      lta                                                     186    

 1    UNITED STATES DISTRICT COURT
      EASTERN DISTRICT OF NEW YORK
 2    ------------------------------x
      THE CITY OF NEW YORK, et al.,
 3    
                         Plaintiffs,          89 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.                          May 12, 1992
                                              10:00 a.m.
11    MOSBACHER, et al.,
      
12                       Defendants.
      ------------------------------x
13    
      Before:
14    
                 HON. JOSEPH M. McLAUGHLIN,
15    
                                              Circuit Judge
16    
      
17

18

19

20

21

22

23

24

25


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census                                       187    

 1               (Trial continued)

 2               (In open court)

 3               THE COURT:  Good morning, ladies and gentlemen.
                                                                

 4               Are there any preliminary or housekeeping 

 5    matters?  

 6               MR. RIFKIND:  I have one, your Honor.  I'm 

 7    advised that Mr. Leon Segen, who represents Hudson County, 

 8    New Jersey, would like to make a brief statement to the 

 9    Court.

10               THE COURT:  Sure.  

11               Good morning.  Would you spell your last name for 

12    the record.

13               MR. SEGEN:  Certainly.  S E G E N.

14               THE COURT:  And the first name was Leon?  

15               MR. SEGEN:  Leon, with the firm of Scarinci 

16    Hollenbeck.

17               Your Honor, we represent Hudson County, and I 

18    just want to make a record to make certain that the rights 

19    of Hudson County are preserved in this court.

20               As I understand the position that is being 

21    pursued by plaintiffs, it is solely in regard to an 

22    adjustment of the entire census based on the arbitrary and 

23    capricious standard.  Hudson County filed a complaint and 

24    intervention which included other claims.  I want to put 

25    defendants on notice that these claims are still being 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census                                       188    

 1    pursued by Hudson County, and that we do intend to argue 

 2    them at the appropriate time.

 3               Your Honor, these claims involve two things:

 4               No. 1.  We certainly join plaintiffs in their 

 5    submission that the entire census head count needs to be 

 6    adjusted in accordance with the PES, and that one of those 

 7    prongs regarding the arbitrary and capriciousness of the 

 8    Commerce Department's decision has to do with the bias 

 9    involved with the differential undercount of minorities.

10               However, Hudson County also submits that even if 

11    the Court should decide that the entire census does not have 

12    to be adjusted, that, nevertheless, because of the bias 

13    involved in regard to the differential undercount, because 

14    of the historical record of discrimination and undercount in 

15    regard to minorities, that the Court could, nevertheless, 

16    find that as to the minorities, that the PES has to be used.

17               The second claim has to do with the Voting Rights 

18    Act.  It is Hudson County's contention that the differential 

19    undercount results in a loss of minority districts under the 

20    Voting Rights Act, and that, therefore, the PES must be 

21    used; an adjustment must be used in regard to the census 

22    count of 1990.

23               THE COURT:  Will we be seeing you or hearing from 

24    you between now and the end of the suit?  

25               MR. SEGEN:  I have attempted to coordinate with 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census                                       189    

 1    the committee representing plaintiffs.  And it is my 

 2    understanding, based on conversations with Assistant 

 3    Attorney General Sanford Cohen, and also with the 

 4    Corporation Counsel, David Golden, that all of the evidence 

 5    that Hudson County will need in order to argue its case will 

 6    be presented by them.

 7               I do intend to review the evidence that is going 

 8    to be submitted, but I certainly defer to the decision of 

 9    the committee that they will present all the evidence, and 

10    that my presence will not be necessary for that purpose.

11               THE COURT:  Well, so there's no -- 

12               MR. SEGEN:  If that's acceptable to your Honor.

13               THE COURT:  Yes.  So that you don't slip through 

14    the cracks, may I ask that you keep an eye on the trial and, 

15    at the end, when I ask for submissions from all the parties 

16    on proposed findings, would you submit your own, if you have 

17    any separate ones?  

18               MR. SEGEN:  Yes, your Honor.  Thank you.

19               THE COURT:  Mr. Millet.  

20               MR. MILLET:  Your Honor, if I could just respond 

21    to that briefly.  

22               All I can say is, as I believe the Court is well 

23    aware, all the plaintiffs who intervened in this case were 

24    intervened and taking the case as they found it.  

25               As to the voting rights claim, your Honor, we can 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                190    

 1    certainly brief that, but I believe that the Voting Rights 

 2    Act does not impose obligations on the Federal Government.  

 3    But we can certainly address that in post-trial briefs.

 4               THE COURT:  Right.  And you can work out my way 

 5    of doing it at the end?  

 6               MR. MILLET:  Yes, your Honor.  

 7               MR. ZIMROTH:  May we proceed, your Honor?

 8               THE COURT:  You may indeed.   

 9               I'll ask shorty to resume the stand.

10               THE WITNESS:  Okay.  I'll put the chair down a 

11    little bit.

12    EUGENE P. ERICKSEN, 
                          

13         having been previously sworn, resumed the

14         stand and continued to testify as follows:

15               THE COURT:  Mr. Zimroth.  

16    DIRECT EXAMINATION CONTINUED

17    BY MR. ZIMROTH:

18         Q.    Good morning, Mr. Ericksen.

19         A.    Good morning.

20         Q.    I believe where we left off yesterday was with 

21    the fourth stage of the census process which is labeled on 

22    that chart "Post-Census Day Coverage Improvement Programs."

23               Would you explain to Judge McLaughlin what those 

24    are?

25         A.    Sure.  At the end of nonresponse follow-up, it's 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                191    

 1    clear that there are millions of people who have not yet 

 2    been counted, and the Bureau has instituted a set of 

 3    procedures called Coverage Improvement Programs, and it's 

 4    been developing since about 1960.

 5               In 1990, the Coverage Improvement Programs 

 6    generally started after a nonresponse follow-up was 

 7    completed.  For example, they went back to every housing 

 8    unit that had been enumerated as vacant to make sure that it 

 9    really was vacant.  And if it turned out that it had been 

10    occupied on April 1st, then they changed to it occupied.  

11    That was called the vacant delete check.

12               A second example of a coverage improvement 

13    program was called the parolee probationer's check.  And in 

14    that procedure, the Bureau obtained names of people who are 

15    on probation or on parole, got their names and addresses 

16    from the relevant parole or probation officer, and checked 

17    whether they had been counted at those addresses; and if 

18    they had not, they added them.  

19               MR. ZIMROTH:  Excuse me, your Honor.  I am 

20    unfortunately having a little trouble hearing.  Is that 

21    microphone on?

22               THE COURT:  It's on.  Maybe it's not loud enough.  

23    I don't know how to control anything in this courtroom.

24               (Pause)

25    BY MR. ZIMROTH:


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                192    

 1         Q.    Had you finished your explanation of parolee- 

 2    probationer?

 3         A.    Yes, I did.

 4         Q.    Okay.  Have there been evaluations of Coverage 

 5    Improvement Programs in prior censuses?

 6         A.    Yes.  The Census Bureau has done evaluations of 

 7    Coverage Improvement Programs.

 8         Q.    Keep your voice up.  I didn't hear you.

 9         A.    Yes.  The Census Bureau has done fairly extensive 

10    evaluations of Coverage Improvement Programs, especially 

11    since 1970.

12         Q.    And what do those evaluations show?

13         A.    Well, in general, Coverage Improvement Programs 

14    don't solve the problem.  After the Coverage Improvement 

15    Programs are completed, there are many people left 

16    uncounted, and there's a substantial differential under- 

17    count.

18         Q.    Will you please look at 474, please, PX 474.  

19    Which is in Volume 3.

20               What is that?  Just tell the Judge what that 

21    document is.

22         A.    Yes.  This is a summary report that describes 

23    certainly most, possibly all of the Coverage Improvement 

24    Programs of the 1970 census.  And if you look on the first 

25    page, under the Table of Contents, it lists the Coverage 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                193    

 1    Improvement Programs that were part of that census.

 2         Q.    And is this one of the evaluations that you were 

 3    referring to?

 4         A.    I'm sorry?

 5         Q.    Is this one of the evaluations to which you were 

 6    referring earlier?

 7         A.    Yes.  This is the Bureau general evaluation, I 

 8    believe, of all the Coverage Improvement Programs of the 

 9    1970 census.

10         Q.    And the conclusion was or the evaluation was?

11         A.    Well, the conclusion was that millions of people 

12    have been left uncounted, and there's also a general 

13    conclusion in there that the problem of differential 

14    undercount in particular and the overall undercount in 

15    general may be proving impossible.

16         Q.    And would you look -- 

17               MR. ZIMROTH:  I offer that at this time, your 

18    Honor, 474.  

19               MR. SITCOV:  No objection.

20               THE COURT:  Received in evidence.  474.

21               (Exhibit PX 474 for identification was received 

22    in evidence) 

23    BY MR. ZIMROTH:

24         Q.    And would you look at 490, please.

25               (Pause)


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                194    

 1         A.    I have it.

 2         Q.    Can you tell the Judge what 490 is.

 3         A.    Yes, 490 is the comparable report for the 1980 

 4    census.  In 1980, the Census Bureau spent a lot more money 

 5    on coverage improvement, and perhaps I could identify a 

 6    correspondingly larger report.

 7               And the conclusions -- the general conclusion was 

 8    that although they had spent a lot of money on coverage 

 9    improvement in 1980, that these Coverage Improvement 

10    Programs occurring as late as they do were hard to mount.  

11    They were expensive.  There were a lot of erroneous 

12    enumerations, a lot of errors in the programs, and there was 

13    a substantial undercount that was left after these were 

14    completed.  

15               MR. ZIMROTH:  All right.  Your Honor, I offer 

16    490.  

17               MR. SITCOV:  No objection.

18               THE COURT:  490 is received.

19               (Exhibit PX 490 for identification was received 

20    in evidence)

21    BY MR. ZIMROTH:

22         Q.    In your opinion, Professor Ericksen, were the 

23    1990 Coverage Improvement Programs successful in eliminating 

24    the differential undercount?

25         A.    No.  While I believe that the Coverage 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                195    

 1    Improvement Programs in 1990 suffered a similar fate to 

 2    those in 1980, as expensive as they were, they are difficult 

 3    to mount.  The rate of error was substantial and, most 

 4    importantly, left a large differential undercount in the 

 5    1990 Census with 5 million net undercount.

 6         Q.    On what do you base your conclusion that the 1990 

 7    Census there was a differential undercount?

 8         A.    Well, there's a differential undercount that's 

 9    been demonstrated not only by the post-enumeration survey 

10    but also by demographic analysis.

11         Q.    Would you tell the Judge what demographic 

12    analysis is?

13         A.    Yes.  Demographic analysis is a procedure that is 

14    relied upon by the Census Bureau for many, many purposes.  

15    Basically, what the demographic analysis involves is the 

16    Census Bureau constructs an alternative estimate of the 

17    population that's based on administrative records.

18               In the United States, demographers feel that we 

19    count births quite accurately and have done so for many 

20    years.  The same is true of deaths.  And the Census Bureau 

21    constructs the population basis on birth records, death 

22    records, records of international migration, and constructs 

23    a national population estimate for blacks, for non-blacks, 

24    and these two groups are separated by gender and by age.  So 

25    we have age, race, sex, group estimates for the nation.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                196    

 1         Q.    When did the Census Bureau first begin to use 

 2    demographic analysis?

 3         A.    Well, demographic analysis -- as I understand it, 

 4    the first article describing demographic analysis that I 

 5    know of was published in the Journal of American Statistical 

 6    Association in 1955 by Andrew Coe.  I believe he developed 

 7    the modeling in cooperation with the Census Bureau.  To the 

 8    best of my knowledge, the Bureau has been using demographic 

 9    analysis at least since some time in the 1960s.

10         Q.    And how far back has the Census Bureau 

11    constructed the national population based on demographic 

12    analysis?

13         A.    Well, the Census Bureau has now developed 

14    estimates of the demographic estimates of the national 

15    population for each of the years 1940, 1950, 1960, 1970, 

16    1980 and 1990.

17         Q.    Now, did the Census Bureau, in June of 1991, 

18    issue figures concerning the demographic estimate for the 

19    1990 Decennial Census?

20         A.    Yes, it did.  There was a press release issued 

21    June 13th of 1991 which provided those figures.

22         Q.    Did that press release also give the figures for 

23    the previous censuses?

24         A.    Yes.  It gave the figures so you can compare the 

25    undercounts in 1990 to the undercounts for the 1940 through 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                197    

 1    1980.

 2         Q.    Will you look at 476 A, please.

 3         Q.    Is that the press release that you were referring 

 4    to?

 5         A.    Yes, it is.

 6         Q.    Now, will you look at this chart which has been 

 7    marked as 476.  

 8               MR. ZIMROTH:  Perhaps, your Honor, if Professor 

 9    Ericksen could come up to the chart and explain what the 

10    chart is.

11               THE COURT:  Sure.  Yes.

12               (Pause)

13         A.    The numbers that this chart is based upon are 

14    given in Table 3 of that press release, and what we have 

15    here is the estimated undercounts by the demographic 

16    analysis for the black population and for the non-black 

17    population.  That's everyone -- the non-black population is 

18    everyone who is not classified as black.  For 1940 through 

19    1990.

20               In 1940 the black undercount was 8.4 percent, and 

21    the non-black undercount was 5.0 percent.

22               As we move from 1940 through 1980, we can see 

23    that there's been a reduction in the net undercount for 

24    blacks.  The figure was about 7.5 percent in 1950, about 6.6 

25    percent in 1960, about the same in 1970.  It went down to 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                198    

 1    about 4 1/2 percent in 1980.  And then in 1990 it went back 

 2    up to 5.7 percent.

 3               Looking at non-blacks, we can see that there was 

 4    a parallel reduction in the non-black undercount from 5.0 

 5    percent in 1940 down to 1980, when it was actually less than 

 6    1 percent.  But in 1990, it went back up to 1.3 percent.  

 7         Q.    Now, let me just move this way a little bit, 

 8    Professor Ericksen.

 9               And I ask you to look at the chart that has been 

10    marked 477.  And can you explain that chart and how it 

11    relates to 476?

12         A.    Certainly.

13               Here we have estimates of the differential 

14    undercount for the same years, 1940 through 1990.

15               If you go back to this chart, recall that the 

16    black undercount was 8.4 percent, the non-black undercount 

17    was 5.0 percent.  You subtract the one from the other.  The 

18    differential is 3.4 percentage points.  So we have the 

19    differential of 3.4 percentage points.

20               As you can see, that the differential undercount 

21    has been more or less flat in between 1940 and 1990.  It 

22    actually increased slightly through 1970, went down a little 

23    bit in 1980.  It went back up again in 1990.

24               So the differential undercount in 1990 is 

25    actually slightly greater than it was for any of the other 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205121census      Ericksen - direct                199    

 1    years between 1940 and 1980.  But the point really is that 

 2    although the net undercount has been reduced, the 

 3    differential undercount has gone up slightly during the 

 4    period from 1940 through 1990.

 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    


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                200    

 1         Q.    Professor Ericksen, I want you just to look at 

 2    this chart that we have -- or this listing for the major 

 3    components of the census process, 454.

 4               In your opinion, -- oh, I am sorry -- I neglected 
                                                                   

 5    to move the introduction of 476 A and the two charts, 476 

 6    and 477.  

 7               MR. SITCOV:  No objection.  

 8               THE COURT:  All three are admitted.

 9               (Plaintiff Exhibits 476 A, 476 and 477 for 

10    identification were received in evidence) 

11         Q.    Professor Ericksen, in your opinion, as long as 

12    the census is conducted in essence using the procedures like 

13    the ones that you have described earlier that were used in 

14    the 1990 census, in your opinion, can there be a substantial 

15    amealioration of the differential undercount?

16         A.    No.  While I believe that the census does as good 

17    a job as it reasonably could do in taking the census, I 

18    believe that the differential undercount is inherent in the 

19    methods that are used, and as long as we take the census in 

20    the way that we do it, the way we do it is a good way of 

21    taking the census, but the differential undercount will 

22    nonetheless result.  

23               I don't see it being reduced if we continue to 

24    take the census in the way that we have taken it in the 

25    recent past.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                201    

 1         Q.    In your opinion, has the census bureau found a 

 2    way to ameliorate the differential undercount?

 3         A.    Yes, it has.

 4         Q.    What is that way?

 5         A.    That way is to adjust the census for the 

 6    differential undercount, using data collected by the 

 7    post-enumeration survey.

 8         Q.    Before we get into that, can you explain to the 

 9    judge why these demographic estimates could not be used -- I 

10    should ask you, in your opinion, can they be used to adjust 

11    the census?

12         A.    No, I do not believe that they should be used to 

13    adjust the census.

14         Q.    And why not?

15         A.    Because they only provide national estimates, and 

16    we need estimates for local areas.  Demographic analysis 

17    does not provide estimates for local areas.

18         Q.    What about the accuracy of various components in 

19    the demographic analysis?

20         A.    Well, that is the second problem with demographic 

21    analysis.  The Census Bureau, when it constructed the 

22    estimate of the demographic population for 1990, realized 

23    that there was, like any other statistical estimate, 

24    uncertainty in the estimate, and in my opinion the 

25    uncertainty inherent in the demographic estimates for the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                202    

 1    subgroups defined by age, sex and race is too great to be 

 2    used.

 3         Q.    Now, so what is the way that you think that the 

 4    census could be improved in order to ameliorate the 

 5    differential undercount again?

 6         A.    I believe it should data collected by the 

 7    post-enumeration survey.

 8         Q.    What is the post-enumeration survey?

 9         A.    The post-enumeration survey is a large scale 

10    survey mounted by the Census Bureau to estimate the 

11    undercount for the nation, for states, cities, counties and 

12    other local areas.

13         Q.    What is the post-enumeration survey designed to 

14    measure?

15         A.    The post-enumeration survey is designed to 

16    measure the proportion of people who are not counted in the 

17    census and the proportion of people who were counted in the 

18    census correctly.

19         Q.    Now, you used the phrase "survey," and 

20    "post-enumeration survey," you used that phrase as well. 

21               Could you explain to the judge what you mean by 

22    "survey."

23         A.    Yeah.  Perhaps a more apt phrase would be to 

24    describe it as a sample survey.  When you take a sample 

25    survey -- if I can just give an example, let's assume that 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                203    

 1    the people in the automobile industry wanted to know how 

 2    many American families bought a car last month.  They might 

 3    take a sample of a thousand families, they would go out and 

 4    interview someone from each of the one thousand families and 

 5    they might learn that 80 of them had bought a car in the 

 6    last month.  

 7               Then they would take the number 80, divide it by 

 8    a thousand, and get an estimate of 8 percent of American 

 9    families had bought a car in the last month.  It would 

10    realize, of course, that this estimate was based on a 

11    sample, and because it is based on a sample, they know that 

12    the figure 8 percent was not exactly correct for the entire 

13    population.  

14               So they would estimate a quantity known as a 

15    sampling error, and for this particular example, they would 

16    say that 8 percent plus or minus 2 percent of American 

17    families had bought a car in the last month.

18         Q.    And in that example the characteristic being 

19    sampled for is what?

20         A.    Whether or not the family had bought a car in the 

21    last month.

22         Q.    What is the characteristic being sampled for in 

23    the post-enumeration survey?

24         A.    Whether or not a person was counted.

25         Q.    I am sorry.  I want to go back to the car example 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                204    

 1    for a minute.

 2               In that example, I think you told us that you 

 3    were doing the sample for a nation as a whole?

 4         A.    That's correct.

 5         Q.    Is it possible also to do surveys for subgroups?

 6         A.    Oh, yes.  In fact, it would be quite likely that 

 7    the people in the automobile industry would be much more 

 8    interested in knowing what parts of the American population 

 9    were more interested -- more likely to have bought cars than 

10    less likely, so they might get estimates by north and south, 

11    east and west, by cities and suburbs, for black families, 

12    for white families, for families with children, for career 

13    couples without children, for single people.  

14               They would want to know in what sectors of 

15    American society many people are buying cars and what 

16    sectors of American society few people were buying cars 

17    because that might be the sector of American society where 

18    they would want to change their advertising and selling 

19    strategies.

20         Q.    In your opinion, would that be a more useful kind 

21    of survey?

22         A.    I think it would be more useful to the automobile 

23    industry, yes, I do, because they would take action based on 

24    the results of the survey.

25         Q.    What is that process that you've just described 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                205    

 1    been called?

 2         A.    It's the process of separating the sample into 

 3    subgroups is known as stratification.

 4         Q.    Once again, the characteristic being a sample for 

 5    in the post-enumeration survey is what?

 6         A.    The proportion of people who were counted in the 

 7    census.

 8         Q.    Did the Census Bureau only in 1990 compute only a 

 9    national estimate?

10         A.    No.  The Census Bureau computed estimates for a 

11    substantial number of subgroups.  It refers to the subgroups 

12    as post-strata, P O S T, S T R A T A, and it computed 

13    estimates for 1,392 of these post-strata.

14         Q.    Could you tell Judge McLaughlin just roughly how 

15    the bureau stratified?

16         A.    Sure.  Now, the way that any sampling 

17    statistician would stratify a sample would be to try to 

18    identify those characteristics which could be measured and 

19    which are related to the quantity that you're trying to 

20    estimate, so in this case the Census Bureau would want to 

21    post-stratify by those characteristics which were related to 

22    the undercount.

23               So the first thing the Bureau did was to create 

24    strata based on what's called census divisions; in other 

25    words, groups of states that are next to each other.  Within 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                206    

 1    the census division, the Census Bureau stratified further on 

 2    whether the area was a central city, a suburb or outside 

 3    metropolitan areas.

 4               It also stratified on race, whether or not 

 5    someone was of Hispanic origin.

 6         Q.    And "race" meaning black or white?

 7         A.    In general, race meant black or white.  In areas 

 8    where the Census Bureau considered the Asian population to 

 9    be concentrated, it stratified for them.  There was also a 

10    separate stratum made for Native Americans living on 

11    reservations.

12         Q.    You were up to Hispanic origin?

13         A.    And Hispanic origin.  Within central cities, the 

14    Census Bureau also stratified on whether the person rented 

15    the house or owned the house because their research 

16    following the 1980 census had found that to be a variable 

17    related to the probability of being missed in the count of 

18    the census.

19         Q.    The end of the stratification, how many groups 

20    were they after these strata sort of intersected?

21         A.    1,392.

22         Q.    Is it possible for a person to be in more than 

23    one stratum?

24         A.    No.

25         Q.    How is that?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                207    

 1         A.    Because every person is a member of a certain 

 2    age, race and sex group.  

 3               To give an example, they only lived in one place, 

 4    they either lived in renting housing or owned or occupied 

 5    housing.  It is an exhaustive status of groups.

 6         Q.    Can you give Judge McLaughlin just some examples 

 7    of what a particular post-stratum might be?

 8         A.    Yes.  One particular post-stratum could have been 

 9    20 to 29 year old males living in rental housing in New York 

10    City.

11               A second post-stratum could have been white 

12    females age 30 to 44 living in the suburbs of New York City.

13         Q.    Is it the case, Professor Ericksen, that every 

14    single person living within one of those -- defined by one 

15    of those post-stratum has exactly the same likelihood of 

16    being counted in the census?

17         A.    No, they don't have exactly the same likelihood 

18    of being counted.  I think the point is that people in one 

19    post-stratum often have a very different likelihood of being 

20    counted than a person in a different post-stratum.

21               For example, black males 20 to 29 years of age 

22    living in rented housing in New York City have a 

23    substantially lower chance of being counted than white 

24    females age 30 to 44 living in suburbs outside New York 

25    City.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                208    

 1               Within a post-stratum there will be variations, 

 2    but those variations will be small compared to the 

 3    variations between the different post-strata.

 4         Q.    Now, how did the Census Bureau actually conduct 

 5    this post-enumeration survey?

 6         A.    Well, the Census Bureau conducted the 

 7    post-enumeration survey in stages.  The first stage was to 

 8    select the sample.  There are millions of blocks in the 

 9    United States.  They made a list of the 5,000 -- I think it 

10    is 5 million blocks in the United States.  They made a list 

11    of these blocks, stratified them and selected a sample of 

12    5,000 of them.

13         Q.    How many households were on those 5,000 blocks?

14         A.    In the blocks that were selected into the sample, 

15    there were approximately 170,000 households.

16         Q.    And how many people?

17         A.    Approximately 400,000 people lived in the 170,000 

18    households.

19         Q.    In your experience, is that a large survey?

20         A.    It's a large survey.  

21               THE COURT:  What's a block? 

22               THE WITNESS:  Well, a block is usually exactly 

23    what you would think of it.  

24               THE COURT:  Just one straight street? 

25               THE WITNESS:  That's right.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                209    

 1               THE COURT:  You turn the corner and that's 

 2    another block? 

 3               THE WITNESS:  Right.  This is 1st Street and this 

 4    is 2nd Street and this is Elm Street and that is Maple 

 5    Street.  That's a block.  

 6               THE COURT:  The square is the block? 

 7               THE WITNESS:  Yes.  

 8               THE COURT:  Oh. 

 9               THE WITNESS:  Now, there are some cases where a 

10    block may have enormous population on it, like 2,000 people 

11    living in apartments.  In that case the Census Bureau would 

12    actually subdivide the block and create for sampling 

13    purposes two blocks within the actual physical block because 

14    that would be too large an area to constitute one local 

15    sampling area.

16               Out in the country where you don't necessarily 

17    have those four streets, they might use a river or some 

18    other identifiable boundary to define the block.

19         Q.    Okay, so now they've picked these 5,000 blocks. 

20               What's next?

21         A.    Okay.  After these 5,000 blocks were selected, 

22    they then located each of the 5,000 blocks on the map and 

23    drew the sample blocks on a map and handed the map to Census 

24    Bureau field workers, who then went out to the blocks and 

25    walked around them and listed all the residential addresses 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                210    

 1    that they could find on those blocks.  That's where the 

 2    170,000 households were identified.

 3         Q.    Was the identity of these blocks kept 

 4    confidential?

 5         A.    Yes.  Because the post-enumeration survey was 

 6    being used to estimate the undercount, it was necessary for 

 7    the Census Bureau to conduct the post-enumeration survey in 

 8    a way that it couldn't affect the way the census was being 

 9    done.  

10               So the post-enumeration survey was conducted 

11    totally separately and independent from the conduct of the 

12    census.  To my understanding, the post-enumeration survey 

13    operations took place in different buildings than the census 

14    operations.  

15               THE COURT:  Before you leave that, I'm intrigued. 

16               For the purposes to see who was missed the first 

17    time, why do you go to different buildings? 

18               THE WITNESS:  No.  What I meant was that the 

19    census operations would generally take place in the district 

20    offices, and because they wanted to make sure that the 

21    people on the post-enumeration survey didn't know what was 

22    going on with the census and the people working on the 

23    census didn't know what was going on with the 

24    post-enumeration survey, the post-enumeration survey was 

25    totally separate.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                211    

 1               What I meant by "different buildings," the 

 2    post-enumeration survey office would be in a totally 

 3    different location than the census office.  

 4               THE COURT:  Oh, oh, I misunderstood you. 

 5               THE WITNESS:  I am glad you asked.

 6         Q.    Now I've lost my train.  Where are we up to in 

 7    this process?

 8         A.    Well, I believe that I just explained to you that 

 9    the census field workers went out to the 5,000 blocks and 

10    listed the 170,000 housing units.

11         Q.    Now, earlier you told us that the Bureau 

12    post-stratified into these 1,392 post-stratum?

13         A.    That's right.

14         Q.    Why, then, are we now interested in collecting 

15    data on the basis of blocks?

