a draft document or a final 

20    document?

21         A.    I think it's a final document.  

22               MR. SUBAR:  Your Honor, we offer Defendant's 

23    Exhibit 21 into evidence.  

24               MR. COHEN:  Object again on relevance grounds 

25    again, your Honor.  
                                                              1790

 1               THE COURT:  Overrule.  I will admit it as 

 2    Defendant's Exhibit 21.

 3               (Defendant's Exhibit 21 marked for identification 

 4    was received in evidence.) 

 5    BY MR. SUBAR:

 6         Q.    Other than results from fixing the computer error 

 7    that you have been mentioning, are there results more 

 8    generally from the selective cluster review?

 9         A.    Yes.

10               Remember the purpose of this is to try to take 

11    some blocks that we thought had major impact on the 

12    estimated undercount rate, go back and redo them and if you 

13    find mistakes in them correct them.  

14               That is some way of getting the error in the PES 

15    estimate out of it, not complete, of course, because you are 

16    only doing 100 clusters, but it is 100 clusters that might 

17    have a large effect on the estimate.

18               And so we have gone back and done those clerical 

19    operations and fixed mistakes where we found them.

20         Q.    Could you please take a look at Defendant's 

21    Exhibit 31.

22         A.    Okay, I have 31.

23         Q.    What is that document?

24         A.    These are the minutes from the CAPE committee 

25    meeting that was held on April 6 and some attachments to 
                                                              1791

 1    those minutes as well.

 2         Q.    What are the attachments? 

 3               You don't necessarily have to tell us --

 4         A.    I will tell you attachment one.

 5               Attachment one combines both the correction of 

 6    the computer coding error and the correction for the hundred 

 7    blocks --  it's actually 104 blocks when we did this review 

 8    --  and reruns the dual system estimation process 

 9    incorporating both things.

10         Q.    What does it show with respect to the results of 

11    that?

12         A.    If you will turn to the first page of attachment 

13    two there, that shows the estimated undercount for each of 

14    the states and for the U.S. in total.  

15               The U.S. in total on the last line named national 

16    totals, under the colume named, titled U for undercount, you 

17    read down on the bottom, it says now our estimated 

18    undercount is 1.58 percent correcting for the computer 

19    coding error and whatever we found in reviewing those 104 

20    blocks.  

21               MR. SUBAR:  Your Honor, we offer Defendant's 

22    Exhibit 31 into evidence.  

23               MR. COHEN:  Your Honor, we were not given these 

24    documents until I believe after Mr. Bounpane's deposition 

25    actually commenced.  We had outstanding discovery requests 
                                                              1792

 1    that --  well, April 6, of course, came after his 

 2    deposition.

 3               We continue our objection on relevance grounds.  

 4               THE COURT:  The objection is, again, overruled.

 5               Exhibit 31 is admitted.

 6               (Defendant's Exhibit 31 marked for identification 

 7    was received in evidence.) 

 8    BY MR. SUBAR:

 9         Q.    Mr. Bounpane, do you think that the net national 

10    undercount is 1.58 percent?

11         A.    No, I'm not sure about that.

12         Q.    Why is that?

13         A.    Well, there are still several sources of error 

14    attached to this number.  We have not removed all of the 

15    bias from it, and there are many points in the whole PES 

16    dual system estimation process that are based on assumptions 

17    and judgment.  I'm still not sure whether those hold and 

18    what the effect of them are, so I can't say that this is a 

19    fully accurate estimate of the undercount in the last 

20    census.

21         Q.    Thank you.  

22               MR. SUBAR:  I have no further questions.  
                 THE COURT:  All right.
23               We will recess for the day.  We will resume 
      tomorrow at 9:30 with cross-examination.
24               (Adjournment taken to Friday, May 22, 1992 at 
      9:30 o'clock a.m.) 
25    