16         A.    Because most blocks include people who were put 

17    into many different post-stratum.  It is much more efficient 

18    both from a statistical point of view and from a cost point 

19    of view to go to one location and find people in different 

20    post-stratum.

21         Q.    After these blocks are put into the computer, 

22    what's next?

23         A.    Okay, after these blocks are put into the 

24    computer -- what you mean by that, I think --

25         Q.    I mean housing units?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                212    

 1         A.    -- the housing units are put into the computer 

 2    and the addresses are put on labels and put onto sheet, 

 3    cover sheets, and given to PES interviewers.

 4               The PES interviewers starting in July of 1990 

 5    went back to these addresses and interviewed the people that 

 6    they found living there and obtained information, such as 

 7    their age, their race, their sex, whether or not they were 

 8    of Hispanic origin, and found out where they were living on 

 9    April 1, 1990.

10         Q.    Then what happens with -- what is done with the 

11    information collected on these questionnaires?

12         A.    Well, the information that is collected on the 

13    questionnaires is put into the data base so there's a 

14    computer record of all the information collected on the PES 

15    questionnaire.

16               The idea is that this information is going to be 

17    matched against the census forms so it can be determined 

18    whether or not each person in the PES sample was counted or 

19    missed.

20         Q.    Could you explain to the judge how this matching 

21    process works, taking it in various stages.

22         A.    Sure.  I think the best way to think of it is 

23    that you've got two piles of forms for each block, and here 

24    you have the PES forms and here you have the census forms.

25         Q.    That's just a metaphor.  In fact, all this 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                213    

 1    information is in the computer?

 2         A.    That's right.  You also have all the census forms 

 3    for the surrounding blocks.  Now, the basic idea is for 

 4    every address, you match the population listed on the PES 

 5    form against the population listed on the census form, and 

 6    those people who match, then obviously they would be 

 7    considered to have been counted in the census.

 8               Now, to make sure that the -- if the people don't 

 9    match, that that is not due to some small error in 

10    addresses, the census forms for all the rest of that block 

11    and for the surrounding block are also searched, and if 

12    someone is found in one of those places, that person would 

13    also be considered to have been matched to the census.

14         Q.    You're now referring to the computer matching?

15         A.    Well, the matching actually takes place on a 

16    series of steps.  The first step is called computer 

17    matching, and the Census Bureau spent most of the decade of 

18    the 1980's developing a procedure called computer matching, 

19    which accounted for, as I recall, about 80 percent of the 

20    PES sample was matched by computer, and so most of the 

21    sample, it was pretty obvious they were counted in the 

22    census and the computer can make that determination.  Then 

23    the Census Bureau was able to get on with the more difficult 

24    cases.

25         Q.    What happened to the people who were found by the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                214    

 1    PES interviewers were not matched by the computer to a 

 2    census form?

 3         A.    Well, cases which were not matched by the 

 4    computer or where the Census Bureau was not sure, they were 

 5    called possible matches, were then sent to the next stage of 

 6    matching, where Census Bureau employees matching -- trained 

 7    matchers would try to compare the two forms to see if they 

 8    can match them.

 9         Q.    Can you give the judge an example of a case that 

10    might be declared not to be a match or a questionable match 

11    by the computer but might be declared a match by one of 

12    these Census Bureau employees.

13         A.    Yes.  If there was, let's say, a woman, we'll 

14    call her Beatrice Jones, and on her census form it was 

15    indicated that she was born in 1927, and living with her was 

16    her teenage daughter Susan Jones, who was born in, let's 

17    say, 1973, then on the PES the same information could have 

18    been collected by the PES interviewer, but when the person 

19    who keypunched the data into the computer could have 

20    reversed the digits of Beatrice Jones' birthday and put her 

21    down as having been born in 1972.  

22               Then the computer would say, well, they were born 

23    on a different year so it can't be the same person.  But the 

24    clerk would see that she was the mother of Susan Jones, that 

25    all the other characteristics like sex and race matched and 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                215    

 1    would declare the person on the PES and the person on the 

 2    census to be the same person and, therefore, to have matched 

 3    the census.

 4         Q.    Were there cases in which this second stage of 

 5    matching even after that you could not tell whether or not 

 6    there was a match?

 7         A.    Well, first of all, I should say that there were 

 8    two stages of human being matching, so it went through two 

 9    of those stages.  After those two stages of human being 

10    matching, those cases where they still couldn't tell went 

11    through a procedure called field follow up.  In field follow 

12    up, Census Bureau interviewers went out to see if they could 

13    obtain the information that was missing or otherwise 

14    inconclusive so they can continue with the matching efforts.

15         Q.    And then they came back?

16         A.    After field follow up was completed, the new 

17    information was sent back to the Census Bureau and the case 

18    went to a fourth stage of matching.

19         Q.    After all the stages of matching, what are the 

20    possible outcomes in terms of the matching of the people 

21    found by the PES against the people who were enumerated in 

22    the census?

23         A.    Well, every person in the PES was considered 

24    either to have been matched to the census or not matched to 

25    the census or unresolved.  An unresolved case is one in 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                216    

 1    which even after all the stages of matching and follow up 

 2    interviewing still cannot be determined whether the person 

 3    was included in the census or not, the proportion of the PES 

 4    sample that was unresolved, if I could refresh my memory --

 5               (Pause)

 6               MR. SITCOV:  Might we know what he is using to 

 7    refresh his memory, your Honor?  

 8               THE COURT:  Might we know what you're using to 

 9    refresh your memory? 

10               THE WITNESS:  Yes.  This is the joint report 

11    submitted by myself and Leobardo F. Estrada, John Tukey and 

12    Kirk Wolter to the Secretary, dated June 21, 1991.  

13               MR. SITCOV:  Is that Exhibit 195? 

14               THE WITNESS:  That would have been a short way to 

15    say it, yes.  I'll get it yet.

16               1.5 percent of the P. Sample was unresolved.

17         Q.    What did the Census Bureau do with respect to 

18    these 1.5 percent cases?

19         A.    It used a standard Census Bureau procedure known 

20    as imputation to estimate a probability of being counted in 

21    the census for each person.

22         Q.    Could you explain to the judge what that is.

23         A.    Yes.  These are people for which a substantial 

24    amount of information was obtained.

25               For example, if we know that somebody in the P. 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                217    

 1    Sample is --

 2         Q.    You referred to a phrase now that I don't think 

 3    we have ever used before so you might want to explain it. 

 4               What do you mean by the P. Sample?

 5         A.    I apologize.  I slipped into a bit of lingo.  

 6               The Census Bureau term for the PES sample I have 

 7    been describing is known as the P. Sample, for population.

 8               Let's say there was a 29 year old black male 

 9    living in rented housing in New York City, there may have 

10    been, let's say, a hundred such people in the PES.  Of those 

11    hundred people, let's say that 80 of them have been found to 

12    have been counted and 20 of them have been found to have 

13    been omitted.  Then there would be a probability of being 

14    counted of 80 percent.  

15               So it is as if 80 such people out of a hundred 

16    had been counted.  So the probability of .8 would be 

17    assigned to that person.

18         Q.    The unresolved case?

19         A.    Yes.  Now, if there were 10 unresolved cases with 

20    just those characteristics, then assigning each of them a 

21    probability of 80 percent would be the same thing as saying 

22    that 8 out of 10 of those guys had been counted and 2 out of 

23    10 of those guys had been omitted, and those proportions 

24    would be about the same as the proportions of similar people 

25    where they had been able to tell had that person been 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                218    

 1    counted or omitted.

 2         Q.    Up to till now you have been talking about, to 

 3    use your lingo, P. Sample?

 4         A.    Right.

 5         Q.    Does the PES also have a way of dealing with 

 6    people who appeared on the census form, but who should not 

 7    have been actually counted?

 8         A.    Yes.

 9         Q.    How does the PES do that?

10         A.    Okay.  Just as we can regard the PES sample as 

11    the sample of people living in those blocks, then we can 

12    also take all the census forms that were obtained in those 

13    blocks and we can do the same kind of matching and follow up 

14    interviewing procedure to determine the proportion of those 

15    that are erroneous enumerations and the proportions of those 

16    that are correct enumerations, because to get the net 

17    undercount, we need to get omission rates and rates of error 

18    enumerations.

19         Q.    How does the Census Bureau go about determining 

20    the error enumeration rate?

21         A.    Well, the first stage, of course, would be the 

22    computer matching that I have just now been talking about. 

23               To the extent that somebody in the PES matches 

24    somebody in the census, then that person would have been 

25    considered to have been correctly counted in the census.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                219    

 1    For all those people in the census --

 2         Q.    I think maybe we -- sorry.  I think you may have 

 3    jumped one step.  Go ahead.

 4         A.    Now, for all those people who are on the census 

 5    forms, we can call them the E Sample.  They're matched 

 6    against the people in the PES interviews.  Those that don't 

 7    match, the next stage is they go to --

 8         Q.    Before you go to the next stage, perhaps you can 

 9    make it more concrete for me if you use the same metaphor of 

10    the two piles and explain what we're now doing that is 

11    different from what the Census Bureau did before.

12         A.    Oh, sure.  Basically if we've got the pile of PES 

13    forms here for the block and the pile of census forms here 

14    for the block, the last time we matched the PES forms 

15    against the census forms to see who was missed.  Now we're 

16    taking the census forms and matching them against the PES 

17    forms to see who was correctly counted.

18         Q.    Now tell the judge how the second process goes.

19         A.    Okay.  Now, the first stage of the second process 

20    is the computer matching that I've already talked about.  In 

21    other words, somebody who exactly links up is considered to 

22    have been counted.

23               Now, from the E Sample side --

24         Q.    You used another lingo.  Why don't you tell the 

25    judge what E Sample is.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                220    

 1         A.    E Sample is a term of lingo for the pile of 

 2    census forms, the best way to say it.

 3               Now, the census forms, they look throughout the 

 4    surrounding block to see what proportion of people were 

 5    duplicated; in other words, counted twice in the same block, 

 6    perhaps the same house got counted twice, that would be a 

 7    duplication.

 8               Then once the search for duplicates has been 

 9    completed and once the computer matching has been completed 

10    and the clerical matching that is followed to be completed, 

11    all the E Sample cases that have not yet been matched are 

12    sent to field follow up.  

13               In field follow up, there is a search to see if 

14    the P. Sample -- it could be the person was correctly 

15    counted and the PES interviewer simply didn't find that 

16    person.  It could be that the person was a fabrication, it 

17    could be that the person's usual residence wasn't there.  

18    All that information is determined.  

19               At the end of the process, every person in the E 

20    Sample is determined to have been correctly counted in the 

21    census, erroneously included in the census, or unresolved.

22         Q.    What happens to the unresolved cases?

23         A.    In the unresolved cases, we go through an 

24    imputation process very similar to the one I've just 

25    described, and a probability of correct enumeration is 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                221    

 1    determined for each person.  The rate of unresolved, the 

 2    proportion of the E Sample that was unresolved was about 1 

 3    percent.

 4         Q.    At the end of of the process, did the Census 

 5    Bureau determine a rate of omission for each of the 

 6    post-strata?

 7         A.    Yes, it did.

 8         Q.    And also a rate of error enumeration?

 9         A.    Yes, it did.

10         Q.    And also in combination a rate of net undercount?

11         A.    Yes, the omission estimate and the error 

12    enumeration estimate together and estimates the percent 

13    undercount for each of the 1,392 post-strata.  

14         Q.    What is the next step in the process after this 

15    net undercount rate?

16         A.    The next step in the process is to calculate a 

17    quantity known as an adjustment factor.

18         Q.    Perhaps you could just go up to the board and 

19    explain to the judge how adjustment factors are calculated.

20         A.    Okay.  Let's say for a post-stratum, the PES 

21    undercount -- sorry -- the PES estimates a net undercount 

22    rate of 5 percent.  This means that 95 percent of the 

23    estimated true population was counted by the census and 5 

24    percent was not.  The adjustment factor is a quantity that 

25    you would multiply by the 95 percent -- 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                222    

 1               MR. SITCOV:  Excuse me, your Honor.  I hate to 

 2    interrupt the witness, but I really must protest the use of 

 3    this chart.  We received this by fax last night for the 

 4    first time and I don't see that there is any way that that 

 5    can be considered the kind of advance notice that we're 

 6    entitled to have for trial exhibits in this case.  We got it 

 7    the evening of the first night of trial.  

 8               THE COURT:  Well, he could have testified without 

 9    any chart and I would be straining to try to write this 

10    stuff down, so I am really not overwhelmed by the claim of 

11    prejudice when it is really not evidence, it is just 

12    illustrative of his testimony.  

13               MR. ZIMROTH:  Thank you, your Honor.

14         A.    Okay.  Going back, the adjustment factor is a 

15    quantity that we multiply by the 95 percent to get up to a 

16    whole population, 100 percent.

17               So to get the adjustment factor, we divide the 

18    hundred percent by the 95 percent, and it comes out to 

19    1.053.  So point 95 times 1.053 is equal to 1.

20               In other words, you have to -- to get from the 

21    census count to the estimated true count, you multiply the 

22    census count by 1.053.  So if the count was 100,000, then 

23    you multiply 100,000 by 1.053, and the estimated count for 

24    that post-stratum would be 105,300.

25         Q.    Now, professor -- thank you.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                223    

 1               (Pause) 

 2         Q.    Is the adjustment factor that you have just 

 3    illustrated the adjustment factor that was actually used by 

 4    the Census Bureau to do the adjustment for the 1990 census?  

 5               THE COURT:  Would you restate that?  I lost you.  

 6               MR. ZIMROTH:  Okay.

 7         Q.    What is this adjustment factor called in the 

 8    lingo that you just described?

 9         A.    The lingo term for that is the raw, R A W, 

10    adjustment factor.

11         Q.    Did the Census Bureau use raw adjustment factors?

12         A.    No.

13         Q.    When it --

14         A.    I am sorry.

15         Q.    It did not?  Okay.

16               In lingo, what is the adjustment factor that you 

17    used to call it?

18         A.    It is called the smooth adjustment factor.

19         Q.    Professor Ericksen -- and for your Honor's 

20    edification, we are going to have another witness who will 

21    testify at some length about smoothing, but just in a 

22    sentence or two can you tell Judge McLaughlin what smoothing 

23    is?  

24               THE COURT:  The opposite of raw.  

25               MR. ZIMROTH:  Excuse me?  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                224    

 1               THE COURT:  The opposite of raw.  

 2               MR. ZIMROTH:  Right, the opposite of raw.

 3         A.    Okay.  What is meant by the opposite, I suppose, 

 4    is that the raw adjustment factors have sampling errors.  We 

 5    like to reduce the effects of the sampling error, and 

 6    smoothing is a well known statistical technique that permits 

 7    the reduction of the effect of the sampling error.

 8         Q.    After the Census Bureau determined the smooth 

 9    adjustment factors, how did it estimate the population of a 

10    particular place, for example, take the city of Washington, 

11    DC?

12         A.    Well, by my count, in the city of Washington, DC 

13    there were 240 post-strata represented, so if we can imagine 

14    a post-stratum, let's say, black females aged 30 to 44, 

15    there might have been 10,000 such people counted in 

16    Washington, DC.  If the adjustment factor for that group, 

17    let's say, was 1.10, then you would take the 10,000 and 

18    multiply it by 1.10, and the estimated true count would be 

19    11,000.

20               If, to give another example, you were concerned 

21    about white females aged 65 and over, and if there were 

22    1,000 of those living in Washington, DC, then the adjustment 

23    factor was 1.02, you would take the 1,000, multiply it by 

24    1.02, and the estimated true count would be 1,020.

25               After you have done that for each of the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                225    

 1    post-strata, you simply add them up, and the sum would be 

 2    the estimated true population of Washington, DC.

 3         Q.    Did the Census Bureau do that for every state in 

 4    the nation?

 5         A.    Yes, the Census Bureau actually applied the 

 6    adjustment factors to every block.  Then it summed the 

 7    estimated true population for each block to get the 

 8    estimates for all the towns and cities and counties and 

 9    states and finally for the nation.  

10               THE COURT:  But again they used the raw 

11    adjustment factors? 

12               THE WITNESS:  No, no.  They used the smooth 

13    adjustment factors.  In other words, that is why they had to 

14    go through that step because the smooth adjustment factors 

15    are the adjustment factors that they used.  

16               MR. ZIMROTH:  Your Honor, I have a bit of a 

17    logistical problem, which is that at this point I wanted to 

18    introduce into evidence the tapes that were produced by the 

19    Census Bureau in accordance with the process that Professor 

20    Ericksen testified about.  Those tapes, as you know, are 

21    under seal and they're now actually physically in Boston, 

22    and I am wondering whether or not we can get a stipulation 

23    -- how do we want to handle this?  These tapes ought to be 

24    in evidence in some way.  

25               THE COURT:  Well, why don't we deem them marked 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                226    

 1    in evidence as a court exhibit.  

 2               MR. ZIMROTH:  Thank you.  

 3               MR. SITCOV:  We can certainly introduce their 

 4    picture of them as a court exhibit.  

 5               MR. ZIMROTH:  That is fine, too.  

 6               THE COURT:  The picture will do instead?  

 7               MR. ZIMROTH:  No.  Actually I think we ought to 

 8    have deemed into evidence the actual tapes.  

 9               THE COURT:  And we'll deem them marked as Court 

10    Exhibit 1.  

11               MR. SITCOV:  Subject to their protective order.  

12               THE COURT:  Subject to the protective order.

13               (Court Exhibit 1 for identification was received 

14    in evidence) 

15         Q.    Did the Census Bureau issue a press release 

16    describing the results of the PES?

17         A.    Yes, it did.

18         Q.    Will you look at 478, please.  

19               Is that the press release that describes the 

20    results of the PES?

21         A.    Yes.  There is a second press release issued June 

22    13, 1991, and this press release presented the results of 

23    the PES.  

24               MR. ZIMROTH:  Just to remind your Honor, the 

25    first release was a previous exhibit also June 13 describing 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                227    

 1    the results of the demographic estimates.

 2         Q.    Now, may we have the next chart, please.

 3               (Pause) 

 4         Q.    Perhaps if you can come because I want to ask you 

 5    some questions about the chart 479.

 6               What is that chart, 479?

 7         A.    Okay, this is a chart that presents the estimated 

 8    undercounts from the PES by race and Hispanic origin.

 9         Q.    And the source of that is the --

10         A.    The source of that, I believe it is Table 4 -- 

11    yeah, it's Table 4 of the June 13 press release.

12         Q.    Just -- because there are two of them, I think 

13    that is 478.  Go ahead.

14         A.    Now, here we have the group of Hispanics.  These 

15    are the people who answered Question 7 on the census form 

16    that they were of Hispanic origin, the PES estimated their 

17    undercount rate to be 5.2 percent.

18               Here we have people who answered Question 4 of 

19    the census form they were black.  Their estimated undercount 

20    rate was 4.8 percent.

21               For people who answered Question 4 that they were 

22    Asian or Pacific Islander, their estimated undercount rate 

23    was 3.1 percent.

24               The estimate for the entire nation was 2.1 

25    percent, and the estimate for all persons other than black 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 2             Ericksen - direct                228    

 1    was 1.7 percent.

 2         Q.    Now, could you explain to the judge who is 

 3    included in this last tower, the all persons other than 

 4    black?

 5         A.    In this tower here the nonblack includes all 

 6    people who were Hispanic who did not also say that they were 

 7    black.  Roughly 4 percent of people who said they were 

 8    Hispanics also said they were black.  So the remaining 96 

 9    percent of Hispanics are included in the nonblack group.  

10    The nonblack group also includes Asian Pacific Islanders, it 

11    includes Native Americans and it includes white people who 

12    are not Hispanic.

13         Q.    If you took out everybody in this last tower 

14    called nonblack other than the white people who were not 

15    Hispanic, what would happen to that tower?

16         A.    It would get shorter and the estimated undercount 

17    for people who are white but who are not Hispanic are 

18    somewhat around 1.1 to 1.2 percent.  

19               MR. ZIMROTH:  Your Honor, at this time I move the 

20    admission of -- no, sorry.  One other thing.

21         Q.    Did this racial difference and ethnic difference 

22    that you have pointed out here also have a geographical 

23    component?

24         A.    Yes, it did.

25               (Continued on next page) 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 229    

 1    (BY MR. ZIMROTH:)

 2         Q.    Would you look at 707 and tell us what that 
                                                             

 3    represents.  

 4               MR. SITCOV:  I don't believe we have a 707, your 

 5    Honor.  

 6               MR. ZIMROTH:  Yes, we did.  You have it.  

 7               MR. SITCOV:  We are going to object to the use of 

 8    this chart.  Again, your Honor, it's one that we got, I 

 9    think, on Saturday.

10               THE COURT:  I don't have 707 in front of me.

11               (Pause) 

12               MR. ZIMROTH:  Your Honor, may I just for a 

13    moment, I hope, short-circuit this?

14               This chart is nothing more than a graphic 

15    illustration of what appears in the press release.  That's 

16    all this is, as Mr. Sitcov well knows.  

17               MR. SITCOV:  How am I supposed to know that?  

18               MR. ZIMROTH:  Because we told you when we sent it 

19    to you.  

20               MR. SITCOV:  The day before we left to come here.

21               THE COURT:  I'll permit it.  Go ahead.

22    BY MR. ZIMROTH:

23         Q.    Go ahead.

24         A.    Now I've forgotten the question.

25         Q.    The question is:  The racial differential that is 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 230    

 1    exhibited in 479, does that have a geographical component?

 2         A.    Yes.  What this chart presents is the net 

 3    undercounts for Washington, D.C. and the 50 states.  At this 

 4    end of the chart, Washington, D.C. has a net undercount of 5 

 5    percent.  And at this end of the chart, Rhode Island has a 

 6    net undercount, as I recall, of 0.4 percent.

 7               Now, Washington, D.C., is about 75 percent 

 8    minority, a bit less than that.  The second highest 

 9    undercount is in New Mexico, where the percent minority is 

10    about 50 percent.  In New Mexico, the net undercount is 

11    about 4 1/2 percent.

12               The third undercount is California, where the 

13    percent minority is about 43 percent.  The net undercount is 

14    3.7 percent.

15               And similarly, if you go down the states, at this 

16    end, the percent minority tends to be high.

17               At the end down here, you had states like Rhode 

18    Island, Massachusetts, Connecticut, New Hampshire, 

19    Pennsylvania and Wisconsin, where the proportions of 

20    minority are quite low, relative to the national average, 

21    and we can see that the percent undercounts are low in these 

22    states.

23               So in states where you have concentrations of 

24    minorities, the percent undercount is usually high.  In 

25    states where you have a low proportion of minority, the net 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 231    

 1    undercount is usually low.  

 2               MR. ZIMROTH:  At this time, your Honor, I move 

 3    the admission of PX 478, 479 and 707.  

 4               MR. SITCOV:  We object to 707.

 5               THE COURT:  The objection is overruled.  All 

 6    three are admitted.

 7               (Exhibits PX 478, 479 and 707 for identification 

 8    were received in evidence) 

 9    BY MR. ZIMROTH:

10         Q.    Professor Ericksen -- 

11               THE COURT:  Might this be a good point to take a 

12    break?  

13               MR. ZIMROTH:  Sure.

14               THE COURT:  All right.  Let's take a 15-minute 

15    mid-morning stretch.

16               (Recess)

17               (In open court) 

18    EUGENE P. ERICKSEN, resumed

19               THE COURT:  Mr. Zimroth.  

20    DIRECT EXAMINATION CONTINUED 

21    BY MR. ZIMROTH:

22         Q.    Professor Ericksen, did the Census Bureau study 

23    the question of -- or compare the accuracy of, on the one 

24    hand, the original enumeration, and, on the other hand, the 

25    original enumeration as supplemented by the information 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 232    

 1    obtained through the PES?

 2         A.    Yes, it did.

 3         Q.    And what was the conclusion by the Census Bureau 

 4    on that subject?

 5         A.    The conclusion of the Census Bureau is that the 

 6    original enumeration supplemented by the information 

 7    collected by the PES produced a set of estimates that were 

 8    more accurate than the estimates provided by the original 

 9    enumeration -- 

10               MR. SITCOV:  Objection, your Honor.  I move to 

11    strike the testimony.  He's testifying in the present tense 

12    and there is no evidence that that is the present view of 

13    the Census Bureau.

14               If he wants --

15               THE COURT:  With that understanding, I'll accept 

16    it.  They concluded at some point.  

17               MR. SITCOV:  Fine.

18               THE COURT:  Whether they adhered to it or not 

19    will be developed.  You've sensitized me.  

20               MR. SITCOV:  Thank you.  

21    BY MR. ZIMROTH:

22         Q.    Will you look at PX 55, please.

23               Do you have that?

24               (Pause) 

25               Do you have that document, Professor Ericksen?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 233    

 1         A.    Yes, I do.

 2         Q.    Do you know what that document is?

 3         A.    This is the recommendation of the director of the 

 4    Census Bureau, Barbara Bryant as to whether or not the 1990 

 5    Census should be adjusted.

 6         Q.    I just want to read to you two sentences and ask 

 7    your opinion about them.

 8               "Extraordinary efforts" --

 9               THE COURT:  Page?  

10               MR. ZIMROTH:  Page 3, your Honor, paragraph 4.

11         Q.    "Extraordinary efforts were made in 1990 to 

12    reduce the differential undercount.  The differential was 

13    not produced."

14         Q.    Do you agree with that conclusion?

15         A.    Yes, I do.

16         Q.    Now, would you turn to page 16, please.

17               And I'm going to read a few sentences from the 

18    bottom of that page --

19               MR. ZIMROTH:  The paragraph beginning at the 

20    bottom of the page, your Honor, and ask this witness to 

21    comment and give his opinion about that.

22         Q.    "The PES, supported by demographic analysis, 

23    estimates that the resident population of the United States 

24    on April 1, 1990, was approximately 5.3 million greater than 

25    was counted in the census.  The fact that both these Census 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 234    

 1    Bureau research projects, including the one based on 

 2    administrative records, rather than census data, produced 

 3    nearly the same 5 million number is strong evidence that 

 4    these residents of the United States exist."

 5               Do you agree with that conclusion?

 6         A.    Yes, I do.

 7         Q.    "Logic also supports the existence of people who 

 8    cannot or will not be counted, although logic cannot confirm 

 9    their numbers."

10               Do you agree with that?

11         A.    Yes, I do.

12         Q.    "In my opinion, not adjusting would be denying 

13    that these 5 million persons exist.  That denial would be a 

14    greater inaccuracy than any inaccuracies that adjustment may 

15    introduce."

16               Do you agree with those conclusions?

17         A.    Yes, I do.  

18               MR. ZIMROTH:  Your Honor, at this time I want to 

19    introduce PX 55.  

20               MR. SITCOV:  No objection.

21               THE COURT:  55 is admitted.

22               (Exhibit PX 55 for identification was received in 

23    evidence) 

24    BY MR. ZIMROTH:

25         Q.    Now, in the last sentence that I just quoted to 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 235    

 1    you is a reference to possible inaccuracies that adjustment 

 2    may introduce.

 3               Do you see that?

 4         A.    Yes, I do.

 5         Q.    Do you agree that adjustment might introduce 

 6    certain inaccuracies?

 7         A.    Yes, I do.

 8         Q.    And why is that?

 9         A.    Well, first of all, the post-enumeration survey 

10    is based on a sample.  And like any estimate based on a 

11    sample, there's going to be sampling error.  Even after the 

12    smoothing is done to reduce the effect of the sampling 

13    error, some will still be left.

14               It's also the case that the PES is not a perfect 

15    survey.  Some people who are missed by the census are also 

16    missed by the PES.

17         Q.    Now, does that latter thing you say have a 

18    technical name?

19         A.    That particular name is known as "correlation 

20    bias."

21         Q.    Could you just repeat to the Judge what 

22    correlation bias is?

23         A.    Yes.  Correlation bias is kind of an error that 

24    is created by the fact that some people were not counted in 

25    the census will also not be counted by the PES.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 236    

 1         Q.    And what is the effect of that sort of error or 

 2    bias?

 3         A.    The effect of that sort of error is to cause the 

 4    estimate of the undercount to be too low.

 5         Q.    And what other sorts of errors?

 6         A.    Well, another example would be matching error.  

 7    It's possible that some people will be counted as matches 

 8    when they were really nonmatches.  Other people we counted 

 9    as nonmatches when they are really matches.

10         Q.    And what would the effect of those errors be?

11         A.    Well, logically the effect could go either way.  

12    In practice, the effect of matching error is to cause the 

13    estimate of the undercount to be too high.

14         Q.    Okay.  Did the Census Bureau study the various 

15    sources of possible error that might have -- might have an 

16    effect on the PES?

17         A.    Yes, it did.

18         Q.    And did it compare the magnitude of those errors 

19    with the magnitude of the errors in the original enumera- 

20    tion?

21         A.    Yes, it did.

22         Q.    And what was the conclusion?

23         A.    That the magnitude of the errors caused by the 

24    PES were smaller than the magnitude of the errors caused by 

25    the original enumeration, both in terms of estimating the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 237    

 1    total population as well as its distribution.

 2         Q.    "...as well as its" --

 3         A.    Distribution.

 4         Q.    Thank you.

 5               In your opinion, Professor Ericksen, did the 

 6    Census Bureau undertake sufficient scrutiny or did they 

 7    scrutinize sufficiently the PES in order to come up with its 

 8    conclusion?

 9         A.    As far as I can tell, the 1990 post-enumeration 

10    survey may be the most scrutinized survey that's ever been 

11    taken.

12               In my opinion, they certainly exercised 

13    sufficient scrutiny.

14         Q.    How did it come about that the Census Bureau 

15    exercised the kind of scrutiny that you just testified 

16    about?

17         A.    Well, you will recall that the Census Bureau in 

18    1980 elected not to adjust the 1980 census.  As an 

19    organization, I think it's fair to say that the Census 

20    Bureau has always been skeptical about the possibility of 

21    adjustment.

22               MR. SITCOV:  Objection, your Honor.

23               THE COURT:  Sustained.  

24    BY MR. ZIMROTH:

25         Q.    What happened in 1980?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 238    

 1         A.    In 1980, the director of the Census Bureau made a 

 2    decision not to adjust the 1980 census for the differential 

 3    undercount.

 4         Q.    And how then did the Bureau come to scrutinize 

 5    the 1990 Census PES the way that you testified that it did?

 6         A.    Well, I think that the Census Bureau, if it was 

 7    going to adjust, wanted to make sure that it has -- 

 8               MR. SITCOV:  Object.  There's no way that this 

 9    witness can give that testimony.

10               THE COURT:  Sustained.  

11               MR. ZIMROTH:  Your Honor, there are many, many 

12    documents that support what this witness is going to say.  

13    There is a whole history here of what happened during the 

14    decade of the 1980s.

15               THE COURT:  Point out what the documents say 

16    rather than simply stating conclusions.

17               MR. ZIMROTH:  This witness has read hundreds of 

18    documents.  He was intimately involved with a lot of 

19    processes.  He reviewed a lot of the documents as a member 

20    of the Special Advisory Panel.

21               THE COURT:  Well, still, the rules of evidence 

22    should apply at least in a gross fashion.  He's invading my 

23    function when he starts making conclusions like that.

24    BY MR. ZIMROTH:

25         Q.    Are you personally aware of the research that the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 239    

 1    bureau did during the course of the 1980s --

 2         A.    Yes.

 3         Q.    -- on the question of adjustment?

 4         A.    Yes, I am.

 5         Q.    Just tell the Judge what you were personally 

 6    aware of.

 7         A.    Well, first of all, the Census Bureau established 

 8    a substantial budget in the millions of dollars for research 

 9    on the undercount.  Then the Census Bureau, a group called 

10    the Undercount Research Staff was created.  The director of 

11    the Undercount Research Staff was Kirk Wolter, together with 

12    his colleagues they identified.  

13               I believe it was eight sources of error that 

14    affected their judgment of the estimates of undercount in 

15    the 1980 census.  And they developed a series of research 

16    projects that would permit them to develop procedures that 

17    would reduce the effects of those errors.  They also 

18    developed methods which would allow them to measure the size 

19    and the magnitude of those errors.  They commissioned 

20    research to be done by statisticians across the United 

21    States.  There were research projects that took place at 

22    many different universities.  It was as if they were 

23    engaging the statistical community in this new statistical 

24    project called "undercount estimation."

25               Many of the projects that the Census Bureau 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 240    

 1    completed became scientific articles that were published in 

 2    journals.

 3         Q.    Did the Census Bureau test its methodology during 

 4    the period of the '80s?

 5         A.    Yes, they did.  There were many small scale 

 6    tests.  I could not tell you what they all were.  A major 

 7    test took place in 1986.

 8               In 1986, the Bureau took a test census in a 

 9    section of Los Angeles County, California, and they took a 

10    test PES there as well, and went through all the operations 

11    that they expected to use in 1980, estimated the undercount 

12    for that local area, and also evaluated the apparent errors 

13    of PES.

14               They did test censuses in St. Louis, Missouri, in 

15    1988.  They did another one in an area of East Central 

16    Missouri.  They did a third one in a rural area of 

17    Washington State.  And applying whatever lessons they 

18    learned from 1986, they used their revised PES methods and 

19    did the same thing there.  Afterwards, they made whatever 

20    modifications they deemed appropriate for use in the 1990 

21    PES.

22         Q.    And did the Census Bureau also develop tests 

23    criteria upon which it would judge the success of the PES?

24         A.    Yes, it did.

25         Q.    And did it develop tests to determine whether the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 241    

 1    PES met those criteria?

 2         A.    Yes, it did.

 3         Q.    Are you aware that under the stipulation and 

 4    order the defendants were required to publish guidelines for 

 5    decisions in this case?

 6         A.    Yes.

 7         Q.    And in your capacity as co-chair of the Special 

 8    Advisory Panel, did you receive a draft preliminary 

 9    technical guideline?

10         A.    Yes, I did.

11         Q.    Will you look at PX 451 B.  

12               MR. SITCOV:  Your Honor, I may as well object now 

13    before he starts testifying.

14               This is a draft of the guidelines which you 

15    approved in your order that's reported in 739 F.Supp.  It is 

16    not the one that was adopted.  It includes attachments that 

17    were never even published in the Federal Register when the 

18    defendants complied with the requirement in the stipulation 

19    and order that they publish draft guidelines.

20               The plaintiffs have attacked the final guide- 

21    lines.  Your Honor has rebuffed their attack, and in fact 

22    found it to be valid and to be in full compliance with the 

23    stipulation and order, and there can't be any relevant 

24    reason for their introduction now, to say nothing of the 

25    many foundational problems about introducing this document.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 242    

 1               THE COURT:  Mr. Zimroth, what's the relevancy of 

 2    the draft?  

 3               MR. ZIMROTH:  I can adopt Mr. Sitcov's entire 

 4    statement.  He has pointed out its relevance that these 

 5    guidelines were not adopted.

 6               For example, one the guidelines that are here 

 7    says that the test should be down to the census block level.  

 8    That was omitted from the final guidelines.

 9               Mr. Sitcov is making precisely the point that we 

10    want to make through the introduction of this draft.  In 

11    other words, that this draft shows that block level accuracy 

12    was removed from the final version as it was promulgated.  

13               MR. SITCOV:  It does no such thing, your 

14    Honor, --

15               MR. ZIMROTH:  We can argue about that down the 

16    road.  

17               MR. SITCOV:  -- because the final guidelines 

18    mention the same thing.  The fact of the matter is these 

19    guidelines were not adopted, they are clearly marked a 

20    draft, you have approved the guidelines that we are 

21    adopting, and we've been operating under them for the past 

22    few years -- 2 1/2 years anyway.  

23               MR. ZIMROTH:  Your Honor, these guidelines also 

24    set forth standards which were ultimately published by the 

25    Census Bureau in technical operation plans which led to all 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 243    

 1    the evaluative studies that this witness is going to testify 

 2    about.  

 3               MR. SITCOV:  Then let them introduce the plans.

 4               THE COURT:  Well, one of the countless issues 

 5    that seem to be arising here is the issue of good faith of 

 6    the Census Bureau.  

 7               MR. ZIMROTH:  Of the Commerce Department.

 8               THE COURT:  Pardon? 

 9               MR. ZIMROTH:  Secretary of Commerce.

10               THE COURT:  Yes.  And as one of the tiles in that 

11    mosaic of bad faith that the plaintiffs are trying to paint, 

12    a draft would be usable.  

13               MR. SITCOV:  Well, your Honor, they made the same 

14    argument at the time that they attacked the guidelines.  And 

15    in fact, Mr. Zimroth himself, personally, during the 

16    argument, mentioned the document that begins about seven 

17    pages in.  It says:  "Preliminary technical guidelines for 

18    the Decennial Census adjustment."  He made that a point of 

19    his argument at the time --

20               THE COURT:  For what?  You mean I rejected as a 

21    matter of law any claim of bad faith?  

22               MR. SITCOV:  No.  The point is that these draft 

23    guidelines are not the guidelines that were adopted.

24               THE COURT:  We all know that.  

25               MR. SITCOV:  Okay.  They made a charge of bad 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 244    

 1    faith at the time.

 2               THE COURT:  Right.  

 3               MR. SITCOV:  Your Honor rejected that charge and 

 4    it was based partly on this.

 5               THE COURT:  Partly.  

 6               MR. SITCOV:  Yes.

 7               THE COURT:  Right.  That's the fallacy in the 

 8    argument.  

 9               MR. SITCOV:  Well, I don't believe it is, your 

10    Honor, with all --

11               THE COURT:  The objection is overruled.  The 

12    draft is admissible, 451 B, if not yet offered, is deemed 

13    offered.  

14               MR. ZIMROTH:  And admitted.

15               (Exhibit PX 451 B for identification was received 

16    in evidence) 

17               MR. ZIMROTH:  Actually, your Honor, this is my 

18    fault.  Your Honor already admitted this document because it 

19    is part of one that's already in evidence --

20               THE COURT:  It's already in evidence then?  

21               MR. ZIMROTH:  -- as 451.  Yes, it's part of 451.

22               THE COURT:  So we are tilting at old windmills.  

23               MR. ZIMROTH:  I apologize, your Honor.  

24               MR. SITCOV:  Well, your Honor, then I will only 

25    point out, if I might, that this Ericksen affidavit is the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 245    

 1    one, I believe, that was submitted -- yes, it is -- at the 

 2    time of plaintiff's attack on the guidelines.

 3               THE COURT:  Okay.  It's noted.  

 4    BY MR. ZIMROTH:

 5         Q.    Did the draft that has just been admitted, 451 B, 

 6    include criteria for determining the accuracy of the PES?

 7               Let me say:  Did it call for evaluative 

 8    procedures in that respect?

 9               (Pause)

10         Q.    You might look at Exhibit D.

11         A.    Yes, it does.

12         Q.    And as Mr. Sitcov just testified, those 

13    evaluative procedures were not included as part of the final 

14    guidelines, is that correct?

15         A.    That's correct.

16         Q.    What happened to this?  

17               MR. SITCOV:  Object.  How could he possibly know?

18               THE COURT:  If you know.

19         A.    All I know is that they weren't there.

20         Q.    And do you know what happened to the plans that 

21    were accomplished in these?  

22               Were they promulgated in a subsequent Bureau 

23    document?

24         A.    Yes.  They -- we received a document entitled 

25    "Technical Operations Plans."  I believe it was either in 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 246    

 1    March or April of 1990.

 2         Q.    Look at PX 491.  Is that the document to which 

 3    you refer?

 4         A.    Yes, it is.

 5         Q.    And did that document set forth the criteria for 

 6    judging the accuracy of the PES?

 7         A.    Yes, it did.

 8         Q.    And did it also call for evaluative procedures to 

 9    determine whether the criteria were met?

10         A.    Yes, it did.

11         Q.    What were these evaluative procedures called?

12         A.    They are generally called P studies and D 

13    studies.  The P studies were a series of about 22 studies 

14    that were used to estimate the effects of correlation bias, 

15    matching error --

16         Q.    What were the purpose of the P studies?

17         A.    To evaluate the accuracy the PES estimates.

18         Q.    And can you be more specific in terms of how the 

19    Bureau broke that down in the P study?

20         A.    Sure.  There was a P study to evaluate the effect 

21    of correlation bias.  That was P 13.

22               There was a study to evaluate the effect of 

23    matching error.  Actually, there were two of those, P 7 and 

24    P 10.

25               There was a study to study the consistency 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 247    

 1    between the PES and demographic analysis.  That was P 17.

 2               There was a study, P 16, to evaluate the combined 

 3    effect of all the errors.

 4         Q.    Okay.

 5         A.    All of this is in addition to a series of 11 

 6    studies which evaluated the quality of the demographic 

 7    estimate.

 8               Now, I have put before you four binders that are 

 9    marked 540 A.  What do you recognize those to be?

10         A.    These three binders over here on my left 

11    (indicating) are the body of the reported research from the 

12    P studies.  And this fourth binder here (indicating) is the 

13    executive summaries of the P studies.

14         Q.    Those were all completed before the July 15th, 

15    1991, correct?

16         A.    Yes.

17         Q.    Now, at the time -- 

18               MR. ZIMROTH:  I move those into evidence, your 

19    Honor.  

20               MR. SITCOV:  We welcome them in.

21               THE COURT:  They are admitted with gratitude.

22               (Exhibit PX 540 A for identification was received 

23    in evidence) 

24    BY MR. ZIMROTH:

25         Q.    At the time you --


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 248    

 1         A.    May I put these over there?

 2         Q.    Sure.  I'll help you.

 3               (Pause)

 4               At the time you wrote your report on June the 

 5    21st, I believe, or submitted it on June the 21st, had you 

 6    received a final version of all these?

 7         A.    No.

 8         Q.    Had you received underlying documentation 

 9    relevant to those studies?

10         A.    Yes, I have.

11         Q.    All right.  You have four volumes in front of you 

12    now.  Will you just identify for the judge what they are.

13         A.    This volume is the same as this volume.

14         Q.    Oops.  Okay.  Well let's take that volume out.

15               (Pause)

16               Okay.  Would you identify what they are.

17         A.    Yes.  These are some of the P studies materials 

18    that I received to perform my duties as a member of the 

19    Special Advisory Panel.  

20               MR. ZIMROTH:  I move those into evidence, your 

21    Honor.  

22               MR. SITCOV:  They could have them.

23               THE COURT:  They are all admitted.

24               (Exhibit 540 B for identification was received in 

25    evidence)


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 249    

 1               THE COURT:  Don't they have a number?  

 2               MR. ZIMROTH:  Yes.  I wanted to just describe to 

 3    the Judge how both of these exhibits are organized.

 4               540 A are the final P studies, and they are 

 5    tabbed, each one, so that 540 A 1 is the first P study.

 6               540 A 2 is the second and seriatim until the end.           

 7    And these are the B series, or the materials that Professor 

 8    Ericksen testified that he received, and they are organized 

 9    similarly, 540 B 1 through whatever, relating to whichever P 

10    study -- its related to. 

11               (Discussion off the record) 

12               MR. ZIMROTH:  Your Honor, apparently I neglected 

13    to move the admission of PX 49.  The technical operation 

14    plans that the witness testified about earlier.  

15               MR. SITCOV:  The draft document?  

16               MR. ZIMROTH:  It's 491.

17               THE COURT:  I think that is the draft.  

18               MR. ZIMROTH:  No, that one is not.

19               THE COURT:  No?  

20               MR. ZIMROTH:  I don't believe. 

21               MR. SITCOV:  We have no objection to that.

22               THE COURT:  491 is admitted.

23               (Exhibit PX 491 for identification is received in 

24    evidence)

25    BY MR. ZIMROTH:


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 250    

 1         Q.    Mr. Ericksen you testified earlier -- let us get 

 2    back on the stream before I got pulled off by lugging all 

 3    these exhibits up -- that the Bureau combined the various 

 4    sources of error, the various studies into which you call 

 5    the total error model.

 6         A.    That is correct.

 7         Q.    And what was the point of the total error model?

 8         A.    Well, the point of the total error model was to 

 9    estimate the quality of the PES estimate of both for the 

10    nation but also for the distribution of population.

11         Q.    And what was the conclusion concerning the total 

12    error model?

13         A.    Well, the conclusion was that the effect of error 

14    in the PES on the adjusted distribution of population was 

15    small.

16         Q.    In your opinion, Professor Ericksen, when the 

17    Bureau came to that conclusion, did it set a high standard 

18    for itself?

19         A.    Yes, it did.

20         Q.    And in your opinion, when it came to the 

21    conclusion it did, was it correct that it met its high 

22    standard?

23         A.    Yes, it did.

24         Q.    And in terms of the quality of the PES, on what 

25    do you base your opinion that it was correct; that it had 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 251    

 1    met its high standards?

 2         A.    Well, first of all, I did review the P study 

 3    materials that I was sent.  I also obtained from the Census 

 4    Bureau access to the PES data files themselves.  I was able 

 5    to analyze the PES data and conduct my own research based on 

 6    a remote terminal in the Census Bureau office in 

 7    Philadelphia.

 8               So based on my review of what the Census Bureau 

 9    did, and based on my own analysis of the data, I concluded 

10    that the P studies were of high quality.

11         Q.    What is it that you were looking for when you 

12    came to that conclusion?

13         A.    I wanted to, first of all, satisfy myself that 

14    the errors of the PES hadn't been appropriately measured.           

15    But, more importantly than that, I wanted to know how the 

16    errors of the PES would have affected the estimated 

17    distribution of the population across areas.

18         Q.    And what did you conclude on those two grounds?

19         A.    I concluded that the effect of the error in the 

20    PES was small.

21         Q.    And did you look at evidence external to the PES 

22    in coming to your conclusion?

23         A.    Yes, I did.

24         Q.    Why did you do that?  Why would it be important 

25    to you to look at evidence that was external to the PES?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 252    

 1         A.    Oh, I could step back for a minute.  Any 

 2    scientific study has to be based on some kind of theoretical 

 3    expectations.  If you go looking at data without any 

 4    scientific theory about what you are going to find, you'd be 

 5    observing chaos.

 6               My own theory of the undercount is based 

 7    substantially on the years of research that had been done on 

 8    it, both inside the Census Bureau and elsewhere.  So what 

 9    you look for for external corroboration is to find evidence 

10    of the undercount which had suffered from the PES.  And you 

11    ask a question of:  Are the results of the PES consistent 

12    with this external evidence?

13         Q.    Okay.  Now, from sources other than the PES now, 

14    external to the PES, what do you know about whether 

15    minorities are likely to be undercounted?

16         A.    Well, we know from demographic analysis that 

17    minorities are more likely -- I should say, more precisely, 

18    we know that blacks are more likely than non-blacks to have 

19    a high undercount.

20         Q.    Do we know from any other source besides 

21    demographic analysis?

22         A.    Well, we know from ethnographic studies that 

23    minority populations in general are likely to have a high 

24    undercount.

25         Q.    Okay.  And what is it that the PES showed us 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 253    

 1    about the likelihood of minorities being undercounted in the 

 2    1990 Census?

 3         A.    The PES showed that minorities had a high 

 4    undercount.

 5         Q.    Okay.  Independent of the PES, in your opinion, 

 6    is anything known about the relationship between mailback 

 7    rates and the likelihood of being counted?

 8         A.    Yes.  There are a number of studies that were 

 9    done of the 1980 census, and these studies showed 

10    consistently that in areas where the mailback rate was low, 

11    the undercount was high.

12         Q.    And why would that be?

13         A.    Because, as I explained yesterday, where the 

14    mailback rate is low, the nonresponse follow-up workload is 

15    high, the census is harder to take.  And where the census is 

16    harder to take, you'd expect the undercount to be high.

17         Q.    Was there also a study in New York City that bore 

18    upon this discussion?

19         A.    Yes.  The studies that look at the relationship 

20    between the mailback rate and the undercount were done both 

21    inside and outside the Census Bureau.  One of those studies 

22    was the study of the 20 district offices of New York City.  

23    The Census Bureau estimated the omission rate for each of 

24    these district offices --

25         Q.    Excuse me.  That's of the 1980 census?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 254    

 1         A.    That's of the 1980 census.  And in those district 

 2    offices where the mailback rate was lowest, the omission 

 3    rate tended to be highest.  In those offices where the 

 4    mailback rate was highest, the omission rate tended to be 

 5    lower.

 6         Q.    And what is it that the PES showed about the same 

 7    phenomenon?

 8         A.    The PES showed that in areas where the mailback 

 9    rate was low, the admission rate was high and vice versa.  

10               MR. ZIMROTH:  Okay.  I put up PX 494, and could 

11    you perhaps come up and explain this to the Judge.  

12               THE WITNESS:  Okay.  This particular chart --

13               MR. ZIMROTH:  Wait for His Honor.

14               THE COURT:  That's all right.

15               THE WITNESS:  This particular chart is based on 

16    an analysis of the 1,392 poststratum that I told you about 

17    this morning.  And what I did here was to take those 1,392 

18    poststrata and group them by what the mailback rate was, so 

19    I created eight categories of the poststrata depending upon 

20    what the mailback rate was.  

21               So in the category over here on the far side 

22    (indicating), this is the category of poststrata where the 

23    mailback rate was 85 percent or greater.  So that means that 

24    less than 15 percent of the forms failed to come back.

25               Now, the next category is where between 15 and 20 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 255    

 1    percent of the forms failed to come back.

 2               And we go 20 to 25, 25 to 30, 30 to 35, 35 to 40, 

 3    40 to 45.  And over here then is a category of poststrata 

 4    where more than 45 percent of the forms failed to come back.           

 5    So this is the category where the mailback was successful 

 6    and this is the extreme where the mailback was not 

 7    successful.

 8               The white bars with red on them indicate the 

 9    omission rate.  And these omission rates were estimated 

10    using data provided to me by the Census Bureau as part of my 

11    duties on the Special Advisory Panel.

12               We can see that as more people failed to return 

13    their forms, that the omission rate increases.

14               It goes from about 3 percent at this end to 4.87, 

15    5.92, 8.33, 10.54, 12.96, 14.37 and finally up to 18.68 

16    percent.

17               Now, the rate of erroneous enumeration, the other 

18    kind of census error, is indicated by the solid green bars.           

19    Now, in the lowest category, the rate of erroneous 

20    enumeration was actually greater than the rate of omission.  

21    Here the rate of erroneous enumeration was 4.04 percent.

22               As we go along the access, it actually declined 

23    to 3.65 percent; went back up to 4.99 percent, to 5.33 

24    percent, to 6.24 percent, to 7.14 percent, to 8.07 percent, 

25    and finally to 9.88 percent.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 256    

 1               So as the non-mailback rate went up, so too did 

 2    the rate of erroneous enumeration, though not as rapidly as 

 3    the omission rate.

 4               The blue and white striped bars are the estimated 

 5    net undercount.  Remember, the net undercount, in essence, 

 6    is the difference between the omission rate and the rate of 

 7    erroneous enumeration.  Here, where the rate of erroneous 

 8    enumeration was greater than the rate of omission, there was 

 9    actually a net overcount estimated.  In other words, the 

10    census count there was 1.38 percent to break.  As the 

11    non-mailback rate increases, so too does the net undercount 

12    rate, to 0.67 percent, to 0.37 percent, to 2.48 percent, to 

13    3.50 percent, to 5.11 percent, to 5.68 percent, and finally 

14    to 8.50 percent.

15               So we see that where the mailback rate is high, 

16    the undercount is low.  Where the mailback rate is low, the 

17    undercount is high.

18         Q.    Why is that significant -- is that significant to 

19    you in terms of judging the successor the accuracy of the 

20    PES?

21         A.    Yes, it is.

22         Q.    And why is that?

23         A.    Well, first of all, just a matter of common 

24    sense, we know that the census is hard to take where the 

25    mailback rate is low.  Where the mailback rate is low, we 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205123census      Ericksen -direct                 257    

 1    would expect to have problems.

 2               The mailback rate has nothing to do with the PES, 

 3    so it's an independent source of information about the 

 4    undercount.

 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    
      


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                258    

 1         Q.    And why is it corroboration, in your mind?
                                                           

 2         A.    It's corroboration because we understand why the 

 3    undercount should be high, and it is high where we expect it 

 4    to be high.

 5         Q.    You mean it is estimated to be high by the PES in 
                                                                   

 6    those same areas?

 7         A.    That's right.  

 8               MR. ZIMROTH:  Your Honor, I move the admission of 

 9    494.

10               MR. SITCOV:  No objection.  

11               THE COURT:  494 is admitted.

12               (Plaintiff Exhibit 494 for identification was 

13    received in evidence) 

14         Q.    Independent of the PES, Professor Ericksen, is 

15    anything known about the likelihood of being counted based 

16    on the relationship with the person filling out the form?

17         A.    Yes.  I believe I testified yesterday that people 

18    who are closely related to the person filling out the form 

19    are more likely to be included in the form, people who are 

20    less closely related to the person filling out the form are 

21    less likely to be included.

22         Q.    And I believe yesterday you referred us to two 

23    reports, which are 464 and 465.  I believe they're already 

24    admitted into evidence.  Do you want to take a look at that? 

25               Do you have those in front of you?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                259    

 1         A.    I do.

 2         Q.    Those are some sources for the information that 

 3    you testified yesterday?

 4         A.    Yes.  Those are Census Bureau studies of the 1980 

 5    census and the 1986 Los Angeles census.

 6         Q.    What did the PES find about the same phenomenon? 

 7               MR. ZIMROTH:  And I have put up, your Honor, a 

 8    chart that doesn't have an exhibit number, but it is nothing 

 9    more than a blowup of PX 195, Appendix A, Table 6.  195 is 

10    Professor Ericksen's report already in evidence and this is 

11    one of the charts that was in the report.

12         A.    Okay.  This is my own study of data collected in 

13    the 1990 PES, and the Census Bureau indicated which of the 

14    people on the PES were the person filling out the form, the 

15    spouse of the person or the child of the person.  

16               The omission rate for those people was 6.98 

17    percent.  The omission rate for other relatives, like uncles 

18    and cousins, was 17.88 percent.  The omission rate for 

19    people who were not related was 22.56 percent.  

20               So we see in the 1990 census that people who were 

21    less closely related to the person filling out the form were 

22    much more likely to be left off the form.

23         Q.    As estimated by the PES?

24         A.    As estimated by the PES.

25         Q.    In your opinion, is that, what you have just 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                260    

 1    testified to, corroboration of the accuracy of PES?

 2         A.    Yes, it is, along with all the other research, 

 3    the ethnographic studies, the fact that this would come out 

 4    in the PES is corroboration of the PES, yes.

 5         Q.    Independent of the PES, do we know anything about 

 6    whether or not it's hard to count people in inner city 

 7    ghettos?

 8         A.    Yes, we do.

 9         Q.    What do we know and from what sources independent 

10    of the PES?

11         A.    Well, first of all, we know everything from the 

12    fact that the census is hard to take there, we know that 

13    minority populations are concentrated there, we know that 

14    minority populations have high undercounts.  We know about 

15    the problems of census taking as described by the 

16    ethnographic studies.

17         Q.    What is it that the PES found about whether or 

18    not it's harder to count people in minority areas and inner 

19    cities?

20         A.    Okay.  Working with the same data file --

21         Q.    Before we move along, for Mr. Sitcov's benefit, 

22    this is table -- we put a blowup of People's Exhibit 195, 

23    Table 8.  

24               MR. SITCOV:  Which people?  

25               MR. ZIMROTH:  My old prosecutorial days, I 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                261    

 1    apologize.  I knew it would happen once at least. 

 2               MR.  SITCOV:  Plaintiff's Exhibit 195, from 

 3    Appendix C, Table 8.  Go ahead.

 4         Q.    Professor Ericksen, would you explain that chart 

 5    to Judge McLaughlin.

 6         A.    Okay.  Now here I was working with the same data 

 7    file of the 1,392 post-stratum, and I defined minorities 

 8    were those people in minority post-stratum, and minority 

 9    post-strata were blacks, Hispanics and in some location 

10    Asians.

11               Now, the Census Bureau found that in the original 

12    enumeration 11.3 percent of the counted population were 

13    minorities living in central cities.  21.4 percent of the 

14    population was counted in these minority post-strata, and 

15    78.6 percent of the population was counted elsewhere.

16               The of the 11.3, the 5.2 and the 4.9 are minority 

17    post-strata in central cities, suburbs, that says remainder 

18    of SMSA, remainder of metropolitan areas, and 

19    nonmetropolitan areas --

20         Q.    Excuse me.  The remainder of SMSA's are, in 

21    essence, suburbs?

22         A.    Suburbs, yes.  The 11.3 and 5.2 and 4.9 sums to 

23    to 21.4.  Here we have the same calculations for 

24    non-minorities, 19.1, 39.8 and 19.7 total the 78.6.  The 

25    21.4 and the 78.6 total 100 percent.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                262    

 1               Now, what we're concerned with is the shares of 

 2    population among central city minorities.  11.3 percent of 

 3    the count was of central city minorities.  Looking only at 

 4    omissions, 18.8 percent of all the omissions were of central 

 5    city minorities.  That 18.8 percent is 7.5 percent greater 

 6    than the 11.3 percent.

 7               16.7 percent of all the error enumerations were 

 8    of central city minorities.  The 16.7 percent is 5.4 percent 

 9    greater than the 11.3 percent.

10               The net undercount, you recall, is, in essence, 

11    the difference between the share of omissions and the share 

12    of error enumerations.  When you subtract the error 

13    enumerations from the omissions across the country and get 

14    the net undercounts, the net undercount was 5.3 million 

15    people.

16               32.4 percent of the 5.3 million people were in 

17    the central city minority post-strata.  The 32.4 percent is 

18    21.1 percent greater than the 11.3 percent.

19         Q.    Now, in your opinion, does the information 

20    represented on that chart corroborate the accuracy of the 

21    PES?

22         A.    Yes, it does.

23         Q.    Why is that?

24         A.    Because from all the external evidence, based on 

25    the 1980 post-enumeration program and all the other studies 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                263    

 1    I told you about, we would expect the undercount to be 

 2    greater than it was.

 3         Q.    It was, you mean, as found by the PES?

 4         A.    As found by the PES.

 5         Q.    Does this chart mean that in every central city 

 6    minority post-strata or every central city in the United 

 7    States there is going to be a higher than average net 

 8    undercount?

 9         A.    No, it does not.

10         Q.    What does it mean?

11         A.    It means that if you look at all central cities 

12    in aggregate, the undercount is going to be high.  

13               There are some central cities where the 

14    undercount is not high.  Those may be central cities where 

15    perhaps the census did a very good job.  They could be 

16    central cities where the rate of error enumeration is 

17    unusually high and cancels out the omissions.

18               But in general, and for the very large majority 

19    of central city minority populations, it means that those 

20    were undercounted substantially relative to the rest of the 

21    United States.  

22               MR. ZIMROTH:  Thank your Honor.  

23               Your Honor, would it be possible to have a lunch 

24    break?  This is a convenient time in the testimony.  

25               THE COURT:  Any problems?  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                264    

 1               MR. SITCOV:  No.  I am wondering if Mr. Zimroth 

 2    might know how much longer the direct will be of this 

 3    witness.  

 4               MR. ZIMROTH:  Probably about an hour.  

 5               THE COURT:  All right.  We will resume at a 

 6    quarter of 2:00.

 7               (Luncheon recess)

 8               (Continued on next page) 

 9    

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                265    

 1               AFTERNOON SESSION 

 2                           1:45 pm

 3               (Trial resumes)

 4               (In open court)

 5               THE COURT:  Mr. Zimroth, are we ready to resume?  

 6               MR. ZIMROTH:  Yes, your Honor.

 7               Your Honor, I have one piece of housekeeping 

 8    business, and that was my unfortunate mixup with these 

 9    volumes on Plaintiff's Exhibit 540 B.  Apparently your Honor 

10    had the four correct volumes, the government had the four 

11    correct volumes.  The only one who didn't didn't have the 

12    four correct volumes was the witness, but there are, in 

13    fact, four volumes, 9 through 12, and they're organized in a 

14    way -- should we have have the witness look at them and make 

15    sure that they're there, that they include -- I think he 

16    identified these as materials he had received relevant to 

17    the various piece lists.

18               (Pause)

19               THE WITNESS:  Yes.  

20               MR. ZIMROTH:  And they, like the 540 A series, 

21    are organized with 540 B 1, which is material relevant to 

22    the one study, 540 B 2, et cetera, to the end.  Those are 

23    already in evidence, your Honor.  

24               THE COURT:  Yes.

25    EUGENE P. ERICKSEN, resumes 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                266    

 1    DIRECT EXAMINATION continued 

 2    BY MR. ZIMROTH:  

 3         Q.    Professor Ericksen, yesterday and this morning 

 4    you described to his Honor, Judge McLaughlin, the way in 

 5    which the census was taken in 1990, and I want to ask you 

 6    whether or not those methods that you see on the board that 

 7    you described, 454, are the same methods that have been used 

 8    throughout history?

 9         A.    No, they're not.  The first census was taken in 

10    1790, and as far as I can tell from the descriptions of 

11    enumerators, they typically supplied their own paper.  They 

12    did not actually count individual people.  

13               What was done in 1790 was to go to a house and 

14    find out how many free white males lived there and so forth.  

15    The census didn't actually enumerate individual people until 

16    1850.  Beginning in the 1850 census, individual's names were 

17    written down together with their characteristics.  Although 

18    the form didn't necessarily change, the nature of the census 

19    changed dramatically in 1880, 1890 when the computer was 

20    first introduced.

21               In 1940 was the first census in which sampling 

22    was used.  There has been a trend over the history of the 

23    census in general to reduce the number of questions that are 

24    asked of respondents.  In 1940, with the use of sampling, it 

25    meant that instead of having to get income for every single 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                267    

 1    person, you might just have to get it from a sample, so not 

 2    everybody would get a long form.

 3               In 1960, the Census Bureau introduced the idea 

 4    that individual people would fill out their own census 

 5    forms.  Before 1960 it was as if the whole procedure was 

 6    like nonresponse follow up.  1960 they dropped the forms 

 7    off, people filled them out and mailed them back.

 8               Beginning in 1970 the idea of mailing the forms 

 9    to people and having the people mail them back was 

10    introduced.

11         Q.    Was that, in your opinion, an important change, 

12    the mailout/mailback procedure?

13         A.    Yeah, from the point of view of my understanding 

14    of the survey research science of the census, I considered 

15    the introduction of the mailout/mailback to be perhaps the 

16    most important change in census taking since sampling was 

17    introduced.  

18               The reason that it's important is that for the 

19    first time you have individuals filling out their own 

20    information, and they can do it and they can take the amount 

21    of time they need to do it.  If they need to go consult 

22    their records to find out what their electricity bill was 

23    last year, they can do it.  There are a lot of advantages 

24    connected with having individuals fill out their own census 

25    forms.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                268    

 1         Q.    Are there also certain assumptions concerning the 

 2    mailout/mailback methodology?

 3         A.    Yes.  I believe, as I discussed yesterday, there 

 4    is the assumption that people will be on the address list, 

 5    there is the assumption that people will receive the form, 

 6    open it, fill it out, fill it out correctly, and send it 

 7    back.

 8         Q.    Are those assumptions, in your opinion, 

 9    quote-unquote true?

10         A.    As a statistician, it is not easy to talk about 

11    assumptions being true.  The issue with assumptions is how 

12    well they comport with the real world.

13               Obviously, there are some people who don't do 

14    that.  We have talked about people don't mail back their 

15    forms, we have talked about errors that are made in mailing 

16    back the form.  We are talking about people who are left off 

17    the forms.  

18               The point is that mailout/mailback is an 

19    improvement over the alternative, which is to take the whole 

20    census by the equivalent of nonresponse follow up.

21         Q.    Now, in 1970 were there also important changes, 

22    in your opinion, made in the way the census was taken?

23         A.    Yes, I believe there were.

24         Q.    Can you give us one example of that.

25         A.    Well, one example of that would be the procedure 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                269    

 1    known as the National Vacancy Check.

 2         Q.    Can you tell Judge McLaughlin what the National 

 3    Vacancy Check was.

 4         A.    Yes.  According to reports of the procedure, 

 5    Census Bureau reports of the procedure that I've read, there 

 6    was a realization that the count for 1970 was coming in 

 7    short, and it appears a likely possibility that too many 

 8    houses were being enumerated as vacant.

 9               So Census Bureau statisticians selected a sample 

10    of 13,546 vacant households, and this sample, they 

11    stratified it into 12 stratum, the four regions, and then in 

12    each region there are three groups defined by how the census 

13    was taken in those areas. 

14                In each of these strata, they selected a sample 

15    of vacant houses and went back and reenumerated them.  It 

16    turned out that the rate varied from a high of about 19 

17    percent to a low just below 10 percent, as I recall, but on 

18    average 11 percent of the sample was found to, in fact, have 

19    been occupied.

20         Q.    And so what did the -- well, did the National 

21    Vacancy Check use sampling techniques?

22         A.    Yes, it did.

23         Q.    Did it use stratification?

24         A.    Yes, it did.

25         Q.    Did it use imputation?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                270    

 1         A.    Yes, it did.  The way in which it used imputation 

 2    was that about 11 percent of the sample of 13,500, which 

 3    would be approximately 1500 in the sample, 1500 of the 

 4    vacant houses in the sample were found to be occupied.  On 

 5    the basis of this, the Census Bureau then took a sample of 

 6    all the vacant households in the country and imputed the 

 7    status of "occupancy" to the selected households, and then 

 8    in those households which had been imputed to be occupied, 

 9    they then imputed people into them.

10               What I mean by imputed people is they link the 

11    house which had just been classified as occupied to another 

12    house that was also on the computer tape, and whoever lived 

13    in the other house, that record was duplicated and put into 

14    the house that had just been created to be occupied.  That 

15    is called imputation.  In that process, 1,069,000 people 

16    were added to the census.

17         Q.    I am sorry if I am repeating myself because I 

18    went a little bit out of order, please forgive me.

19               The National Vacancy Check used sampling 

20    techniques?

21         A.    It did.

22         Q.    And it used stratification?

23         A.    It did.

24         Q.    And it used imputation?

25         A.    It did.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                271    

 1         Q.    On the basis of these techniques, was there an 

 2    adjustment made to the 1970 decennial census?

 3         A.    Yes, there was.

 4         Q.    Did the Census Bureau do tests of the accuracy of 

 5    the methods?

 6         A.    Yes, they recorded tests of the accuracy of the 

 7    method.

 8         Q.    In your opinion, were those tests as stringent as 

 9    the tests supplied with the 1990 PES?

10         A.    No.

11         Q.    You mentioned that you had read a report, a 

12    Census Bureau report of this National Vacancy Check.  Will 

13    you look at 474, please.

14               is that the report to which you were referring?

15         A.    It is.  

16               MR. ZIMROTH:  Your Honor, I move its admission.

17               MR. ZIMROTH:  No objection.  

18               THE COURT:  474 is admitted.

19               (Plaintiff Exhibit 474 for identification was 

20    received in evidence) 

21         Q.    In 1970 was there also a procedure called the 

22    Post-Enumeration Post Office Check?

23         A.    Yes, there was.

24         Q.    Can you tell Judge McLaughlin what that procedure 

25    was.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                272    

 1         A.    Yes.  The Census Bureau also believed that a 

 2    number of housing units had been left out in rural areas in 

 3    the south, so in the 16 states that constitute the south in 

 4    the Census Bureau definition, that's the area stretching 

 5    from Delaware down through Texas, in rural areas they asked 

 6    the Post Office to check the Census Bureau addresses against 

 7    their records.

 8               When the Post Office record indicated a house 

 9    that hadn't been included in the census, they checked it.  

10    They did this on a sample basis because they didn't have 

11    time or resources to check every single housing unit that 

12    the Post Office had that the Census Bureau didn't.

13               They estimated that a certain proportion of the 

14    Post Office addresses had not been included, and they added, 

15    through imputation, 484,000 people to the 1970 census based 

16    on this procedure.

17         Q.    Did the Post Office Post-Enumeration Post Office 

18    Check -- 

19               MR. ZIMROTH:  Your Honor, please forgive me, in 

20    the lingo that is called PEPOC.

21         Q.    -- did the 1970 PEPOC use sampling techniques?

22         A.    It did.

23         Q.    Did it use matching techniques?

24         A.    It did.

25         Q.    Did it use imputation?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                273    

 1         A.    Yes, it did.

 2         Q.    On the basis of these procedures, was there an 

 3    adjustment made to the 1970 decennial census?

 4         A.    Yes, there was.  

 5         Q.    Is that procedure described in the same document 

 6    that has just been admitted into evidence?

 7         A.    Yes, it is.  

 8               MR. ZIMROTH:  474.

 9         Q.    Now, in 1980 was there imputation used in the 

10    1980 decennial sentence?

11         A.    Yes, there was.

12         Q.    Will you explain that to the judge.

13         A.    Yeah.  What I am talking about here is the 

14    imputation of whole people.  At the end of the --

15               THE COURT:  Whole people? 

16               THE WITNESS:  Yes, as opposed to the imputation 

17    of characteristics.  

18               THE COURT:  I thought you meant half people? 

19               THE WITNESS:  For the income it might be 

20    two-thirds.  Look, the idea is that in 1980, at the end of 

21    the day the Census Bureau had a list of addresses where they 

22    still didn't know whether the structure was occupied or 

23    vacant.  They knew it existed.  They couldn't ignore the 

24    fact that it existed.

25               So they imputed an occupancy status to those 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                274    

 1    houses, and they did that by linking the undetermined house 

 2    to another house that was located nearby in the computer 

 3    tape, and if the linked house was occupied, then the 

 4    previously undetermined house was deemed to be occupied.  

 5               If the link house was vacant, the previously 

 6    undetermined house was deemed to be vacant.  For those 

 7    previously undetermined houses that were deemed to be 

 8    occupied --

 9         Q.    How did the Census Bureau then determine how many 

10    people to impute into those occupied households?

11         A.    It was again through imputation.  They link the 

12    nearly deemed occupied house to another occupied house and 

13    duplicated the records, so the computer, in essence, created 

14    762,000 people who were included in the 1980 census count.

15         Q.    The 1980 census used imputation methods, did it?

16         A.    Yes, it did.

17         Q.    There was an adjustment in the population based 

18    on those methods?

19         A.    Yes.  

20               MR. ZIMROTH:  I think your Honor can take 

21    judicial notice of the fact that this method in 1980 was 

22    challenged, they changed a seat in the House of 

23    Representatives.  It was challenged in court and upheld.  

24               MR. SITCOV:  If your Honor is going to take 

25    judicial notice of anything, your Honor should take judicial 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                275    

 1    notice of the fact that in the opinion Judge Sprizzo wrote 

 2    in Cuomo v. Baldwin, 674 F. Supp., and I believe it begins 

 3    on 1089, there is a passage in there in which Judge Sprizzo 

 4    recounts virtually the identical testimony from Dr. Ericksen 

 5    that was given in the -- 

 6               MR. ZIMROTH:  Your Honor, this is --

 7               THE COURT:  Let him finish.  

 8               MR. SITCOV:  -- given in the 1980 trial, and 

 9    Judge Sprizzo rejected all of them as being support for 

10    adjustment.  So if you want to take judicial notice of what 

11    Mr. Zimroth wants, you should take judicial notice of that 

12    as well.  

13               THE COURT:  This court, being the direct 

14    successor of Solomon, will take judicial notice of both.

15         Q.    In the 1990 decennial census, did the Census 

16    Bureau -- how did the Census Bureau deal with the question 

17    of federal employees stationed abroad, outside the United 

18    States?

19         A.    As I understand it, the federal employers, the 

20    agencies that employed these people, were asked to send in 

21    the names of these people with their addresses, and the 

22    Census Bureau added about 922,000 overseas people and added 

23    them to the states they were indicated as their home of 

24    record.

25         Q.    Is that different from the way this issue was 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                276    

 1    dealt with in 1980?

 2         A.    It was not done in 1980.

 3         Q.    What was not done in 1980?

 4         A.    The adding of federal employees living overseas.

 5         Q.    And you say that these people were added to the 

 6    states in which was indicated?

 7         A.    That's right.

 8         Q.    Were they added to a particular block?

 9         A.    No.

10         Q.    Of the state?

11         A.    No.

12         Q.    They weren't?

13         A.    No, they weren't.

14         Q.    So if you add up in the 1990 decennial census all 

15    the blocks that were added, in each state will it come to 

16    the total that was reported in each state?

17         A.    It will come to the total that's reported for 

18    each state that adds 248.7 million for the United States.  

19    It will not come to the total of that number plus 922,000 

20    that was used for the apportioning of Congress.

21         Q.    You'll have to explain this.  Two different 

22    numbers were used?

23         A.    Yes.

24         Q.    Could you explain that to the judge.

25         A.    I want to make sure -- (Pause) yeah, I want to 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                277    

 1    make sure I got those numbers right.

 2         Q.    What are you referring to?

 3         A.    I believe that the director of the Census Bureau 

 4    indicates what they are.

 5         Q.    What --

 6         A.    On page 1.

 7         Q.    Of?

 8         A.    Of her report to the Secretary.

 9         Q.    Which is what number?

10         A.    It is Exhibit 55, page 1.

11         Q.    Go ahead.

12         A.    And the 248,709 -- 248,709,873 is the sum of the 

13    state populations, that is called the resident populations.

14         Q.    Is that also the sum of the block populations?

15         A.    It is.

16         Q.    Okay.

17         A.    And if you add the 922,000 overseas military and 

18    federal employees and their dependents, you get the 

19    apportionment population of 249,632,692.

20         Q.    The 922,000 were not put into particular blocks?

21         A.    That is correct.

22         Q.    Professor Ericksen, have you read the decision 

23    document -- will you look at PX 9, please.

24         A.    I have it.

25         Q.    Have you read that?  That is the Secretary of 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                278    

 1    Commerce's decision document?

 2         A.    Yes, it is.

 3         Q.    Have you read that document?

 4         A.    Yes, I have.  

 5               MR. ZIMROTH:  I would like to offer that document 

 6    into evidence at this time.  

 7               MR. SITCOV:  Happy to have it.  

 8               THE COURT:  Received as Exhibit 9.

 9               (Plaintiff Exhibit 9 for identification was 

10    received in evidence) 

11         Q.    Would you look at page -- actually, it is on the 

12    bottom Bates stamped a lot of zeroes and a 3.  It is the 

13    first page of text, I believe -- yes.  Before I ask you 

14    about that particular page, let me ask you, do you believe 

15    that -- you have read this document, have you not?

16         A.    Yes, I have.

17         Q.    Do you believe that the Secretary of Commerce 

18    presents valid arguments for rejecting the Census Bureau's 

19    conclusions?

20         A.    No, I do not.

21         Q.    Can you look at page 3, 003 -- it is an 8.  

22    Sorry.  I want to read you this sentence:  "The 1990 census 

23    is one of the two best censuses ever taken in this country."  

24    That is the first sentence of the third full paragraph.  

25               MR. SITCOV:  It is Bates stamped 8. 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                279    

 1               THE WITNESS:  That is page 8.  

 2               THE COURT:  8?  I am sorry.  3.  

 3               MR. ZIMROTH:  Sorry.  Mine looked like a 3.  

 4               THE COURT:  I've got it.  It's 8.  It does look 

 5    like a 3.  So you're only deaf, not blind.  

 6               MR. ZIMROTH:  I'm deaf and blind, not only half 

 7    deaf.  Let me read it again.

 8         Q.    "The 1990 census is one of the two best censuses 

 9    ever taken in this country."  Do you see that?

10         A.    Yes, I do.

11         Q.    Can you think of any standard by which the 1990 

12    census could be judged one of the two best ever taken?

13         A.    The only standard I can think of would be in 

14    terms of the net undercount.  It is the case that the net 

15    undercount is the second lowest measured from 1940 to 1990.

16         Q.    In your opinion, is that an appropriate test to 

17    judge the the success of the 1990 decennial census?

18         A.    No.  What the census is supposed to do is to 

19    distribute the population among areas.  The relevant 

20    statistic is the differential undercount.  As that chart 

21    over there indicates, if anything, the differential 

22    undercount is higher in 1990 than it has been in any year 

23    since 1940.  

24               One could argue that this census comes in 6th.  

25    It is not really a meaningful thing to say, but certainly it 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 4             Ericksen - direct                280    

 1    is getting no better than it has been since 1940.  To say it 

 2    is the second best census or it is one of the two best 

 3    censuses ever, to me, misunderstands the entire issue, which 

 4    is to reduce and possibly to eliminate the effects of the 

 5    differential undercount.  

 6               MR. ZIMROTH:  I have nothing further, your Honor.  

 7               THE COURT:  Thank you.  Are you ready to go or do 

 8    you want some time?  

 9               MR. SITCOV:  Can we take about five minutes?  

10               THE COURT:  Sure.  We'll take a five minute 

11    recess.

12               (Recess)

13               (Continued on next page) 

14    
      

15    
      

16    
      

17    
      

18    
      

19    
      

20    
      

21    
      

22    
      

23    
      

24    
      

25    
      


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 281    

 1    EUGENE P. ERICKSEN, resume
                                

 2               THE COURT:  Mr. Sitcov.  
                                          

 3               MR. SITCOV:  Good afternoon, your Honor.  

 4    CROSS-EXAMINATION 

 5    BY MR. SITCOV:

 6         Q.    Dr. Ericksen, do you recall that you testified 

 7    this morning that you agreed with a number of the 

 8    conclusions in Dr. Bryant's recommendation on adjustment to 

 9    Secretary Mosbacher?

10         A.    Yes, I do.

11         Q.    And you agree with Dr. Bryant's recommendation 

12    that the 1990 Census should be adjusted using the PES 

13    results, is that right?

14         A.    That's correct.

15         Q.    Could you turn to Plaintiff's Exhibit 55, and 

16    turn to page 3, please.

17               The last sentence on that page says: 

18               "Adjustment is an issue about which reasonable 

19    men and women of the best statisticians and demographers can 

20    disagree."  

21               Do you see that?

22         A.    I see it.

23         Q.    And you agree with that, don't you?

24         A.    As a hypothetical question, I have not yet seen a 

25    reasonable argument against adjustment.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 282    

 1         Q.    Well, I'm not asking you as a hypothetical 

 2    question.  I'm asking:  Do you agree with the statement on 

 3    page 3 that I just read?

 4         A.    Well, I don't really know whether I do or I 

 5    don't.  What I'm telling you is I have not yet seen a 

 6    reasonable argument against adjustment.

 7               THE COURT:  It would seem to suggest you disagree 

 8    with this conclusion.

 9               THE WITNESS:  Yes.

10         Q.    Okay.  You testified as an expert witness for 

11    some of these same plaintiffs in their attempt to force an 

12    adjustment of a 1980 decennial census, didn't you?

13         A.    I did testify in a lawsuit brought by the City of 

14    New York concerning the adjustment of the 1980 census.

15         Q.    Is that a yes to my question?

16         A.    That's an answer.  

17               MR. ZIMROTH:  He testified to what he testified 

18    to.  If you want to characterize the lawsuit anyway you want 

19    to characterize it, characterize it.  Put it in a complaint.

20               THE COURT:  I think all he asked him was did he 

21    testify, and I take it the answer was yes.  

22               MR. ZIMROTH:  Yes.  

23    BY MR. SITCOV:

24         Q.    And did you submit an affidavit on behalf of the 

25    plaintiffs in that case?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 283    

 1         A.    Yes, I did.

 2         Q.    And did you argue in that affidavit that because 

 3    the Census Bureau used limited adjustment techniques in the 

 4    1970 Decennial Census, that adjustments should be 

 5    appropriate for 1980?

 6         A.    I don't think that your question captures the 

 7    logic of that affidavit.  I don't recall everything that was 

 8    in that affidavit.  If you'd like to show me -- it was 

 9    written, what, eight years ago?  12 years ago?  Let me take 

10    a look at it, and that could refresh my memory as to exactly 

11    what was said in.

12         Q.    Okay.  If Judge Sprizzo described your affidavit 

13    as stating -- 

14               MR. ZIMROTH:  Objection, your Honor.

15               THE COURT:  Overruled.

16         Q.    If Judge Sprizzo described your affidavit as 

17    stating that because the Bureau used limited adjustment 

18    techniques in the 1990 Decennial Census, the Bureau should 

19    now be foreclosed from arguing that adjustments suggested by 

20    the plaintiffs for the 1980 Census are technically 

21    unfeasible, would he have incorrectly summarized your 

22    affidavit?  

23               MR. ZIMROTH:  Your Honor, the witness said he 

24    didn't remember precisely what was in his affidavit.  If Mr. 

25    Sitcov wants to give him the affidavit and have him read it 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 284    

 1    and summarize it and characterize it, the witness says he's 

 2    pleased to do so.

 3               THE COURT:  Are you in a position now, without 

 4    looking at the papers, to answer the question whether Judge 

 5    Sprizzo's summary is accurate?

 6               THE WITNESS:  Look, I wrote the affidavit.  I 

 7    gave the reasons explaining why there had been an undercount 

 8    in the 1990 Census.  I talked about methods for adjustment.  

 9    I talked about the fact that in 1970 that a sample had been 

10    taken, and the population had been changed as a result of 

11    inferences made from the sample to a larger population. 

12               That's what I remember.  I don't remember what 

13    Judge Sprizzo said about my affidavit.  What I do remember 

14    about Judge Sprizzo's decision is that he ruled against 

15    adjustment.

16         Q.    Okay.  Do you recall that Judge Sprizzo said that 

17    none of the adjustments in 1970 were even remotely similar 

18    to the type of wholesale adjustments that the plaintiffs 

19    were suggesting in 1980?

20         A.    I don't. 

21               MR. ZIMROTH:  Objection, your Honor.

22               THE COURT:  Overruled.  

23               MR. ZIMROTH:  Why don't we just put the decision 

24    into evidence, your Honor?

25               THE COURT:  Let him do it the way he wants to.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 285    

 1    It's his cross.  

 2               MR. ZIMROTH:  Okay.

 3         A.    Like I said, I don't recall what Judge Sprizzo 

 4    said.  I do recall what his decision was.  That's what I can 

 5    tell you. 

 6         Q.    Okay.  Do you have Plaintiff's Exhibit 195 in 

 7    front of you?  That was this rather interesting tone.

 8               I believe you testified yesterday that you were 

 9    one of the authors of this report?

10         A.    I was.  Or, I did.

11         Q.    You were and you did.

12               THE COURT:  Well, he did and was.  

13               MR. SITCOV:  He did and was.  That's why I'm a 

14    lawyer and not an English teacher.

15         Q.    Your recommendation to the Secretary discusses 

16    smoothing, does it not?

17         A.    I'm sorry?

18         Q.    Your recommendation to the Secretary, that is 

19    Plaintiff's Exhibit 195, discusses smoothing, doesn't it?

20         A.    This is not my recommendation to the Secretary.  

21    This is a report I wrote with three colleagues on the panel 

22    that supported my recommendation.  But this was not my 

23    recommendation.

24         Q.    And this report that supported your 

25    recommendation discusses smoothing, does it not?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 286    

 1         A.    Yes, it does.

 2         Q.    And it discusses loss functions, does it not?

 3         A.    It may mention loss functions, but I want to make 

 4    clear that in writing this report, we did not have access to 

 5    the Census Bureau's conclusions on loss functions for the 

 6    1990 Census.

 7         Q.    You did include in this document discussions of 

 8    loss function analyses, did you not?

 9         A.    Yes.

10         Q.    Okay.  And this report also includes a discussion 

11    of fabrications, does it not?

12         A.    Yes, it does.

13         Q.    And it also includes a discussion of computa- 

14    tions, does it not?

15         A.    It does.

16         Q.    You are familiar with the term standard error, 

17    are you not?

18         A.    I am.

19         Q.    And you don't know whether you rely on estimated 

20    differential undercount that was in magnitude less than 

21    twice the standard error, is that right?

22         A.    I don't understand your question.

23         Q.    Will you rely on an estimated differential 

24    undercount that was a magnitude less than twice its standard 

25    error?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 287    

 1         A.    It depends on the purpose that I wanted to use it 

 2    for.

 3         Q.    Do you know what T statistic is?

 4         A.    Yes, I do.

 5         Q.    And one way of computing a T statistic is to 

 6    divide an estimate by a standard error, isn't that right?

 7         A.    There are estimating situations where you would 

 8    calculate T statistic in that way, yes.

 9         Q.    There is a generally-accepted cutoff for T 

10    statistics in the sense that a T statistic below that level 

11    is in the statistically significant?

12         A.    I would not say there was a generally-accepted 

13    cutoff.  Statisticians use different cutoffs depending on 

14    their purpose.

15         Q.    Now, you not only submitted an affidavit in the 

16    Cuomo case, you actually testified in court on behalf of 

17    plaintiffs in that case, did you not?

18         A.    I did.

19         Q.    And did you not testify in that case that a T 

20    statistic of 2 is about as low as you would want to go?

21         A.    For the purpose that that T statistic was used 

22    for, I did make that testimony.  It does not mean that one 

23    would extrapolate from that one situation that for every 

24    possible situation where a statistician was using numbers 

25    that you would use a T statistic.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 288    

 1         Q.    And that was that exact situation you were 

 2    testifying about was a regression, was it not?

 3         A.    That situation was -- the situation that I was 

 4    testifying about concerned the following question:

 5               When you have a whole set of variables that might 

 6    possibly become independent variables in the regression 

 7    equation, you need a criterion to decide which variable 

 8    should be included.  And the criterion that I used to decide 

 9    which variable should be included was, among other things, 

10    that the regression coefficient should be twice the standard 

11    error.

12         Q.    So, my question was:  Were you saying that the 

13    T statistic shouldn't be lower than 2 in the context of a 

14    regression analysis?

15         A.    What I was testifying to was that when you have 

16    the problem that I faced in the smoothing model that I was 

17    working on of trying to decide which independent variable 

18    should go in that particular regression equation, that is a 

19    criterion that I decided to use.  There are many criteria 

20    that different statisticians will use to solve that 

21    particular problem.  That's one that I happened to choose in 

22    that particular situation.

23         Q.    With a two-tailed test, a T statistic of about 2 

24    corresponds to the 5 percent significance level, is that 

25    right?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 289    

 1         A.    That is correct.

 2         Q.    You are familiar with confidence intervals?

 3         A.    Yes, I am.

 4         Q.    Would it be reasonable for a statistician to use 

 5    the 95 percent confidence level in considering the issue 

 6    assessment adjustment?

 7         A.    It depends on what part of census adjustment you 

 8    are talking about.  I can imagine some situations where it 

 9    might be.

10         Q.    So you don't have an opinion about what is the 

11    most accurate confidence level for determining the accuracy 

12    of adjustments?

13         A.    The concept "accuracy of confidence level" 

14    doesn't really have a concrete definition for a 

15    statistician.  You try to estimate a confidence interval 

16    accurately by calculating the standard error accurately, but 

17    it's not a matter of choosing the accurate confidence level.

18         Q.    Is the 95 percent confidence level a wrong one to 

19    use if you are trying to estimate the net national 

20    undercount?

21         A.    Well, if I was the Census Bureau statistician 

22    presenting the net national undercount calculated from -- I 

23    assume -- well, perhaps I shouldn't quite start answering 

24    your question.  

25               Could I ask for a short question of clarifi- 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 290    

 1    cation?

 2         Q.    Do you not understand my question?

 3         A.    Well, are you talking about the PES?

 4         Q.    If you are trying to -- if you are trying to 

 5    estimate the net national undercount on the basis of the 

 6    PES --

 7         A.    Okay.  That clarifies it.

 8         Q.    Okay.

 9         A.    With all due apologies, could you give me the 

10    question again?

11         Q.    Sure.

12               Are you certain or not whether the 95 percent 

13    confidence level is the appropriate one to use if you are 

14    trying to estimate the net national undercount?

15         A.    Okay.  Now, were I the Census Bureau statistician 

16    that had to make the decision about how to present the net 

17    national undercount, what I would do is present the net 

18    national undercount and its standard error.  I don't think 

19    that I would draw a conclusion about which confidence 

20    interval should be used.

21         Q.    Would it be inappropriate to do so in your 

22    opinion as an expert in statistics?

23         A.    I don't think it would be necessary to do so.  I 

24    think that what one should do is present the estimate with 

25    the standard error.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 291    

 1         Q.    No.  My question is:  Would it be inappropriate 

 2    as a statistical matter in your opinion as an expert in 

 3    statistics to present that confidence interval?

 4         A.    Well, I suppose if you presented a confidence 

 5    interval, another statistician could back out the standard 

 6    error and do whatever she or he wishes to do with it.  So it 

 7    would not be inappropriate.

 8         Q.    You are familiar with the smoothing process that 

 9    was used in the 1990 Decennial Census?

10         A.    I am.

11         Q.    Net process used the regression model, did it 

12    not?

13         A.    Yes, it did.

14         Q.    And the Bureau's regression model required a 

15    selection of variables to include in the regression 

16    equation, is that right?

17         A.    The Census Bureau's production, regression 

18    equation requires that variables be selected?  Yes.

19         Q.    When you testified for the City of New York in 

20    the Cuomo trial, you used a regression equation to smooth 

21    undercount estimates, didn't you?

22         A.    Yes, I did.

23         Q.    And you selected variables for the regression 

24    equation that you believed to be related to the undercounts, 

25    didn't you?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 292    

 1         A.    I think that your question misstates in some part 

 2    what I did.  What I did, I started out with a list of 

 3    variables that I believed to be related to the undercount, 

 4    and then the variable selection procedure that I used helped 

 5    me decide which of those would be best to include in the 

 6    equation.

 7         Q.    Isn't it a fact that you started with eight 

 8    variables that in your mind you considered in advance to be 

 9    associated with the undercount?

10         A.    I started with eight variables that I believed in 

11    advance could be related to the undercount, that is true.

12         Q.    And that's not what the Bureau did in 1980, is 

13    it?

14         A.    They certainly did in the start with the eight 

15    variables that I started with, that's true.

16         Q.    Well, in fact, didn't the Bureau adopt a step- 

17    wise approach in which they started out with 29 variables?

18         A.    I don't recall the exact number that the Bureau 

19    started out with.  They did start out with a number that 

20    could have been about 29.  If you could show me the document 

21    that they produced and that could refresh my memory.

22         Q.    All right.  But it was around 29 something,, 

23    about that size?

24         A.    Look, if somebody said it was -- 

25               MR. ZIMROTH:  Your Honor, the witness said he 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 293    

 1    would like his recollection refreshed.

 2               THE COURT:  Do you wish to pursue it, --

 3               MR. SITCOV:  Sure.

 4               THE COURT:  -- Mr. Sitcov?  

 5               MR. SITCOV:  Sure.

 6               (Pause)

 7    BY MR. SITCOV:

 8         Q.    Do you recall testifying in the following fashion 

 9    on page 2423 of your transcript, dated February 23rd, 1984, 

10    in Cuomo versus ~Baldridge?  

11               MR. ZIMROTH:  Excuse me.  Do you have a copy of 

12    that that I can follow?

13               MR. SITCOV:  If I may.  May I step over?

14               THE COURT:  Yes, sure.  

15               MR. ZIMROTH:  This is 1984 testimony during the 

16    1984 trial?

17               THE COURT:  That's what he said.  

18               MR. SITCOV:  Yes.  

19               MR. ZIMROTH:  What page are you on?  

20               MR. SITCOV:  2423. 

21               MR. ZIMROTH:  Will you show the witness as well?  

22               MR. SITCOV:  I'm going to show him to see if it 

23    refreshes his recollection.  

24         Q.    2423, starting on page 21.  

25               MR. ZIMROTH:  Wait.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 294    

 1               (Pause)

 2         Q.    "I would also like to contrast the approach that 

 3    I took.  I started with the eight variables in my own mind I 

 4    considered in advance were relevant which is very different 

 5    than the step-wise approach where the Census Bureau started 

 6    with 29 variables, many of which have only a tangentially 

 7    seeming effect of the undercount and they toss them all in, 

 8    or by chance you are bound to get some kind of regression if 

 9    you tried that."

10         A.    I misheard you.  I thought you were talking about 

11    the 1990 census.

12         Q.    No, 1980.

13         A.    Okay.  Then the 29 -- I didn't remember that 

14    number either.  But if I said it, I must have had a reason 

15    to say it then.  That sounds about right anyway.

16         Q.    And does the passage I just read for you refresh 

17    your recollection at all about whether or not you were 

18    critical of the Bureau's approach?

19         A.    Yes.  I thought they didn't do as good a job as 

20    they could have in picking the 29 variables they started out 

21    with.

22         Q.    In fact, you testified they just threw 29 

23    variables into the hopper, didn't you?

24         A.    Whatever it said, it said.

25         Q.    In 1990 the Bureau's production regression model 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 295    

 1    started out with more than 48 candidate variables, didn't 

 2    it?

 3         A.    Again, that sounds reasonable.  I don't recall 

 4    what the number was.  I do know it was a substantial number.

 5         Q.    When you come up with a potential candidate for 

 6    an independent variable to use in a regression equation, it 

 7    is generally the case that it has to fit the general pattern 

 8    that applies across the observations, is that right?

 9         A.    (No response) 

10               MR. SITCOV:  Your Honor, the witness is obviously 

11    struggling.  I just wonder, if you don't mind, that I 

12    suggest if he needs the question reread, that he could ask 

13    for it to be reread.

14               THE COURT:  If you are having trouble with the 

15    question, you are entitled to ask that it be reread.

16               THE WITNESS:  It's just that the question is a 

17    little bit long, and I want to make sure I heard it right 

18    before I answer.

19         Q.    Would you like me to repeat it?

20         A.    I would appreciate it.

21         Q.    When you come up with a potential candidate for 

22    an independent variable to use in a regression equation, it 

23    is generally the case, isn't it, that it has to fit the 

24    general pattern that applies across the observations?

25         A.    Okay.  The answer to that question is that if you 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 296    

 1    have -- the situation that you've got is you want to know 

 2    which variable should go into the smoothing model.  

 3               Now, you don't know, in advance, what these 

 4    variables should be.  You have a general idea of the 

 5    variables that might be good to use.  The variables that are 

 6    going to get used will generally be the ones that fit the 

 7    general pattern across the observations.

 8         Q.    So is your answer yes, it would generally be the 

 9    case that it have to fit the general pattern.

10         A.    The answer is that the variables that would be 

11    selected among the candidate variables would usually be the 

12    ones that fit the general pattern across the observations.

13         Q.    And is it also the case that each of the 

14    variables you select for your regression should make sense 

15    individually?

16         A.    They should make sense, certainly in terms of 

17    what we call the zero order correlation, which is the basic 

18    correlation of the variable with the undercount.

19               Now, when you get a lot of the independent 

20    variables in together, all of which are related to the 

21    undercount, you generally get a statistic known as a partial 

22    regression co-efficient, which is a very different kind of 

23    statistic.

24         Q.    So are you saying you do agree or you don't agree 

25    with my question?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 297    

 1         A.    I do agree that a variable that goes into a 

 2    regression equation where the regression equation is used to 

 3    predict the undercount, you should be able to understand the 

 4    zero order relationship, which is the bivariate relation- 

 5    ship.

 6               For example, if you wanted to use the percent 

 7    minority to predict the undercount, then because we know 

 8    that the undercount is higher for minorities, that the 

 9    correlation between the percent minority and the undercount 

10    should be positive.  But that's -- that's what my answer is.

11         Q.    Okay.  And is it also the case that when you are 

12    selecting variables for a regression equation, you look at 

13    the data to see if it's noise or if it's really telling a 

14    story?

15         A.    Your term "look at the data" is a bit imprecise.  

16    I'm not exactly sure what you mean.  

17               You look at the results of your regression 

18    equation to try to make that determination.  The data are 

19    the elements that go into calculate the regression equation.  

20    So the answer to your question, as I've described the 

21    difference between data and the results of the calculation 

22    based on the data, would be no.

23         Q.    Do you recall that I asked you the following 

24    question at your deposition and you gave your response as 

25    follows --


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 298    

 1               MR. ZIMROTH:  Would you tell us the page?  

 2               MR. SITCOV:  Page 43 

 3               MR. ZIMROTH:  Just give us a minute, please.  

 4    BY MR. SITCOV:

 5         Q.    Line 16.

 6               "Q.    Then the next paragraph says" --

 7         A.    Can I read it?  Do I get to look at it?

 8         Q.    Pardon me?

 9         A.    Do I get to look at it?

10         Q.    It would be quicker if I just read it to you.  

11               MR. ZIMROTH:  It might not be quicker because 

12    there's a context here.  

13               MR. SITCOV:  Your Honor, I'm entitled to correct 

14    his recollection.

15               THE COURT:  To prejudice a recollection.  

16               MR. SITCOV:  Absolutely not.  I would never do 

17    such a thing.

18               THE COURT:  I thought you said you were entitled.  

19               MR. SITCOV:  I'm from the government, I --

20               THE COURT:  That's one of the three biggest lies.  

21               MR. SITCOV:  And one we can't tell.

22               THE COURT:  2.  

23               MR. SITCOV:  I can get another copy if you want.

24               THE COURT:  Would you?  

25               MR. SITCOV:  Sure.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 299    

 1               THE COURT:  I think it may speed things up. 

 2               (Pause) 

 3         A.    This is trial transcript.

 4         Q.    Yes.

 5         A.    I thought you said my deposition transcript.  

 6               MR. ZIMROTH:  That's what I thought you said.  

 7               MR. SITCOV:  I'm sorry.  I thought I said trial 

 8    transcript.  Wait, I take that back.  No, that is his 

 9    deposition transcript.

10               THE COURT:  But you handed him the trial 

11    transcript.  He wouldn't expect that.  

12               MR. SITCOV:  I am sorry.

13               THE WITNESS:  Now that I've got it, I'm keeping 

14    it.

15         Q.    Dr. Ericksen, your government wants a refund.

16               There you go, page 43.

17         A.    Thanks.

18         Q.    Line 16:

19               "Q.    Then the next paragraph says:  'So all of 

20    these make sense is a pattern to me and as a statistician, 

21    what I look for in the data to see if it's noise or if it's 

22    really telling a story is whether that is a pattern that I 

23    can see there that makes sense.  And I look at these 

24    regression equations, and to me there is such a pattern.'

25               "Do you agree with that opinion that makes sense?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 300    

 1               "A.    Yes, I agree with that opinion."           

 2               (Pause)

 3         A.    Yes.  In that particular regression equation, 

 4    which had three independent variables in it, it was possible 

 5    to look at the coefficients, and they did make sense, yes.

 6         Q.    Okay.  But is that something that is not 

 7    generally applicable to regressions?

 8         A.    It's applicable to regressions sometimes when the 

 9    number of variables in the equation is small.  When the 

10    number of variables in the regression equation gets into the 

11    double digits or even into 15 or 20 variables, you would not 

12    expect to make a ready interpretation of each coefficient of 

13    each regression equation -- of each regression coefficient.

14         Q.    In your opinion, low education levels are 

15    associated with undercounts, is that right?

16         A.    In my opinion, low education levels are 

17    associated with census error.  In this situation, where 

18    omissions are much more numerous than erroneous 

19    enumerations, I would expect that a lack of education would 

20    be associated with a higher rate of net undercount in a zero 

21    order relationship.

22         Q.    Education was not a variable in the Bureau's 

23    production regression model for 1990, was it?

24         A.    No.  I believe that the variables the Census 

25    Bureau used were better than education.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 301    

 1         Q.    It is your opinion that crime is associated with 

 2    census undercounts, isn't it?

 3         A.    It is certainly my opinion that census under- 

 4    counts are higher in high crime neighborhoods.

 5         Q.    And crime is not a variable in the Bureau's 

 6    regression model for 1990, was it?

 7         A.    It would not have been possible to use crime 

 8    rates for 1392 poststrata, given the poststrata findings.  

 9    So the answer to your question is no.

10         Q.    It is your opinion that drug use is associated 

11    with census undercounts, isn't it?

12         A.    It is my opinion that in areas where drug use is 

13    high, it's difficult to take the census.  I had never 

14    thought about the question as to whether drug users 

15    themselves, above and beyond the fact they may be poor or 

16    they may be minorities, are undercounted in the census.

17         Q.    Would you turn to page 2 of the -- I guess it's 

18    the main report that's Exhibit 195.

19               And you say in the third full paragraph: 

20               "Differential undercount is a result in the way 

21    in which the census is taken.  The census works best with 

22    the typical middle class suburban household with well- 

23    educated parents and children and reliable mail delivery.  

24    The opposite is true where living conditions are different, 

25    especially in poor neighborhoods with high crime rates, 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 302    

 1    rampant drug use and a great deal of irregular housing."  

 2               Do you see that?

 3         A.    I do.  I believe that's what I just told you.  In 

 4    neighborhoods where the rates of drug use are high, I 

 5    believe that the census is harder to take, which is also 

 6    what I believe that sentence say.

 7         Q.    So, in your opinion, is drug use associated with 

 8    census undercount?

 9         A.    Again, areas where drug use are high would be 

10    associated with areas where the net undercount is also high.

11         Q.    Drug use was not a variable in the Bureau's 1990 

12    regression model, was it?

13         A.    Again, I don't know how the Bureau could possibly 

14    have measured drug use at the level of 1392 poststrata.  So 

15    the answer to your question is no.

16         Q.    So it's your opinion that irregular housing is 

17    associated with undercounts in the census, isn't it?

18         A.    Yes, it is.

19         Q.    And irregular housing was not available in the 

20    Bureau's regression model for 1990, was it?

21         A.    No.

22               Let me again make a point clear, that you can 

23    either draw a distinction between considerable in the 

24    regression equation and actually include it in the --

25               MR. SITCOV:  Your Honor, could you instruct him 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 303    

 1    just to answer my question?

 2               THE COURT:  If you can, answer his question with 

 3    a yes or no preferably or at least a short answer.

 4               THE WITNESS:  Well, he's asking an imprecise 

 5    question that I'm trying to make sure that my answer isn't 

 6    misinterpreted because the distinction to be made between 

 7    variables considered for inclusion and the variables 

 8    actually included in the regression equation.

 9         Q.    I'm not asking about things that were considered 

10    for inclusion.  I'm asking only about things that were 

11    included.

12         A.    That actually made it in?

13         Q.    Yes.

14         A.    No.

15         Q.    Okay.  In your opinion, it is impossible to 

16    assess the reliability of a regression model without knowing 

17    the dependent variables it uses, isn't that right?

18         A.    Well, it's impossible to read a regression 

19    equation without knowing what the dependent variables is.

20         Q.    And when I took your deposition about five weeks 

21    ago, you didn't know the dependent variable used in the 1990 

22    production regression model, did you?

23         A.    I knew that the dependent variable was either the 

24    percent undercount or the reciprocal of it.  I don't recall 

25    the exact passage that you are referring to.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 304    

 1         Q.    Did you testify that "at this moment I don't 

 2    recall the exact form of variable"?

 3         A.    Yes.  I didn't know whether it was the adjustment 

 4    factor or the reciprocal of the adjustment factor at that 

 5    moment.

 6         Q.    I believe -- well, I don't know if you were asked 

 7    this or not on direct.

 8               Is it your opinion that the production smoothing 

 9    model is reliable enough to use for adjusting the 1990 

10    Decennial Census?

11         A.    It certainly is reliable to be a part of the 

12    process, yes.

13         Q.    The smoothing process that was employed in 

14    preparing the adjusted counts for the 1990 Decennial Census 

15    involved two steps, didn't it?  

16               MR. ZIMROTH:  Involved -- excuse me, what?

17         Q.    Two steps.

18         A.    I missed the first half of your question.

19         Q.    Were there two steps in the smoothing process for 

20    the 1990 Decennial Census?

21         A.    If you consider the presmoothing on the variance 

22    to be one step and the calculation of the regression 

23    equation to be the second step, we could think of that as 

24    two steps.

25         Q.    Okay.  Now, I want you to exclude the 1990 PES 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 305    

 1    from your consideration in this question.

 2               None of the other surveys done by the Census 

 3    Bureau used smoothing models of the type used for the 1990 

 4    estimate, isn't that right?

 5         A.    I can't tell you what the Census Bureau does in 

 6    all its surveys.

 7         Q.    You don't know of any?

 8         A.    No, I don't know of any of the Census Bureau 

 9    regular surveys that do that.

10         Q.    And you are an expert on sample surveys, aren't 

11    you?

12         A.    Sure, the Census Bureau has used smoothing.  When 

13    they used smoothing in this exact form I cannot answer.

14         Q.    And don't you know of any other Census Bureau's 

15    surveys in which variances are presmoothed?

16         A.    You mean regular continuing surveys?

17         Q.    Yes.

18         A.    No.

19         Q.    And I know you are familiar with the so-called 

20    P studies because you identified them this morning, is that 

21    right?

22         A.    Yes, I know the P studies.

23         Q.    The Bureau did not smooth and presmooth the data 

24    in the P studies, did it?

25         A.    No.  That's not what smoothing and presmoothing 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 306    

 1    is intended for.

 2         Q.    And you and Drs. Estrada, Tukey and Meier 

 3    submitted a report to the Secretary advocating an adjustment 

 4    to the 1990 Decennial Census, did you?

 5         A.    We submitted a report which submitted our 

 6    individual recommendations.  And on individual 

 7    recommendations, we all four recommended adjustment.

 8         Q.    You assisted in the preparation of Appendix A of 

 9    the report, didn't you?

10         A.    Oh, I would say I did more than assisted in it.

11         Q.    Okay.  If you turn to Appendix A, Plaintiff's 

12    Exhibit 195.  And I'd like you to turn in particular to the 

13    bottom of page 51, to the paragraph that carries over to the 

14    top of 52.  There you were presenting some statistics from 

15    the New York City Department of Planning, is that right?

16         A.    That's correct.

17         Q.    These data were not smoothed or presmoothed, were 

18    they?

19         A.    You wouldn't use smoothing and presmoothing if 

20    you were presenting one estimate for the totality.  

21    Smoothing and presmoothing is a method that you use when you 

22    want to compute local estimates; say, on individual blocks 

23    or census tracts.  That's not what I was doing here, so, 

24    naturally, I didn't smooth or presmooth the estimates.

25         Q.    I'm not asking if you smoothed or presmoothed 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 307    

 1    them.  Did the City of New York smooth or presmooth them?

 2         A.    No, neither did they.

 3         Q.    Okay.  And you presented a 95 percent confidence 

 4    interval there, didn't you?

 5         A.    Let me check.

 6         Q.    Try page 52, the third line from the bottom.

 7         A.    Yes, I used 95 then.

 8         Q.    And that's because that's the standard confidence 

 9    level used in the field statistics, isn't it?

10         A.    That's the confidence interval that I decided to 

11    use when I helped to write this report.

12         Q.    In 1985 and 1989, you published papers in the 

13    Journal of American Statistical Association advocating 

14    adjustment of the 1980, census, is that right?

15         A.    I certainly published two papers on adjustment in 

16    those years in the Journal of American Statistical 

17    Association along with coauthors.

18         Q.    And you advocated adjustment of the 1980 census, 

19    didn't you?

20         A.    Certainly argued that adjustments of the 1980 

21    census would have improved the data, yes.

22         Q.    And you didn't advocate presmoothing in those 

23    papers, did you?

24         A.    No.  For the kinds of data that we were working 

25    with, in the research that we did underlying that we 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 308    

 1    considered presmoothing, it didn't seem necessary for those 

 2    data, for that particular estimating situation, so the 

 3    answer to your question is no.

 4         Q.    At the time I took your deposition five weeks ago 

 5    you didn't know the assumptions in the smoothing model, did 

 6    you?

 7         A.    I did not have them committed to memory.

 8         Q.    Well, in fact, you didn't know whether you knew 

 9    them at the time you wrote the report to the Secretary 

10    urging him to adjust the census on the basis of the amount, 

11    isn't that right?

12         A.    What I believe I told you at the deposition was 

13    that in our discussions with the Census Bureau statisticians 

14    we had discussed a variety of smoothing models.  I knew in 

15    general what all the different models they were considering 

16    were.  At this time I did not know the exact ones they had 

17    selected.

18         Q.    So is the answer to my question yes, that you 

19    didn't know whether you knew the assumptions in the 

20    smoothing model at the time you wrote your report to the 

21    Secretary?

22         A.    Yes, I didn't know at the time of the deposition, 

23    that's --

24               MR. ZIMROTH:  Programs we can look at the 

25    deposition.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 309    

 1               THE COURT:  He has that.

 2         Q.    Why don't you turn to page 103, Mr. Ericksen.

 3         A.    (Complies)

 4         Q.    And you'll see that I asked you, on line 14:

 5               "Q.    Do you know the assumptions that are 

 6    included in the 1990 production smoothing model?

 7               "A.    As we sit here, I could not list them all.

 8               "Q.    Did you know them at the time you wrote 

 9    your report to the Secretary?

10               "A.    I don't recall."  

11               Does that refresh your recollection that you 

12    didn't know whether you knew the assumptions in the 

13    smoothing model at the time you wrote your report to the 

14    Secretary?

15         A.    Yes.  I don't know whether I knew the exact model 

16    on June 21st that they had decided to use that, that's 

17    right.

18         Q.    The major goal of smoothing the adjustment 

19    factors was removing sample error in the adjustment factor, 

20    is that right?

21         A.    No.  The major goal was to reduce the effect of 

22    sampling error.

23         Q.    The major goal of presmoothing the variances of 

24    the adjustment factors was the reduction of sampling errors 

25    in the variances themselves, isn't that right?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 310    

 1         A.    That's correct.

 2         Q.    And in your opinion, both smoothing and 

 3    presmoothing were successful, correct?

 4         A.    Yes.

 5         Q.    So the difference between the raw and the 

 6    presmooth variances is due to nothing but sampling error, is 

 7    that right?

 8         A.    The difference between the raw and the presmooth 

 9    variances is due to nothing but sampling error.

10               The presmooth variances would certainly have 

11    smaller standard errors you would expect than the raw 

12    variances, now, due to the fact that you've used the model 

13    to presmooth the variances that could introduce some error.  

14    But in general you would expect that the errors of the 

15    presmooth variances would be less than the errors of the raw 

16    variances.  And to make sure that you know what I'm talking 

17    about:

18               The variances of the 1392 poststrata are 

19    themselves measured with were based on sample, and they have 

20    variances.  I'm talking about reducing the variances of 

21    those variances.

22         Q.    So is your testimony today that there is 

23    something other than sampling error that is the difference 

24    between the raw and the presmooth variances?

25         A.    Yes.  Because you use a model to presmooth the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 311    

 1    variances, you would expect some error in the smooth 

 2    variances to be introduced by use of the model.

 3         Q.    So some of the difference represents bias?

 4         A.    Nonsampling error, yes.

 5         Q.    Now, let's focus on the adjustment factors 

 6    themselves.

 7               Is the difference between the smooth and the raw 

 8    adjustment factor simply sampling error?

 9         A.    That's not a question that seems sensible to me 

10    because that suggests that if you subtract the smooth from 

11    the raw adjustment factor, the difference would be sampling 

12    error.

13               That's confusing.

14         Q.    Well, didn't you say the purpose of the smoothing 

15    was to reduce the sampling error?

16         A.    The standard errors of the smooth adjustment 

17    factors would in general be less than the standard errors of 

18    the raw adjustment factors.  That's why you do the 

19    smoothing.

20         Q.    Well, is the smoothing removing something other 

21    than sampling error?

22         A.    As I explained to you before, it seems reasonable 

23    to say that the smoothing procedure would introduce a small 

24    amount of nonsampling error.

25         Q.    So, in other words, some of the differences 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 312    

 1    between the raw adjustment factors and the smooth adjustment 

 2    factors represents the bias, is that right?

 3         A.    Well, I would say that the difference would be 

 4    mainly due to the reduced effect of sampling errors, the 

 5    difference would be partly due to nonsampling error.

 6         Q.    That's bias.

 7         A.    That's one way of stating that, yes.  Bias is a 

 8    kind of nonsampling error.

 9         Q.    Right.  

10               MR. SITCOV:  I am not sure, your Honor, exactly 

11    what number of the hoards of plaintiffs' exhibits I'm about 

12    to duplicate.  But I am going to show Dr. Ericksen what 

13    is -- 

14               MR. ZIMROTH:  Are you finished with the 

15    deposition now?  

16               MR. SITCOV:  Pardon me?  

17               MR. ZIMROTH:  Are we finished with the 

18    deposition?  

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

20               It's Defendant's Exhibit 104.  You have it in 

21    front of you.  I gave your clerk some exhibits, and I also 

22    gave some to the witness.

23               THE WITNESS:  I have two volumes that are both 

24    labeled 101 through 103.

25               THE LAW CLERK:  I have two that are 104 --


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 313    

 1               MR. SITCOV:  Now we know the answer.  

 2               I'll just refer to it as my document, your Honor, 

 3    but it's already in evidence.  I don't want to duplicate it.  

 4               MR. ZIMROTH:  Are all these documents in 

 5    evidence?  

 6               MR. SITCOV:  A good number of them are.  Since we 

 7    didn't get your exhibits until after we had sent ours in, it 

 8    was difficult for us to know what you would have.

 9               THE COURT:  As I understand it, your documents 

10    are, by and large, duplicates of the plaintiffs' documents 

11    that are already in evidence.  

12               MR. SITCOV:  A number of them are.  I wouldn't 

13    say -- "by and large" is outside the bounds of the sampling 

14    error, your Honor.

15               THE COURT:  Can we smooth it?  

16               MR. SITCOV:  I think some people think you can 

17    smooth anything.

18               THE COURT:  Does anybody at plaintiffs' table 

19    have a complete list of what they put in? 

20               MR. ZIMROTH:  What we have put in?

21               THE COURT:  Yes, what you have put in.  

22               MR. ZIMROTH:  We have a list -- I think the 

23    answer is yes.

24               THE COURT:  All right.  May I ask whoever that 

25    wonderful person is to keep an eye on her documents and her 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 314    

 1    list and let us know when he is duplicating one of your 

 2    documents.

 3               MS. KRUEGER:  Yes.  

 4               MR. SITCOV:  I'll be able to tell when I'm 

 5    duplicating one.  I just won't know the number.

 6               THE COURT:  Okay.  That makes your job easier.  

 7    BY MR. SITCOV:

 8         Q.    Could you turn to Defendants' Exhibit 104, which 

 9    is the June 13th, 1991, press release.

10               And I'd like you to look at Table 1.

11               It is the fifth page of text.  And it is entitled 

12    "An Illustrative Set of Alternative PES Estimates of 

13    Under/Overcount Rates by State." 

14               MS. KRUEGER:  476 A.  

15               THE COURT:  Plaintiffs' --

16               MS. KRUEGER:  478.

17               THE COURT:  Plaintiffs' 478.  

18               MR. SITCOV:  Your Honor, I would be happy to do 

19    it under the table with plaintiff's counsel, and we can, 

20    when we take breaks, work it out.

21               THE COURT:  It's better to put right in the 

22    record now.  It will read better a year from now.  

23               MR. SITCOV:  I hope when reading in --

24               THE COURT:  I'm in no great hurry.  

25    BY MR. SITCOV:


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 315    

 1         Q.    Do you have Table 1 in front of you there, Dr. 

 2    Ericksen?

 3         A.    In the "Illustrative Set of PES Estimates of 

 4    Under/overcounts Rates by State"?

 5         Q.    Yes.

 6         A.    Yes, I do.

 7         Q.    You see there is a title "Preliminary PES"?

 8         A.    Yes.

 9         Q.    And the preliminary PES is the raw dual system 

10    estimates without smoothing or presmoothing, right?

11         A.    That's right.

12         Q.    And do you see one that says "Selected Method"?

13         A.    I see it.

14         Q.    And the selected method is with smoothing and 

15    presmoothing, is that right?

16         A.    Yes, it is.

17         Q.    And in the first row in both of those columns 

18    they gave the same estimated national undercount rate of 2.1 

19    percent, is that right?

20         A.    Yes.  They don't give the same estimate of 

21    populations, but the two different estimated populations, 

22    both come out to 2.1 percent.

23         Q.    And that's due to benchmarking, isn't it?

24         A.    As I sit here today, I don't recall that fact for 

25    sure.  That may be true.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205125census      Ericksen - cross                 316    

 1         Q.    Okay.  I would like you to take a look at Exhibit 

 2    -- Defendant's Exhibit 105.  

 3               MR. SITCOV:  That has not been duplicated, I 

 4    don't think, your Honor.

 5               THE COURT:  Defendant's 105.

 6         Q.    Do you have it?

 7         A.    Yes.

 8         Q.    This is a letter dated July 11th, 1991, to former 

 9    Secretary Mosbacher, and it is four pages long, and it is 

10    signed by Eugene Ericksen and John Tukey.  Do you see that?

11         A.    Yes, I see it.

12         Q.    And you are the Eugene Ericksen that signed that?

13         A.    Sure am. 

14               MS. KRUEGER:  That is in evidence.  It's PX 142.

15               THE COURT:  Plaintiff's 142 is the same as 

16    defendant's 105. 

17               (Pause)

18               (Continued on next page) 

19    
      

20    
      

21    
      

22    
      

23    
      

24    
      

25    
      


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 317    

 1               MR. SITCOV:  Shall I continue, your Honor?  
                                                             

 2               THE COURT:  Yes.
                                 

 3         Q.    You were the principal author of this letter, 

 4    weren't you?

 5         A.    I don't know what you mean by principal author.  

 6    I wrote the first draft of it.  Professor Tukey and I 

 7    discussed it, and ample revisions were made.  That is what I 

 8    recall.

 9         Q.    Did you do the most work on this letter?

10         A.    I didn't keep time records.  It is probably 

11    reasonable that I spent more time on it than John did.  That 

12    is fair enough.

13         Q.    Do you recall testifying at your deposition on 

14    page 110 that you did most of the work on this letter?

15         A.    I don't recall it, but if it is in there, it is 

16    in there.

17         Q.    Would you turn to page 2, please, of the letter. 

18               According to this letter, if you smooth the raw 

19    adjustment factors without presmoothing their variances, the 

20    estimated net national undercount rate is 1.15 percent.  Is 

21    that right?

22         A.    That's the result that came out of such a 

23    calculation, yes.

24         Q.    If you smooth and presmooth, the estimated net 

25    national undercount rate is 2.1 percent.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 318    

 1               Is that right?

 2         A.    That's right.

 3         Q.    In other words, presmoothing before smoothing 

 4    makes the estimated undercount rate go up from 1.15 percent 

 5    to 2.1 percent.  Is that right?

 6         A.    No.  What presmoothing with smoothing does is it 

 7    retains the net national undercount of 2.1 percent.  It's 

 8    for rather particular reasons.  If you would like me to, I 

 9    would be happy to explain it if we have to get into it.  

10               It is necessary to do the presmoothing to prevent 

11    the smoothing model from extorting the results at the 

12    national level, and this is an issue that the Census Bureau 

13    was concerned about for at least well over a year.  We 

14    discussed it many times at panel meetings, and as a result 

15    of our discussions and their considerations, this is the 

16    decision they came to.

17         Q.    My question is, if you smooth without 

18    presmoothing, the undercount rate is 1.15 percent, is it 

19    not?

20         A.    Yes.

21         Q.    If you add presmoothing, the undercount, the net 

22    undercount, is 2.1 percent.  Is that right?

23         A.    That's a factual question, so the answer is yes.

24         Q.    When I took your deposition five weeks ago, you 

25    didn't know whether that difference was statistically 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 319    

 1    significant, did you?

 2         A.    I had not conducted a test of statistical 

 3    significance of that difference, that is true.

 4         Q.    So is the answer is you didn't know if the 

 5    difference was statistically significant?

 6         A.    That's right.

 7         Q.    You didn't know why the estimated undercount rate 

 8    is increased by presmoothing before smoothing, didn't you?

 9         A.    I had not studied in great detail why the absence 

10    of presmoothing reduced the undercount to 1.15 percent, 

11    that's true.

12         Q.    The answer is you didn't know why it happened. 

13               Is that right?

14         A.    That's right.

15         Q.    Nonetheless, you considered that increase from 

16    1.15 percent to 2.1 percent to be an argument in favor of 

17    presmoothing, didn't you?

18         A.    Oh, yes.

19         Q.    Even though you didn't know why presmoothing had 

20    the effect that it had?

21         A.    I would say I didn't know why the absence of 

22    presmoothing had had the effect that it had.

23         Q.    On the other hand, if the effect of presmoothing 

24    before smoothing had been to cut the undercount rate from 

25    1.15 to 0.58, you weren't certain when I took your 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 320    

 1    deposition whether you would recommend in favor of 

 2    presmoothing.  Isn't that right?

 3         A.    That was a long question.  I really don't know 

 4    what you're talking about.  I need to look at my deposition 

 5    to understand what was going on at that time.

 6         Q.    When I took your deposition five weeks ago, isn't 

 7    it a fact that your testimony was that if presmoothing 

 8    before smoothing had caused the undercount rate to be cut in 

 9    half, you weren't certain whether you would recommend in 

10    favor of presmoothing?  

11               MR. ZIMROTH:  Do you want to show him the place 

12    in the deposition.  Perhaps he can look at it.

13         Q.    Do you need to see the deposition to answer that 

14    question?

15         A.    I do.  

16               MR. ZIMROTH:  That is what he said before.

17         Q.    Why don't you look at page 137.  

18               MR. ZIMROTH:  1 what?  

19               MR. SITCOV:  137.  

20               MR. ZIMROTH:  Thank you.

21         Q.    Do you recall that I asked you the following 

22    question:  

23               "Q.    If presmoothing had caused the total 

24    undercount to go from 1, the total net undercount to go from 

25    1.15 percent to .58 percent, would you have recommended in 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 321    

 1    favor of the use of presmoothing?  

 2               "MR. COREWIN:  Are you asking this as just a 

 3    hypothetical?  

 4               "MR. SITCOV:  Yes.  

 5               "THE WITNESS: I don't know."

 6               Does that refresh your recollection that when I 

 7    took your deposition five weeks ago, you didn't know whether 

 8    you would recommend in favor of presmoothing if its effect 

 9    had been to cut the undercount rate from 1.15 to 0.58 

10    percent?  

11         A.    Yeah.  As I indicated to you, I need to know the 

12    reasons why that happened.

13         Q.    I see.  So before you advocate presmoothing, you 

14    didn't need to know why it caused an increase in the net 

15    national undercount, but if it caused a decrease in that 

16    estimate, you needed to know why?

17         A.    That is not what I am telling you at all.  

18               What I am telling you is that the issue that was 

19    presented to us at a meeting of the special advisory panel.  

20    When statisticians from the Census Bureau came to us, it was 

21    important to them that the smoothing procedure that they 

22    used would not distort the major patterns in the data.  

23               So from the point of view of the Census Bureau, 

24    had the net national undercount been reduced from 2.1 to 

25    1.15 percent, for example, they would have considered that 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 322    

 1    something was wrong with the smoothing model.  I happen to 

 2    agree with that particular opinion.

 3         Q.    My question is, when I took your deposition five 

 4    weeks ago, isn't it a fact that you testified that you would 

 5    use these presmoothing and smoothing if it caused an 

 6    increase in the net national undercount even if you didn't 

 7    know why, but if it caused a decrease in the estimate, you 

 8    would need to know why it caused that decrease?

 9         A.    No, I don't think that was my testimony.  My 

10    testimony is and --

11         Q.    No.  I am asking about your testimony at your 

12    deposition.

13         A.    No, I don't think that's correct.

14         Q.    The smoothing process was done separately for 

15    each of the four census regions.  Is that right?

16         A.    That's right.

17         Q.    One of those regions was the northeast region. 

18               Is that right?

19         A.    That's right.

20         Q.    New York State and New York City are both in the 

21    northeast region, aren't they?

22         A.    Yes, they are.

23         Q.    And there were a total of 1,392 post-strata 

24    nationwide, correct?

25         A.    There were.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 323    

 1         Q.    300 of them were in the northeast?

 2         A.    Somewhere in the range of 300 to 350 is my 

 3    recollection of that.

 4         Q.    It is the case, isn't it, that in order to 

 5    calculate the smooth adjustment factor for any particular 

 6    post-stratum in the northeast, the Census Bureau had to 

 7    utilize data from other post-strata in the northeast region?

 8         A.    Well, that is the whole idea of the smoothing 

 9    model, yes.

10         Q.    In fact, the smooth adjustment factor for each 

11    post-stratum in the northeast depends on the raw adjustment 

12    factors of each of the other 299 post-strata in the 

13    northeast.  Isn't that right?

14         A.    The smooth adjustment factor is an average of a 

15    regression estimate and the original raw adjustment factor 

16    for that area.  So if you're talking about the regression 

17    component of the smooth adjustment factor, the answer to 

18    your question is yes.

19         Q.    So the answer is that the smooth adjustment 

20    factor for each post-stratum in the northeast depends on the 

21    raw adjustment factors of each of the other 299 

22    post-stratum?

23         A.    In part, yes.

24         Q.    Now, I'd like you to consider any particular 

25    post-stratum in the northeast, and you don't have to tell me 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 324    

 1    which one it is, I don't care, but for convenience we'll 

 2    just call it Post-Stratum 1.  It doesn't have to be in the 

 3    first post-stratum in the series of 300, it can be any one. 

 4               Do you follow me?

 5         A.    Yes, I am.  

 6               THE COURT:  Pick a card out of the deck.  

 7               MR. SITCOV:  It would be not quite six decks.

 8         Q.    It is possible, isn't it, that in the process of 

 9    smoothing what we're calling Post-Stratum 1, the data from 

10    other post-strata in the northeast region could have a 

11    larger impact on Post-Stratum 1 than Post-Stratum 1 had on 

12    itself?

13         A.    That's a theoretical possibility which should 

14    always need to be considered.  I believe that the only time 

15    that that would occur would be for the regression component 

16    of the smoothing, of the smoothed estimates.  So for that 

17    part of it, the answer to your question is possibly, yes.

18         Q.    You consider that likely?

19         A.    For the one that I picked?

20         Q.    For any one?

21         A.    Well, it's quite possible that there would be one 

22    out there, or there might be a few that that might happen.

23         Q.    Do you consider it likely that that was a 

24    commonplace happening?

25         A.    It depends on what you mean by "commonplace."


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 325    

 1         Q.    In your opinion, is it likely it happened 15 

 2    percent of the time?

 3         A.    I really don't know.

 4         Q.    25 percent?

 5         A.    I don't know.

 6         Q.    50 percent?

 7         A.    I don't know.

 8         Q.    75 percent?

 9         A.    I don't know.

10         Q.    80 percent?

11         A.    I don't have a way of thinking about the 

12    percentage of time that happened.  I've not really 

13    considered that issue specifically so I wouldn't want to 

14    give any kind of quantification of that.  

15         Q.    If that happened 80 percent of the time, would 

16    that cause you to question the reliability of the smoothing 

17    process?

18         A.    I would want to consider that fact along with 

19    everything else that went into the smoothing process in the 

20    production of adjusted estimates.

21         Q.    Just that one fact alone wouldn't be enough to 

22    make you question the validity of the smoothing process?

23         A.    I would want to consider it in context with all 

24    the other errors and features of the smoothing process as 

25    well as the entire adjustment process, so I would consider 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 326    

 1    it, yes.

 2         Q.    So you would consider it?

 3         A.    Oh, sure.

 4         Q.    That would be a problem?

 5         A.    It is something I would want to consider in 

 6    context with everything else, yes.

 7         Q.    But you never even considered that possibility 

 8    before your deposition, did you?

 9         A.    That's really hard to say, but I don't consider 

10    that to be a particularly likely event or serious problem.

11         Q.    Dr. Ericksen, would you turn to Defendant's 

12    Exhibit 66.  

13               MR. SITCOV:  This has not been introduced, your 

14    Honor, I can assure you this has not been introduced by the 

15    plaintiffs.  

16               THE COURT:  This is conclusively a defendant's 

17    exhibit.  

18               MR. SITCOV:  Yes.  No one else would dare use it.

19               (Pause) 

20         Q.    Do you have Defendant's Exhibit 66?  

21               MR. ZIMROTH:  That doesn't look to me like we 

22    have it.  Where is it?  

23               MR. SITCOV:  Can you angle that, Jackie.  It is 

24    an attachment to -- it is Attachment 2 to that, your Honor.  

25               THE COURT:  Attachment to 2 to --


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 327    

 1               MR. SITCOV:  Sorry, it is Enclosure 3.  I take it 

 2    back.  It begins description of smoothing adjustment 

 3    factors.  That is just a blowup of one page.  

 4               THE COURT:  I have it.

 5               MR. ZIMROTH:  I just want to note for the record 

 6    I am not objecting to this, but just for symmetry's sake, I 

 7    should say we just got this about 20 minutes ago.  

 8               MR. SITCOV:  You just got what?  

 9               MR. ZIMROTH:  This exhibit.  

10               MR. SITCOV:  These are cross-examination 

11    exhibits, and you got that one when you got our original 

12    exhibits because that is an exhibit we will be introducing 

13    through one of our witnesses, so you actually got that about 

14    10 days ago.  

15               MR. ZIMROTH:  Okay.  

16               THE COURT:  But he looked at it 20 minutes ago.  

17               MR. SITCOV:  I said it just for symmetry, your 

18    Honor.  

19               MR. ZIMROTH:  The same is true of the exhibits we 

20    gave to him.  

21               THE COURT:  Go ahead.  

22               MR. SITCOV:  I ask your indulgence just for a 

23    second, your Honor.  I have too many papers here.  

24               THE COURT:  Sure.

25               (Pause) 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 328    

 1         Q.    Are you looking at the document entitled 

 2    description of smoothing adjustment factors in the 1990 

 3    post-enumeration survey, Dr. Ericksen?

 4         A.    Yes, I am.

 5         Q.    Would you turn to page 7, please.

 6         A.    Okay.

 7         Q.    Do you recall that I showed you this document at 

 8    your deposition?

 9         A.    I do recall it.

10         Q.    When I showed you this document during your 

11    deposition, you testified that you didn't know if you had 

12    ever seen it before.  Is that right?

13         A.    That's right.

14         Q.    When I showed you that -- when I showed you page 

15    7 at your deposition, you didn't know of your own knowledge 

16    whether or not it stated the assumptions in the 1990 

17    production smoothing model, did you?

18         A.    What I told you was as consistent with the 

19    question you asked before, I did not know the exact model 

20    that they had selected, so the answer is yes.

21         Q.    But you do know that this is the smoothing model 

22    that was selected, don't you?

23         A.    Yes, I do.

24         Q.    There is no P study that describes the 

25    assumptions in the 1990 production smoothing model, is 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 329    

 1    there?

 2         A.    That's right.

 3         Q.    Those P studies were that vast array of materials 

 4    that are sitting in the one, two, three, four, five, six, 

 5    seven, eight fat binders in front of the judge.  Isn't that 

 6    right?

 7         A.    I think that's what they are.

 8         Q.    And the joint report that you and Doctors Estrada 

 9    and Tukey and Wolter submitted to the secretary secretary 

10    did not state those assumptions, either, did it?

11         A.    No, we did not state these assumptions.  

12               MR. SITCOV:  For the record, your Honor, that 

13    report is Plaintiff's Exhibit 195.

14         Q.    Now, I would like you to look at the paragraph 

15    that begins at the 5th line from the bottom on page 7.  

16    You'll see that there is an equation in the fourth line from 

17    the bottom that says capital Y equals small Y plus E.  Do 

18    you see that?

19         A.    I do.

20         Q.    The capital Y represents the raw adjustment 

21    factor as we have been talking about.  Is that right?

22         A.    That's right.

23         Q.    And little Y represents the true adjustment 

24    factor.  Is that right?

25         A.    That's right.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 330    

 1         Q.    And E represents sampling error.  Isn't that 

 2    right?

 3         A.    That's right.

 4         Q.    Immediately to the right of that equation there 

 5    is an assumption.  Is that right?

 6         A.    Yes.

 7         Q.    Now, that assumption reads, E, squiggly line, N 

 8    O, V sub 3, right?

 9         A.    That's what it reads.

10         Q.    V 3 or V sub three in that assumption is the 

11    presmooth covariance matrix for the raw adjustment factors, 

12    isn't it?  

13               THE COURT:  Of course.  

14               MR. SITCOV:  Your Honor, I neglected to make it 

15    clear you had written this document.

16         A.    It says let V 3 denote the smooth estimated 

17    sampling for variance matrix, yes.  

18         Q.    That assumption entails that the dual system 

19    estimate is unbiased.  Isn't that right?

20         A.    That's what this assumption states.

21         Q.    And you don't know whether that assumption is 

22    true or false, do you?

23         A.    I don't think in terms of assumptions being true 

24    or false.  Assumptions are almost never completely perfectly 

25    correct.  The issue is how far from truth in the real world 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 331    

 1    the data are relative to the assumption, but assumptions are 

 2    almost never perfect.  

 3         Q.    This assumption is false?

 4         A.    Like I said before, you don't really think in 

 5    terms of assumptions being true or false; you think in terms 

 6    of how close the data are to being true.

 7         Q.    Well, is the dual system estimate unbiased?

 8         A.    The dual system estimate is not perfectly 

 9    unbiased, no.

10         Q.    Well, the Census Bureau reported a figure for the 

11    amount of the bias in the dual system estimate, didn't it?

12         A.    Yes, it did.

13         Q.    The bias in the dual system estimate is reported 

14    in the P 16 report, isn't it?

15         A.    Yes, it is.

16               MR. SITCOV:  Your Honor, this is a document as 

17    well that plaintiffs have introduced, and it is in that 

18    large volume of binders in front of you, and this is called 

19    total error in post-enumeration survey estimates in 

20    evaluation of post-strata.  Perhaps Mr. Zimroth can tell me 

21    or the court what the exhibit number is.              

22               MR. ZIMROTH:  540 A point 16.  

23               THE COURT:  We have such a document in evidence?  

24               MR. ZIMROTH:  I promise.  Do you want it now?

25               THE COURT:  No, I don't.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 332    

 1               MR. ZIMROTH:  I didn't think so.

 2         Q.    Do you have P 16?

 3         A.    Total error model study?

 4               What page do you want?

 5         Q.    Table 14 -- Tables 14 and 15 I think would be 

 6    best.  Do you have table 14?

 7         A.    Yes, I do.

 8         Q.    And there is a total in the first column of 2.11, 

 9    and that is the undercount rate.  Is that right?

10         A.    That's right.

11         Q.    And then the next column, with a B and a little 

12    funny sign on top and then a capital U in parens with the 

13    same little I funny sign on top at the bottom has a number 

14    0.419.  Do you see that?

15         A.    I do.

16         Q.    That is the bias in the dual system estimate, 

17    isn't it?

18         A.    That is the bias at the national level of the 

19    dual systems estimate, yes.

20         Q.    .49 out of 2.1 is about 25 percent.  Isn't that 

21    right?

22         A.    That ratio is .25.

23         Q.    That's about 25 percent?

24         A.    Yes.

25         Q.    So about 25 percent of the net undercount is due 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 333    

 1    to bias according to Table 14.  Is that right?

 2         A.    No.

 3         Q.    That's wrong?

 4         A.    Yes.

 5         Q.    Would you turn to table 15.  Table 15 says total 

 6    error of the net undercount rate assuming no collision bias. 

 7               Do you see it?

 8         A.    I see it.

 9         Q.    That gives the same figure for the net undercount 

10    as Table 14, doesn't it?

11         A.    Yes.

12         Q.    That gives a bias of 0.73.  Isn't that right?

13         A.    That's right.

14         Q.    And that is about a little more than a third of 

15    the undercount, isn't it?

16         A.    That's a third of the 2.11, but that's not a 

17    meaningful ratio as of the undercount -- net undercount of 

18    national limit could go to zero.  That ratio could become 

19    infinite and cease to have meaning.  

20               That particular ratio is not a particularly 

21    meaningful ratio in terms of considering the biases of the 

22    dual systems estimate.

23         Q.    Now, the assumption that I was referring you to 

24    before, that E with a funny little squiggle, N zero -- sorry 

25    -- let me withdraw that question.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 334    

 1               Oh, yes, the assumption E with the funny squiggle 

 2    N zero V sub three entails that sampling error in the law as             

 3    -- the assumption, that can we call that funny squiggle a 

 4    twiddle if it will make it easier?

 5         A.    It's your cross-examination.

 6         Q.    Great.  Thanks.  

 7               THE COURT:  Does it have a technical name? 

 8               THE WITNESS:  Yeah, I think most statisticians 

 9    would read this to say "is distributed as" 

10               THE COURT:  E is distributed as? 

11               THE WITNESS:  Right, as a normal zero with 

12    variance.

13               MR. SITCOV:  I can live with that, your Honor.  

14               THE COURT:  Good.  Better than a twiddle.  

15               MR. SITCOV:  You know, your Honor, I have it on 

16    very sound information that in some dark corners of the 

17    statistical community, that is actually called a twiddle.  

18               THE COURT:  Oh.

19         Q.    So the assumption that E has normal distribution 

20    mean zero, V sub three entails that the sampling error in 

21    the raw adjustment factors has covariance matrix equal to 

22    the presmoothing covariance matrix for the raw adjustment 

23    factors.  Isn't that right?

24         A.    That's what it says here, yes.

25         Q.    That is an assumption, isn't it?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 335    

 1         A.    That's an assumption.

 2         Q.    That assumption entails that the sampling error 

 3    in the raw adjustment factor follows the normal 

 4    distribution, right?

 5         A.    That's what it means, yes.

 6         Q.    When I took your deposition five weeks ago, you 

 7    had no opinion as to the extent to which that assumption 

 8    holds true for the 1990 PES.  Isn't that right?

 9         A.    I had not considered the question as to whether 

10    the distribution of these variances is exactly normal. 

11               Among all the things that I did consider and 

12    could have considered, that was not one of them, that's 

13    true.

14         Q.    The answer is you had no opinion as to the extent 

15    to which that assumption holds true for the 1990 PES?

16         A.    That's right.

17         Q.    You didn't know whether that assumption was 

18    realistic and warranted, did you?

19         A.    That's right.  

20               MR. SITCOV:  We're progressing.  

21               THE COURT:  Would it be time for a break?  

22               MR. SITCOV:  Sure.  

23               THE COURT:  Let's take a three day break.  15 

24    minutes.

25               (Recess) 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 336    

 1               THE COURT:  Any time you wish, Mr. Sitcov.  

 2               MR. SITCOV:  How about next Tuesday?

 3               (Pause) 

 4         Q.    Dr. Ericksen, do you see the second equation that 

 5    is at the bottom of the third line from the bottom?

 6         A.    I see it.

 7         Q.    Of page 7?

 8         A.    Yes.

 9         Q.    That equation says little Y equals X beta plus V.  

10    Is that right?

11         A.    That's what it says.

12         Q.    And X represents the matrix of explanatory 

13    variables.  Is that right?

14         A.    That's what it represents, yes.

15         Q.    And beta is a perameter to be estimated?

16         A.    Beta is the perameter of regression coefficients 

17    to be estimated.

18         Q.    And V is a disturbance, isn't it?

19         A.    You can call it that, yes.

20         Q.    To the right of that equation is another 

21    assumption.  Isn't that right?

22         A.    That's what that is.

23         Q.    That assumption says V has normal distribution 

24    mean zero sigma square times I, right?

25         A.    That's what it says.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 337    

 1         Q.    Now, that assumption entails that the components 

 2    of V have mean zero, correct?

 3         A.    That's what it assumes, yes.

 4         Q.    And that assumption entails the components of V 

 5    have constant variance equal to sigma square.  Is that 

 6    correct?

 7         A.    That's what it says, yes.

 8         Q.    That is an assumption, isn't it?

 9         A.    That's an assumption.

10         Q.    When I took your deposition five weeks ago, you 

11    weren't able to testify that this assumption was realistic 

12    and warranted enough to use to adjust the 1990 decennial 

13    census, were you?

14         A.    My deposition was three days of questions like 

15    you've just been asking me.  I have almost no recollection 

16    of my deposition.  If you want to show me something, I'd be 

17    happy to read it.

18         Q.    All right.  Do you have your deposition 

19    transcript in front of you?

20         A.    Is this the whole thing?

21         Q.    Well, that's day one of the three you were just 

22    describing.  

23               MR. ZIMROTH:  Do you want to refer him to a 

24    particular line and page, Mr. Sitcov?  

25               MR. SITCOV:  Yes, I think we'll turn to page 176.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 338    

 1               Well, actually let's start at page 175 line 16: 

 2               "Q.    Now, that assumption I just asked you 

 3    about" --

 4         A.    What line are you on?

 5         Q.    Line 16 page 175:  

 6               "Q.    Now, that assumption I just asked you 

 7    about, that is, that the components of V have constant 

 8    variance equal to sigma square, is that to any degree 

 9    realistic and warranted?

10               "A.    I think I answered that question.

11               "Q.    No.  I think you did not -- no.  I don't 

12    think so you did.  

13               "MR. ZIMROTH:  Why don't we read back and see 

14    whether he did or did not.

15               "Q.    I asked you if something else was 

16    realistic and warranted but not that.

17               "A.    What was that?

18               "Q.    I don't remember exactly what.  

19               "MR. COREWIN:  I think it was the question before 

20    so if you could read it back.

21               "Q.    My specific question is, is the assumption 

22    it entails that the components of V have constant variance 

23    equal to sigma square realistic and warranted?  

24               "MR. ZIMROTH:  I think he did answer that 

25    question, Mr. Sitcov, when he answered that your question 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 339    

 1    implied it was a black and white, yes or no.  Therefore, he 

 2    could not answer that question.  

 3               "MR. SITCOV:  No, I think that is not it. 

 4               "THE WITNESS:  It is a matter of how realistic 

 5    and warranted.  

 6               "MR. SITCOV:  Is it realistic and warranted 

 7    enough to use in connection with adjusting the 1990 

 8    decennial census?

 9               "A.    To answer that question, I would have to 

10    go back in history and try to recall all the different 

11    things that we looked at, and I just don't recall the 

12    results of any studies that I have seen as to that question 

13    at this time."

14               Does that refresh your recollection that at the 

15    time I took your deposition, you weren't able to testify 

16    that this assumption was realistic and warranted enough to 

17    use to adjust the 1990 decennial census?

18         A.    When we looked at our research, we looked at so 

19    many different things.  At the time of my deposition nearly 

20    a year later, I did not recall the details of that 

21    particular consideration, that is true.

22         Q.    In fact, you couldn't recall the results of any 

23    studies you had seen as to that question.  

24               Isn't that right?

25         A.    It says what it says.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 340    

 1         Q.    Well, I am asking you, is that right?

 2         A.    That's what it says, so the answer is yes.

 3         Q.    And this assumption I've been asking you about 

 4    entails that the components of V are independent, doesn't 

 5    it?

 6         A.    That's what it entails.

 7         Q.    And that is an assumption, isn't it?

 8         A.    That is an assumption.

 9         Q.    You don't know whether this assumption is 

10    realistic and warranted, do you?

11         A.    As I explained to you in my deposition, it is a 

12    matter of how realistic and how warranted.  Assumptions are 

13    never perfectly true.  In order to decide whether or not you 

14    have a good estimate, you consider how realistic, how 

15    warranted, how far it deviates from perfection.

16         Q.    My question is, you didn't know that at the time 

17    I took your deposition, did you?

18         A.    At the time you took my deposition, I did not 

19    recall what I had considered, what I had done, with respect 

20    to that question, that's true.  

21         Q.    There is a question that E and V are independent.  

22    Isn't that right?

23         A.    That's what it says, yes.

24         Q.    When I took your deposition, you had no idea 

25    whether in the context of the 1990 production smoothing 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 341    

 1    model that assumption is realistic and warranted to any 

 2    extent, did you?

 3         A.    At the time you took my deposition, I don't 

 4    recall what we discussed on that question.

 5         Q.    So you don't know if you had no opinion 

 6    whatsoever about whether or not that assumption was 

 7    realistic and warranted?

 8         A.    I don't remember three days of deposition.  If 

 9    you show me what we talked about, that would refresh my 

10    memory, we can go forward.  

11         Q.    In fact, you didn't even know how you would test 

12    the degree to which the assumption in this smoothing model 

13    held, did you?  

14               MR. ZIMROTH:  The witness just said he would like 

15    his recollection refreshed.  If you have a page number, he 

16    would be happy to look at it.  

17               MR. SITCOV:  I don't think I am required to. 

18               Your Honor asked him on page so and so did you 

19    say this, on page so and so did you say that.  The man has 

20    testified for a day and a half about studies going back till 

21    1970.  I am only asking him about a deposition that is five 

22    weeks' old.  

23               THE COURT:  I don't think there is a pending 

24    question, is there?  

25               MR. SITCOV:  I think there is.  I think the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 342    

 1    question was, when I took your deposition, you didn't even 

 2    know how you would test the degree to which the assumptions 

 3    in the smoothing model held.  Isn't that right?  

 4               THE COURT:  The answer is either yes, no, or I 

 5    don't remember. 

 6               THE WITNESS:  I don't remember.

 7         Q.    Why don't you turn to page 245.

 8               (Pause) 

 9         Q.    Please turn to page 245 line 5, where I asked you 

10    the following question:  

11               "Q.    How could you have measured the degree to 

12    which" -- 

13               MR. ZIMROTH:  He is still looking.  

14               MR. SITCOV:  Oh, I am sorry.

15         Q.    Do you have it, Dr. Ericksen?

16         A.    Sorry?

17         Q.    Do you have page 245?

18         A.    I do.

19         Q.    Starting on line 5 I asked you the question:  

20               "Q.    How could you have measured the degree to 

21    which the assumptions on page 7 held?

22               "A.    Now you're asking me a different question 

23    because first you asked me how do you measure the degree to 

24    which these assumptions hold, and now you try to find some 

25    data that tests the assumption, and oftentimes you don't 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 343    

 1    have the perfect data that you need to so you try to find 

 2    data that might be consistent with the assumptions.  If the 

 3    assumptions were true, then certain other things might 

 4    follow.  That is genuinely what you might do.  What you do 

 5    specifically I can't think at this time.  I'd have to think 

 6    about that for a while."

 7               Does that refresh your recollection that when I 

 8    took your deposition five weeks ago, you didn't know how you 

 9    would test the degree to which the assumptions in this 

10    smoothing model held?

11         A.    Yes, I certainly said I'd have to think about 

12    that.  

13               MR. SITCOV:  Your Honor, I am going to show the 

14    witness what is our Exhibit 68 and I can guarantee beyond 

15    any shadow of a doubt that this exhibit has not been 

16    introduced into evidence.  

17               THE COURT:  All right.  You are going to move to 

18    introduce this thing?  

19               MR. SITCOV:  Yes.  

20               THE COURT:  68 is admitted.  

21               MR. ZIMROTH:  68?  

22               MR. SITCOV:  66 is the one we had.  

23               THE COURT:  66?  68?

24               (Defendant Exhibit 66 for identification was 

25    received in evidence)


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 344    

 1               (Pause) 

 2         Q.    Can you see both of these, Dr. Ericksen?

 3         A.    I can see the one on the left.

 4         Q.    Can you see this one?

 5         A.    My vanity precludes the purchase of glasses at 

 6    this time.  If you would be kind and move it about three 

 7    feet closer, I would do better.

 8               (Pause)

 9         A.    I can read this one.

10         Q.    Okay.  

11               MR. SITCOV:  Your Honor, can you see these these?  

12    I think it would be helpful.  

13               THE COURT:  Unfortunately, I can.  

14               MR. SITCOV:  This isn't the most interesting 

15    thing you have ever seen?

16         Q.    Dr. Ericksen, Defendant's Exhibit 68 is an 

17    exhibit that has been prepared for defendants by an expert 

18    we have retained for this litigation.  You are certainly 

19    free to check it, but it is identical to Defendant's Exhibit 

20    66, with one exception.  

21               You will see that in the second equation in the 

22    smoothing model, that is, Y equals X beta plus V with V 

23    having a normal distribution mean zero, sigma squared, V sub 

24    3 in 66, substituted for I in 66.  Do you see that? 

25         A.    Yes, I do.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 345    

 1         Q.    That is 68 over there?  

 2         A.    Over there, isn't it?

 3         Q.    Yes.  68 is the new one.

 4         A.    All right.

 5               MR. SITCOV:  Do you see the difference, your 

 6    Honor?  

 7               THE COURT:  No.  I am looking carefully now.  

 8               MR. SITCOV:  Okay.  In this third line from the 

 9    bottom, what is in Defendant's Exhibit 66, the second 

10    equation in the smoothing model says Y equals X beta plus V, 

11    where V has a normal distribution, mean zero, sigma squared 

12    I, okay?  

13               The exhibit that has been prepared by our expert, 

14    which is Defendant's Exhibit 68, has the same, exactly the 

15    same first equation in the smoothing model.  The second 

16    equation, however, reads Y equals X beta plus V and V has a 

17    normal distribution with mean zero, sigma squared, V sub 3.  

18               THE COURT:  The difference is the V sub 3?  

19               MR. SITCOV:  Instead of the I.  

20               THE COURT:  Okay.  

21         Q.    Do you recall that I showed you a copy of this 

22    document at your deposition, that is, Defendant's Exhibit 

23    68?

24         A.    Yeah, this particular one seared itself into my 

25    memory, so I do remember this one.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 346    

 1         Q.    Gotcha.

 2         A.    My memory is not completely imperfect.  I do 

 3    remember some things.

 4         Q.    Okay.  I advised you at that time it was prepared 

 5    by one of our experts in this case and that it was identical 

 6    to page 7 of Defendant's Exhibit 66, except for the change 

 7    that I just explained to the judge.  

 8               Is that right?

 9         A.    I even remember that.

10         Q.    Okay.  Now, this document modifies the second 

11    assumption in the smoothing model used to produce the dual 

12    system estimates for 1990.  In particular, instead of 

13    assuming that V has a normal distribution with sigma squared 

14    times I as the real smoothing model does, the modification 

15    assumes that V has a normal distribution sigma squared times 

16    V sub 3.  Is that right?

17         A.    That was a long question, and I think the answer 

18    is yes.

19         Q.    Okay.

20         A.    Basically what I think that boils down to is just 

21    the same, sigma square I and one A sigma squared 3 in the 

22    other.

23         Q.    Thats exactly right.  The alternative contained 

24    in Exhibit 68 assumes that the covariance matrix for the 

25    errors in the second equation is proportional to the matrix 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 347    

 1    to the errors in the first equation.  

 2               Isn't that right?

 3         A.    That's what it assumes, yes.

 4         Q.    And at your deposition you testified that this 

 5    could be a reasonable assumption, didn't you?

 6         A.    That now we have faded out of memory.

 7         Q.    Did would you turn to page 192, that would be the 

 8    first volume of your deposition.  If you want to look back 

 9    on 191, you'll see that, just to get context, you see I am, 

10    in fact, asking you about the equation that is -- assumption 

11    that is in Defendant's Exhibit 68, and on page 192 I asked 

12    you the question, at line 9:

13               "Q.    Is it possible this could be a reasonable 

14    assumption?

15               "A.    It certainly within the realm of 

16    possibility.  I am not going to say how likely it is."

17               Does that refresh your recollection that you 

18    testified that this could be a reasonable assumption?

19         A.    I had said it is within the realm of possibility. 

20               That doesn't mean it is all that likely.  It is 

21    something -- look, of all the things to consider, in 

22    something that is complicated as adjustment of the 1990 

23    census, there there are hundreds of things you might 

24    consider.  All of these things are certainly in the realm of 

25    theoretical possibility.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 348    

 1               What place this comes in, whether this is in the 

 2    25 percent most reasonable possibilities, in the 50 percent 

 3    most reasonable possibilities, in the bottom 25 percent at 

 4    this time I have no opinion.

 5         Q.    But you did testify at your deposition that it's 

 6    possible this could be a reasonable assumption?  

 7               MR. ZIMROTH:  He said what he said at the 

 8    deposition.  

 9               THE COURT:  Is that it?  

10               MR. SITCOV:  I just want to make sure that he --

11         A.    I will certainly agree that I said what I said at 

12    the deposition as recorded here.

13         Q.    You never considered this assumption before I 

14    showed you at your deposition, did you?

15         A.    No.  You learn something new every day.

16         Q.    And you didn't know at your deposition what any 

17    of the experts retained by you in your capacity as a member 

18    of the special advisory panel considered, did you?

19         A.    No.

20               MR. SITCOV:  I would like to just ask for your 

21    indulgence for a moment, your Honor. 

22               THE WITNESS:  Can we put this book away now?  

23               MR. SITCOV:  Which book?  Yes, I think you may.

24               (Pause) 

25               MR. SITCOV:  Would you turn to Exhibit 69. 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 349    

 1               THE WITNESS:  Who's?  

 2               MR. SITCOV:  Mine. 

 3               THE WITNESS:  I don't have an Exhibit 69.  

 4               THE COURT:  You're welcome to mine.

 5               (Pause) 

 6               MR. ZIMROTH:  Do you have another one for us?  

 7               MR. SITCOV:  I thought I did.  I had intended to 

 8    -- oh, wait.  I have one here.  Yes, I think you might have 

 9    an Exhibit 69.  Exhibit 69.  Look in the other book.

10               (Pause) 

11               MR. SITCOV:  Your Honor, you both may have 

12    Exhibit 69.  It is in the volume that begins 101 to 103.  If 

13    I can just approach the witness, I think that I could --

14               THE WITNESS:  Oh, it is in the other book?  

15               MR. SITCOV:  Yes.  That was the problem. 

16               THE WITNESS:  I thought it was with 68.  

17               MR. SITCOV:  No.  That would be too sensible. 

18               THE WITNESS:  It does doesn't say 69 on the 

19    outside cover.  

20               MR. SITCOV:  I know.  I think that is called good 

21    enough for government work, I figure, and I take all the 

22    blame. 

23               THE WITNESS:  I like our government.  

24               MR. SITCOV:  I do, too.  

25         Q.    Dr. Ericksen, have you ever seen this document 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 350    

 1    before?

 2         A.    In my capacity as a member of the special 

 3    advisory panel, we were sent various reports of the results 

 4    of regression models that the Census Bureau had considered 

 5    for use in smoothing.  I can't tell you if this is the exact 

 6    report that we got, but at least looks similar to reports 

 7    that I have studied.

 8         Q.    I will represent to you that this is in the 

 9    administrative record at page 10,582, that it is the -- that 

10    it represents the results of the regression models by region 

11    for production in the 1990 adjustment.

12               Do you recall -- 

13               MR. ZIMROTH:  Are we going to have any testimony 

14    on that or is this the testimony?  

15               MR. SITCOV:  It is in the administrative record, 

16    and in our case we will be offering the administrative 

17    record in evidence.  

18               MR. ZIMROTH:  We will be objecting to it.  Are 

19    you going to have any testimony on this?  

20               THE COURT:  On what, that is it is in the record?  

21               MR. ZIMROTH:  If that is in the record, fine, 

22    what they call the record, fine, that is, in fact, what was 

23    used.  

24               MR. SITCOV:  I think they would be happier if I 

25    said in the so-called record.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 351    

 1               MR. ZIMROTH:  That is fine, but you made other 

 2    representations as well.  

 3               MR. SITCOV:  Well, the document is what it 

 4    purports to be, and that is what the purpose of having a 

 5    sealed document like administrative record is.  

 6               MR. ZIMROTH:  That is not our understanding of 

 7    what that is.  When the time comes, we will object to that.  

 8               THE COURT:  All right.  Wait for the time to 

 9    come.

10               MR. ZIMROTH:  This is all assuming what you say 

11    you can prove.  Is that right?  

12               Is that where we are, your Honor, he is asking 

13    this witness to assume that what Mr. Sitcov says is true?  

14               THE COURT:  I don't know that he said anything 

15    other than this is in the so-called administrative record.  

16               MR. ZIMROTH:  Okay.

17         Q.    I showed you this document at your deposition, 

18    didn't I? 

19         A.    Something like it perhaps.  I do remember seeing 

20    something like this.  Whether this is the exact thing you 

21    showed me I don't remember.

22         Q.    Do you have some doubt that this is the exact 

23    document that I showed you?  

24               THE COURT:  I think we can take your 

25    representation that it is.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 352    

 1               MR. SITCOV:  Okay.  I will represent that it is.  

 2               THE COURT:  All right.

 3         Q.    Dr. Ericksen, I would like you, for the purposes 

 4    of my questions, to assume that this is the production 

 5    regression model and that the dependent variable is the 

 6    adjustment factor for the post-strata in the model.  Are you 

 7    following me?

 8         A.    So far.

 9         Q.    And the adjustment factor for a post-stratum is 

10    the estimated true population divided by the census count.  

11    Are you with me?

12         A.    So far.

13         Q.    Would you turn to page 3 of Defendant's Exhibit 

14    69.

15         A.    I'm there.

16         Q.    Now, that reports the results of the regression 

17    for the northeast region.  Is that right?  Can you tell that 

18    from the preceding pages?

19         A.    That's what it is.

20         Q.    Do you see that there is a variable F 108 in the 

21    second line from the bottom?

22         A.    I see it.

23         Q.    The regression coefficient for F 108 is minus 

24    .093.  Is that right?

25         A.    That's what it is.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 353    

 1         Q.    And that is a partial regression coefficient 

 2    indicating the effect of this variable on the dependent 

 3    variable holding constant all the other variables in the 

 4    equation, correct?

 5         A.    That's what it is.

 6         Q.    And this coefficient has a statistic of negative 

 7    3.977.  Is that right?

 8         A.    That's right.

 9         Q.    Can this negative sign be explained by sampling 

10    error?

11         A.    The negative sign is what it is.  Like any sample 

12    statistic, it's subject to the possible effects of sampling 

13    error.  

14         Q.    How likely is it that estimate of negative .093 

15    is due to sampling error?

16         A.    This is a regression coefficient that was 

17    estimated across the data.  It was certainly not be due just 

18    to sampling error.

19         Q.    Doesn't the statistic of 3.977 suggest that it is 

20    not likely due to sampling error?

21         A.    Well, it that is not a very good interpretation 

22    of what that statistic means, but it's not due just to 

23    sampling error.

24         Q.    So is it your testimony that this negative sign 

25    cannot be explained solely by sampling error?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      jah 5 12 6             Ericksen - cross                 354    

 1         A.    It cannot be explained solely by sampling error.

 2         Q.    Let's focus on black females aged 20 to 29 living 

 3    in owner occupied housing.  Let's compare such persons in 

 4    New York or other large central cities with those in small 

 5    cities or in the suburbs in the northeast region.  Holding 

 6    constant the other variables, like F 19 and F 25, with black 

 7    females aged 20 to 29 in New York have a smaller predicted 

 8    undercount rate, don't they?

 9         A.    There's black females 20 to 29 and you're 

10    comparing those --

11         Q.    Black females 20 to 29 living in owner occupied 

12    housing?

13         A.    Right.

14         Q.    In New York or other large central cities and 

15    we're holding constant all the other variables and we're 

16    comparing those women with like women in small cities or the 

17    suburbs?

18         A.    Okay, so what is the question?

19               (Continued on next page) 

20    
      

21    
      

22    
      

23    
      

24    
      

25    
      


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 355    

 1         Q.    Holding constant the other variables the black 
                                                                

 2    females age 20 to 29 living in owner-occupied housing in New 
                                                                   

 3    York have small predicted undercount rate than black females 

 4    aged 20 to 29 living in owner-occupied housing in small 

 5    cities or suburbs, isn't that right?

 6         A.    Well, however many those kinds of people there 

 7    are, yes.

 8         Q.    And that 9.3 percentage point difference, is that 

 9    just a little difference or is it a significant difference?

10         A.    It's a -- for that small subgroup, it's a fairly 

11    sizable amount.

12         Q.    So a 9.3 percentage point differential undercount 

13    is a significant one?

14         A.    Significance has different meaning to a 

15    statistician, and I don't think that it's an easily-answered 

16    question the way you framed it.

17         Q.    Would you look at page 11.  Actually, page -- 

18    pages 9 through 11.  That reports the results of the 

19    regression model for the West Region, is that right?

20         A.    That's what it reports.

21         Q.    Okay.  Can you look at variable F 108.  Do you 

22    see it?  That's on page 11.

23         A.    No, I need to study the full equation.

24         A.    Okay.

25         Q.    Do you see variable F 108 on page 11?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 356    

 1         A.    I see it.

 2         Q.    That is blacks in Los Angeles and large central 

 3    cities in the west, is that right?

 4         A.    That's right.

 5         Q.    And that has a negative coefficient, is that 

 6    right?

 7         A.    It does.

 8         Q.    And that coefficient is negative .122, is that 

 9    right?

10         A.    That's what it is.

11         Q.    And it has a T statistic of negative 3.646, is 

12    that right?

13         A.    That's right.

14         Q.    Can the negative coefficient for F 108 be 

15    explained by sampling error?

16         A.    It's not an easy question to answer the way 

17    you've framed it.  It's certainly not due just to the 

18    sampling error.

19         Q.    It's not likely that that is due to sampling 

20    error, is it?

21         A.    No.

22         Q.    Can you compute how likely that would be?

23         A.    Not as I sit here today without the relevant 

24    tables, and so forth.

25         Q.    Well, isn't the T statistic of 3 better than one 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 357    

 1    in a thousand?

 2         A.    That's sounds right.

 3         Q.    So it's about an one-in-a-thousand chance that 

 4    that's due to sampling error?

 5         A.    It's about a one-in-a-thousand chance that if you 

 6    had these data for everybody in the West Region, that the 

 7    coefficient would be zero or less.

 8         Q.    And holding constant the other variables, does it 

 9    show that blacks in big central cities in the west have 

10    smaller predicted adjustment factors than blacks outside the 

11    big central cities?

12         A.    That question makes no sense because you'll see 

13    there's also the mail return in the equation.  And any place 

14    where you would have blacks in any substantial numbers, you 

15    would most likely have a low mail return rate.  

16               So if you really want me to make the quite 

17    unrealistic assumption that we are holding the mailback rate 

18    constant, then the answer to your question is yes.

19         Q.    And it would be smaller by about 12.2 percentage 

20    points, is that right?

21         A.    Yes.  But like I said, 12.2 percentage points is 

22    counterbalanced by many other variables in the equation.

23         Q.    Would you turn to page 4 in Defendant's Exhibit 

24    69.  That is the regression model for the South Region.  Do 

25    you see that?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 358    

 1         A.    I do.

 2         Q.    Would you turn to page 6.  I think that 

 3    represents the results of the regression for the South 

 4    Region, is that right?

 5         A.    That's right.

 6         Q.    Now, I want to you consider two poststrata in the 

 7    south.  Both poststrata are male, non-black Hispanic, age 10 

 8    to 19, living in owner-occupied housing.

 9         A.    Wait a minute.

10         Q.    Would you like some paper --

11               MR. ZIMROTH:  Excuse me, your Honor.  Just for 

12    clarification, there's some illegible handwriting on my 

13    copy.  

14               Is that a part of the original document?  

15               MR. SITCOV:  Yes, it is.

16               MR. ZIMROTH:  Well, do you know whose it is or 

17    what it is?  

18               MR. SITCOV:  No, I don't.  

19               I believe Dr. Ericksen is asking for you for 

20    something to write on.  

21               MR. ZIMROTH:  Excuse me?  

22               MR. SITCOV:  I believe he's asking you for some 

23    paper.  

24               MR. ZIMROTH:  And I would also like to point out, 

25    your Honor, that the copy in the deposition transcript is 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 359    

 1    different from the copy we are now looking at.  So I don't 

 2    know what the representation was that was made earlier, but 

 3    this copy that we have that's attached to the deposition 

 4    transcript has a different figure in it from the copy that 

 5    we have been examining the last -- I don't know how many 

 6    minutes, cross-examining about.          

 7               THE COURT:  Mr. Sitcov, what do you say about 

 8    that?  

 9               MR. SITCOV:  Well, all I can say is that I don't 

10    know, but I can represent to you that the document that I 

11    have is the one that is in the record, and there are no 

12    material or substantive differences that I am aware of 

13    between the one that -- 

14               MR. ZIMROTH:  Two things, your Honor.  

15               First of all, this document I hold in my right 

16    hand, which Mr. Sitcov has represented is in the 

17    administrative record, does not have a bates stamp number on 

18    the bottom of it.  So it's hard for us to tell whether this 

19    is, in fact, a document in the administrative record.  

20               MR. SITCOV:  And I checked that it is.  

21               MR. ZIMROTH:  I'm just saying we have no way of 

22    checking because of the bates stamp number.

23               THE COURT:  I'm taking his representation.  

24               MR. SITCOV:  I can give you the bates stamp 

25    record.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 360    

 1               MR. ZIMROTH:  Fine.  I'm just making a record 

 2    now.  

 3               Second of all, the copy that is attached to the 

 4    deposition transcript on page 4 has in it a B, regression 

 5    model for adjustment factors 20 variables.  The 20 is 

 6    crossed out.  There's a 17 written in hand.

 7               Now, that is not in the copy that we've been 

 8    looking at here (indicating).  Now, I haven't gone back and 

 9    checked all the other pages, and so forth.  This is the 

10    first time I've noticed that, but it is a little disturbing.  

11               MR. SITCOV:  Well, I apologize for the 

12    disturbance, your Honor, and I would be happy to check the 

13    administrative record to be certain that there are no 

14    substantive differences between the two documents by any of 

15    the questions I have asked the witness, and tomorrow will 

16    advise the witness.  

17               MR. ZIMROTH:  But there's a different issue here, 

18    your Honor, which is:  What was the document that was shown 

19    to the witness at his deposition?  And I believe there were 

20    some questions and answers based on the representation that 

21    this is the document that was shown to the witness at the 

22    deposition.

23               And maybe -- well, I can just point out to that 

24    one difference that I see right in front of us.

25               THE COURT:  Well, let's proceed on the assumption 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 361    

 1    that the only difference between the two documents is the 

 2    handwriting that appears on one of them and not on the 

 3    other.  And then you can check later -- 

 4               MR. SITCOV:  I will.  And I will check to make 

 5    sure that there are no differences in the numbers that I've 

 6    been asking the witness about.  If there are, I'll bring it 

 7    to the Court's attention.

 8               THE COURT:  That seems to be a reasonable 

 9    solution. 

10    BY MR. SITCOV:

11         Q.    Now, Dr. Ericksen, do you have page 6 of 

12    Defendant's Exhibit 69 in front of you?

13         A.    I do.

14         Q.    Okay.  And that is the result of the regressions 

15    for the South Region?

16         A.    That's what it is.

17         Q.    And now I want you to consider two poststrata --

18         A.    Nice and slow.

19         Q.    Okay.

20               THE COURT:  It's getting late.  

21               MR. SITCOV:  For me too.

22         Q.    "Both poststrata are male, non-black, 

23    Hispanic" --

24         A.    Uh-uh.

25         Q.    I'm sorry. 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 362    

 1         A.    Okay.

 2         Q.    You have both poststrata are male, non-black, 

 3    Hispanic --

 4         A.    So far.

 5         Q.    -- ages 10 to 19.

 6         A.    10 to 19.

 7         Q.    Living in owner-occupied housing.

 8         A.    Owner-occupied housing?

 9         A.    Anything else?

10         Q.    No.  Oh, yes.

11               Both poststrata were enumerated by the 

12    mailout/mailback method.

13         A.    Both enumerated by the mailout/mailback method.

14         Q.    Poststrata No. 1 is in Houston, Dallas, or other 

15    large central city.

16         A.    Oh, we've got two of these, do we?

17         Q.    Yes, I told you we were considering two.

18         A.    Yeah, you did.

19         Q.    Poststrata -- are you ready for poststratum 2?

20         A.    Yes.

21         Q.    Okay.  Poststratum 2 is outside the central 

22    cities?

23         A.    You mean is a suburban one.

24         Q.    Yes.

25         A.    That can't be.  Because I don't think there are 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 363    

 1    any poststrata for owner-occupied suburban nights in the 

 2    South Region.

 3         Q.    Okay.  So that's an impossibility?  What is 

 4    F 113?

 5         A.    F 113 is H D plus C C plus plus C C minus.

 6         Q.    What's C C minus?

 7         A.    That means -- I don't know.  It's not listed 

 8    here.

 9         Q.    You can't tell what this F 113 is?

10         A.    Well, I know that H D is Houston, Dallas, and C C 

11    plus is larger central cities.  C C minus could be smaller 

12    central cities, but --

13         Q.    Could it be a suburban area?

14         A.    I don't think so.

15         Q.    Okay.  Well, if it can't be anything, then I 

16    can't ask you that.

17         A.    Oh, after all this work.

18         Q.    Would you turn to deposition Exhibit -- 

19    Defendant's Exhibit 67.

20         A.    Which book?

21         Q.    That should be -- I believe it's in the second 

22    one.

23               THE COURT:  You believe what?  

24               MR. SITCOV:  I believe it's in the second book.  

25    Yes, it is.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 364    

 1               MR. SITCOV:  Do you have it, your Honor?

 2               THE COURT:  Yes.  Freedman and Navidi.  Freedman 

 3    is here, Navidi is not.  

 4    BY MR. SITCOV:

 5         Q.    Do you recognize this, Dr. Ericksen?

 6         A.    Sure do.

 7         Q.    What do you recognize this to be?

 8         A.    Regression models for adjusting the 1980 census 

 9    by my friend David Freedman, who is sitting at the table 

10    behind you, by W. C. Navidi, who's never there.

11         Q.    Did you have a comment on this paper?

12         A.    Sure did.

13         Q.    Is it in here?

14         A.    I don't know.  Let me look.

15               (Pause) 

16               Better check your page numbers.

17         Q.    No, I think that's right.  I think that whoever 

18    copied this unfortunately did not copy both sides of the 

19    pages, your Honor.

20               THE COURT:  So we have every other page.  

21               MR. SITCOV:  Unless, you may -- there may be 

22    one --

23               THE COURT:  No, 3, 5, 7.  

24               MR. SITCOV:  Okay.  It just so happens that I 

25    have one that has all the pages.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 365    

 1               THE COURT:  Give it to the witness.  

 2               MR. SITCOV:  All right.  I will.

 3               (Pause)

 4    BY MR. SITCOV:

 5         Q.    The Freedman and Navidi article discusses the 

 6    production you proposed in connection with your work in the 

 7    Cuomo case, is that right?

 8         A.    I wouldn't put it quite like that, but I think 

 9    that it was a larger research than just that.  We did use 

10    results from that lawsuit, and Freedman and Navidi certainly 

11    referred to its use in the lawsuit.

12         Q.    So it does discuss the adjustment that you 

13    proposed for the 1980 manufacture Decennial Census?

14         A.    I think of it as discussing results of the 

15    research that I engaged in, yes.

16         Q.    And on page 5, Freedman and Navidi identified the 

17    assumptions in the smoothing model you proposed, isn't that 

18    right?

19         A.    They do discuss assumptions.

20         Q.    And these are the assumptions in the smoothing 

21    model you proposed?

22         A.    Those are some assumptions of the smoothing model 

23    I proposed.  

24               MR. SITCOV:  May I approach the witness for a 

25    second?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 366    

 1               THE COURT:  Sure.

 2               Oh, it's already up here.  Your Honor, this is 

 3    the page 5.  I don't know if you have it or not.

 4               THE COURT:  Yes, I do.  

 5               MR. SITCOV:  Okay.  And the assumptions that I 

 6    was referring to a moment ago are 1 through 7.

 7    BY MR. SITCOV:

 8         Q.    You wrote some comments about the Freedman and 

 9    Navidi article, didn't you?

10         A.    I did.

11         Q.    And so did Dr. Kirk Wolter, didn't he?

12         A.    He did.

13         Q.    And of that's the same Dr. Wolter who is a member 

14    of the Special Advisory Panel?

15         A.    The one and only.

16         Q.    And he is a co-author of the report that you 

17    wrote to the secretary that is Plaintiff's Exhibit 195, is 

18    that right?

19         A.    Yes.

20         Q.    Would you turn to page 25 of deposition 

21    Defendants' Exhibit 67.  

22               MR. ZIMROTH:  Excuse me.  Once again?  

23               MR. SITCOV:  25 

24               MR. ZIMROTH:  Of?  

25               MR. SITCOV:  67.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 367    

 1               (Pause)

 2    BY MR. SITCOV:

 3         Q.    That's part of the Dr. Wolter's comment on the 

 4    Freedman and Navidi article, is that right?

 5         A.    It is.

 6         Q.    Do you see point 4 on page 25?

 7         A.    Would you like me to read it through?

 8         Q.    Well, I'll direct your attention to the paragraph 

 9    that begins in the right-hand column, about halfway down, 

10    after the paragraph small B, it's the paragraph that begins 

11    "The analysis by Freedman and Navidi."  

12               Do you see that?  Do you see that paragraph, Dr. 

13    Ericksen?

14         A.    I'm reading.

15         Q.    Okay.  Let me know when you are done.

16               (Pause)

17         A.    I've read it.

18         Q.    Okay.  Do you agree with Dr. Wolter's opinion 

19    that whether adjusted counts are closer to the truth than 

20    the original census counts is fundamentally intractable 

21    because no one knows the true population counts?

22         A.    As I sit here today, I don't think of it that 

23    way.  In the real world, we need to make estimates based on 

24    all available evidence as to whether one set of estimates is 

25    closer than another set of estimates.  And I do consider 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 368    

 1    that to be an issue that statisticians consider.

 2         Q.    So was Dr. Wolter wrong to state that whether the 

 3    adjusted counts are closer to the truth than the original 

 4    census counts is fundamentally intractable?

 5         A.    I disagree that that is fundamentally 

 6    intractable.

 7         Q.    Would you turn to page 26.  There is some 

 8    material there entitled "Assumption 5."

 9               Now, in that paragraph, Dr. Wolter was commenting 

10    about assumption 5 in the Freedman and Navidi article, is 

11    that right?

12         A.    (No response) 

13               MR. SITCOV:  Assumption 5 is up here, your Honor, 

14    (indicating).  You won't have to keep going back and forth.  

15         Q.    There is an assumption in the 1990 production 

16    smoothing model that is similar to assumption 5 on page 5 of 

17    the Freedman and Navidi papers, isn't that right?

18         A.    Yes.

19         Q.    In fact, in the 1990 model, the expression for V 

20    is similar to the expression in assumption 5, isn't that the 

21    case?

22         A.    Both do assume constant variances.  I'm not going 

23    to get into what you mean by similar or what you don't mean 

24    by similar.

25               (Pause)


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 369    

 1               After you ask your next question, I would like to 

 2    take about a three-minute break.  

 3               MR. SITCOV:  I'd be happy to take that 

 4    three-minute break now.

 5               THE COURT:  Might it make more sense to adjourn?  

 6               MR. SITCOV:  Sure.

 7               THE COURT:  Anybody object to adjourning?  

 8               MR. ZIMROTH:  Only, I take it, there's no way 

 9    that you are going to finish today, Dr. Ericksen.  

10               MR. SITCOV:  No.

11               THE COURT:  If he has to come back tomorrow, we 

12    may as well adjourn now.

13               THE WITNESS:  May I ask, do I have to be here all 

14    day tomorrow?  

15               MR. ZIMROTH:  Do you have any idea how long 

16    you'll be?  We have another witness that's coming in out of 

17    town.  Can you give us an estimate of how long?  

18               MR. SITCOV:  I would guess, at least from my 

19    part, as long as I've gone already.

20               THE COURT:  Interminable is not something I can 

21    quantify.  

22               MR. SITCOV:  You don't think this part's more 

23    exciting.

24               THE COURT:  I have no idea how long you'll be 

25    going.  


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 370    

 1               MR. SITCOV:  Neither do I.  I think I've probably 

 2    been going -- about two hours.  

 3               MR. ZIMROTH:  Two hours.  So we should plan to 

 4    put on our next witness some time at the end of or 

 5    mid-morning tomorrow.  

 6               MR. SITCOV:  Unless you plan to redirect this 

 7    witness.  

 8               MR. ZIMROTH:  Right.

 9               THE COURT:  We don't need this on the record.

10               (Pause)

11               MR. ZIMROTH:  Your Honor, we also have a serious 

12    scheduling problem with the witness that's coming from out 

13    of town, and we would like to accomplish as much as we can, 

14    if we finish with Dr. Ericksen as early as possible, and I'm 

15    just wondering if we took a short break and couldn't go on a 

16    little bit earlier or start earlier tomorrow or both.

17               THE COURT:  All right by me.  Whatever you wish.  

18    I'll accommodate your desires.  

19               MR. SITCOV:  It's fine for me to go on.  I would 

20    just like to sit down for a moment.

21               THE COURT:  Let's take a ten-minute break.

22               (Discussion off the record)

23               (In open court) 

24               MR. SITCOV:  May I have the last question and 

25    answer read?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 371    

 1               (Record read) 

 2    EUGENE P. ERICKSEN, resumed           

 3    CROSS-EXAMINATION CONTINUED

 4         Q.    Dr. Ericksen, would you turn to page 208 of your 

 5    deposition transcript.  That would be in the first volume, I 

 6    think.

 7               Do you have page 208?

 8         A.    I have it.

 9         Q.    Okay.  Starting on line 16:

10               "Q.    Is there an equation that is similar to 

11    the 1990 production smoothing model to assumption 5 in the 

12    Freedman paper?

13               "Mr. Zimroth:  Objection.  What do you mean by 

14    similar -- do you mean -- do they both use Greek letters?  

15    That's similar also.  Is that what you mean by similar?  

16    It's an impossible question to answer unless you give it a 

17    definition, Mr. Sitcov.  Could you please give it some more 

18    definition.  Excuse me.  Will you give it some more 

19    definition?  

20               "Mr. Sitcov:  I think Dr. Ericksen isn't having 

21    any trouble.

22               "Dr. Ericksen:  Yes, I believe the expression for 

23    V is similar to the expression in assumption 5."

24               Does that refresh your recollection that you 

25    testified at your deposition that the expression for V in 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 372    

 1    the 1990 model is similar to the expression in assumption 5, 

 2    in the Freedman and Navidi paper?

 3         A.    What I meant by that answer is that they are 

 4    similar in the sense that they both assume constant 

 5    variances.  However, there are many important ways in which 

 6    they are not similar.  For example, the set of data that the 

 7    model -- 

 8               MR. SITCOV:  Your Honor, I just asked him if it 

 9    refreshed his recollection, if they were similar.  I didn't 

10    ask him for a lecture on it.

11               THE COURT:  All right.  Your answer is it does 

12    refresh your recollection, and you said it was similar. 

13    BY MR. SITCOV:

14         Q.    Now, turning back to the Freedman and Navidi 

15    article -- and you can see that it's up on the board in 

16    front of you -- assumption 6 states that the terms delta 1 

17    through delta 6 are independent from each other.  Isn't that 

18    right?

19         A.    That's assumption 6 on page 5, right.

20         Q.    In fact, the statement in the 1990 model that V 

21    is distributed normal would mean zero and variance matrix 

22    sigma squared i is similar to part of assumption 6 in the 

23    Freedman and Navidi article, isn't it?

24         A.    It's similar in the sense that out of context 

25    they both say the same thing.  But in the context of the 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 373    

 1    estimates that they are actually being used for, in the 1980 

 2    situation --

 3               MR. SITCOV:  Your Honor, I think he's answered my 

 4    question.  

 5               MR. ZIMROTH:  No, he has not, your Honor.  He's 

 6    answering the question.

 7               THE COURT:  Repeat your answer.

 8               THE WITNESS:  Yes, it's very dissimilar in the 

 9    sense that in 1980, we were computing estimates for 66 

10    geographic areas that are quite different than the 1392 

11    poststrata that are at issue in 1990.

12    BY MR. SITCOV:

13         Q.    It is also the case, is it not, that assumption 

14    6, in the 1980 smoothing model, has a similar meaning to the 

15    assumption in the 1990 model that E and V are independent, 

16    isn't that right?

17         A.    Not really because the meaning of independence 

18    when your observation of 66 states and part of states is not 

19    the same as the meaning of independent when you are talking 

20    about 1392 poststrata.

21         Q.    Can you turn to page 221 of your deposition 

22    transcript.

23               Do you recall testifying at your deposition that 

24    assumption 6 in the 1980 smoothing model has a similar 

25    meaning to the assumption in the 1990 model that E and V are 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 374    

 1    independent?

 2         A.    That's what I just said, has a similar meaning 

 3    and has another meaning which is very different.

 4         Q.    So it has a similar meaning?

 5         A.    It has -- among all the meanings, it has one that 

 6    is similar, yes.

 7         Q.    Okay.  You only testified in your deposition that 

 8    it had a similar meaning, isn't that right?

 9         A.    Yes, out of context, the meaning of independence 

10    has a certain theoretical meaning, yes.

11         Q.    It is the case, isn't it, that the only evidence 

12    that you and Drs. Estrada, Tukey and Wolter provided to the 

13    Secretary that adjustment improves the proportional 

14    distribution of a population at subnational levels were the 

15    results of loss function analysis?

16         A.    Is it true -- one more time, please.

17         Q.    Isn't it true that the only evidence that you and 

18    Estrada, Tukey and Wolter provided to the Secretary that 

19    adjustment improves the proportional distribution of the 

20    population at subnational levels were the results of loss 

21    function analyses?

22         A.    No.

23         Q.    What was the other evidence?

24         A.    Well, first of all, I would like to make it clear 

25    that what we were doing was reviewing the work in general 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 375    

 1    that was done at the Census Bureau.  The Census Bureau 

 2    provided evidence of errors both of the sampling and 

 3    nonsampling kind for 13 evaluations poststrata.  

 4               We did do some analysis of our own with the 13 

 5    evaluation poststrata that generally supplemented the work 

 6    that had been done at the Census Bureau.

 7         Q.    Well, evaluation poststrata, isn't that a 

 8    geographical area, isn't it?

 9         A.    Evaluation poststrata is defined in part as 

10    geographical component.

11         Q.    What is the capita of evaluation poststrata 1?

12         A.    I don't know.

13         Q.    Well, there is no capita because it is not a 

14    geographic area, is it?

15         A.    Well, that's like asking me what's the capital of 

16    Brooklyn.

17         Q.    Well, can you turn to page 16 -- I'm sorry, page 

18    18 of P 16.  Do you have the P 16 study in front of you?

19         A.    What page?

20         Q.    18.

21               Okay.  Do you have page 18?  This is the total 

22    error amount.

23         A.    I have it.

24         Q.    Okay.  On page 18, it lists the 13 evaluation 

25    poststrata, is that correct?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 376    

 1         A.    Yes.

 2         Q.    And evaluation poststrata 1 is minorities living 

 3    in central cities in the northeast.

 4         A.    Yes, that's what I mean by geographic component.

 5         Q.    But that's not a geographic area, is it?  You 

 6    can't adjust the northeast.

 7         A.    You calculate the adjustment factor for 

 8    minorities in poststrata and central city in the northeast 

 9    that is not a government recognized entity.  If that's what 

10    you mean, the answer to that question would be yes.

11         Q.    It's not a city, is it?

12         A.    It's not a state.

13         Q.    It's not a state?

14         A.    Nope.

15         Q.    It's not a county?

16         A.    It's not a county.

17         Q.    It's not a census tract?

18         A.    It's not a census tract.

19         Q.    It's not a block?

20         A.    It's not a block.

21         Q.    But the loss function analysis was an attempt to 

22    show that adjustment improves the proportional distribution 

23    of a population at the subnational levels, correct?

24         A.    What loss function analysis?

25         Q.    Wasn't that the purpose of the loss function 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 377    

 1    analysis that you suggested to the Secretary?

 2         A.    I don't recall suggesting that the Secretary do a 

 3    loss function analysis.

 4         Q.    Do you have Plaintiff's Exhibit 195 still?  

 5               Would you turn to page 20 of the main report.

 6               I believe you testified earlier that page 20 

 7    discusses a loss function.  Is that right?

 8               There i-- about three quarters of the way down, 

 9    on page 20, there is an indented line that says sum of 

10    (errors divided by size) squared weight by state."

11               That's a loss function, isn't it?

12         A.    What we were doing in this section --

13         Q.    No.  My question is:  That is a loss function, 

14    isn't it?

15         A.    Wait a minute.  I've lost the context of your 

16    question.  Could you --

17         Q.    Is that material that I just read to you, on page 

18    20 of the main report, which is Plaintiff's Exhibit 195, 

19    that says:  "Sum of (error divided by size) squared weight 

20    by state" -- isn't that a loss function?

21         A.    That's a measure of error.  That could be used in 

22    a loss function, yes.

23         Q.    Okay.  And then in the next sentence, you say 

24    that the empirical evidence is supplied by Wolter and 

25    Causey, is that right?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 378    

 1         A.    That's empirical evidence for the passages under 

 2    A and B.

 3         Q.    And that refers to Appendix G of your report, is 

 4    that right?

 5         A.    That's the Wolter and Causey paper, yes.

 6         Q.    Okay.  Would you turn to Appendix G.           

 7               Do you have it?

 8         A.    I do.

 9         Q.    Would you turn to page 3.

10               There are two equations on that page marked 2.1 

11    and 2.2, aren't there?

12         A.    Right.

13         Q.    And both of those are loss function equations, 

14    aren't they?

15         A.    Yes, these are equations that could be used in 

16    loss functions, that's correct.

17         Q.    And the equation 2.1 is an absolute error loss 

18    function, is that right?

19         A.    It is an absolute error loss function expression 

20    -- it's an expression of absolute error that could be used 

21    in a loss function.

22         Q.    And equation 2.2 is a squared error loss 

23    function, is that right?

24         A.    It is a squared error measure that could be used 

25    in a loss function, yes.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 379    

 1         Q.    Would you turn to page 4, please.  Do you see 

 2    equation 2.3?

 3         A.    Yes, I do.

 4         Q.    That's a squared error loss function, isn't it?

 5         A.    It's a squared error expression that could be 

 6    used in a loss function.

 7         Q.    Is your answer yes?

 8         A.    I said it's a squared error expression that could 

 9    be used in a loss function.

10         Q.    And that's what it was used as here.

11         A.    That's what it's used as, I believe, in this 

12    article.

13         Q.    As were the other two I asked you about.

14         A.    As we sit here today, I don't recall us using all 

15    three of these, but certainly used either two or three of 

16    them, yes.

17         Q.    And you also see there's an equation that's 

18    marked 2.4 on page 4?

19         A.    Yes, I do.

20         Q.    And that is an absolute error loss function, 

21    isn't it?

22         A.    That's an expression of absolute error that could 

23    be used in a loss function, yes.

24         Q.    And that's what this article is using it as, 

25    isn't it?


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 380    

 1         A.    That is what this article is using them as in 

 2    this article, yes.

 3         Q.    And none of the loss function equations on pages 

 4    3 and 4 of the article we're looking at now are the same as 

 5    the loss function equation on page 20 of your main report, 

 6    are they?

 7         A.    That's right.

 8         Q.    And of these five loss functions equations, which 

 9    is the one that is the sort of textbook loss function 

10    equation?

11         A.    I don't think there is such thing as a textbook 

12    loss function equation.  It's different loss functions we've 

13    used in different situations.

14         Q.    Well, are there generally-accepted principles in 

15    statistics about which loss function is considered the most 

16    accurate?

17         A.    I don't think so.

18         Q.    You agree, don't you, that two statisticians 

19    could disagree about which is the appropriate loss function 

20    to use in the context of adjusting the 1990 Decennial 

21    Census, and both can be reasonable in their choice of loss 

22    function?

23         A.    Depending, of course, on what the actual loss 

24    functions being considered are, I think it is the case that 

25    there could be different loss functions that could be 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 381    

 1    considered, that's right.

 2         Q.    That could be --

 3         A.    Considered.

 4         Q.    Reasonable.

 5         A.    (No response) 

 6               MR. ZIMROTH:  There is a question pending.  

 7               MR. SITCOV:  Yes.

 8         Q.    My question is:  Could they both be considered -- 

 9    let me restate it in case there's any doubt.

10               Do you agree that two statisticians could 

11    disagree about which is the appropriate loss function to use 

12    in the context of adjusting the 1990 Decennial Census, and 

13    both could be reasonable in their choice of loss function?

14         A.    I can imagine two different loss functions that 

15    could be used, and I would not think that -- I may not think 

16    that that would be unreasonable.  Other loss functions would 

17    be unreasonable.

18         Q.    So the answer is that there could be more than 

19    one reasonable loss function to use for adjusting the 1990 

20    Decennial Census -- 

21               MR. ZIMROTH:  The answer is what the witness 

22    previously gave.

23               THE COURT:  He's entitled to paraphrase if he can 

24    get an agreement to his paraphrase.

25         A.    Look, you don't usually use loss functions to 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 382    

 1    make the adjustment.  You use loss functions to evaluate the 

 2    quality of the adjustment, and different statisticians will 

 3    use different loss functions to make different evaluations, 

 4    and some of those evaluations will be better than others.

 5         Q.    But there is more than one reasonable loss 

 6    function to use for that purpose, is that right?

 7         A.    Well, there may be.  I'd have to take a look at 

 8    the loss function and take a look at the results and give 

 9    you an opinion as to whether or not that's reasonable.

10         Q.    But I'm asking you:  Is there more than one that 

11    is reasonable?  

12               MR. ZIMROTH:  Your Honor, this witness has 

13    answered this question at least three times.

14               THE COURT:  I'll permit it.

15         A.    I could imagine more than one that would be 

16    reasonable.

17         Q.    I'm not asking you if you can imagine that.  I'm 

18    asking you:  Are there?  Is there more than one?  You've 

19    identified five different ones for me, and I'm asking you 

20    now:  Is it a fact that there is more than one reasonable 

21    loss function --

22         A.    To answer that question, you'd actually -- see 

23    you have to look at the results of loss function, think 

24    about it, and make a consideration in the context of the 

25    actual data.


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 383    

 1         Q.    Well, were all the four loss financial equations 

 2    in the Wolter-Causey paper reasonable?

 3         A.    For the purpose that Wolter and Causey were using 

 4    them, to the best of my recollection, they were reasonable.

 5         Q.    Well, were they reasonable for the purpose of 

 6    deciding whether to adjust the 1990 Decennial Census?

 7         A.    They were not used for that purpose, and this 

 8    section of the report that we were writing, we did not have 

 9    access to the loss function that was calculated by the 

10    Census Bureau.  

11               Based on the information we had, based on the 

12    results of that paper, we considered it likely that the loss 

13    function analysis that would be calculated would show that 

14    the adjusted data bought were an improvement over the 

15    original enumeration.  That's all we were doing here.

16         Q.    Well, is it your testimony then that you don't 

17    know whether or not the four loss function equations that I 

18    just asked you about in the Wolter Causey paper were -- 

19    would be reasonable to use in evaluating whether or not to 

20    adjust the 1990 Decennial Census?

21         A.    The answer is they might be.

22         Q.    So you have no current opinion now whether they 

23    are or aren't they?

24         A.    I want to compare those to the loss function 

25    equations that actually were used, think about it, and I can 


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
      lta 9205127census      Ericksen - cross                 384    

 1    give you an answer.  But as I sit here today, that's not the 

 2    kind of question that you give an instant answer to.

 3         Q.    So you don't have an opinion?

 4         A.    No.  

 5               MR. SITCOV:  Your Honor, would this be a good 

 6    time to break.

 7               THE COURT:  Let's go off the record.

 8               (Discussion off the record)

 9               (Proceedings adjourned to 9:30 a.m., Wednesday, 

10    May 13, 1992)

11    
      

12

13

14

15

16

17

18

19

20

21

22

23

24

25


                 SOUTHERN DISTRICT REPORTERS 212-791-1020
