[Senate Hearing 107-1061]
[From the U.S. Government Publishing Office]



                                                       S. Hrg. 107-1061
 
                       YEAR 2000 DECENNIAL CENSUS

=======================================================================

                                HEARING

                               before the

                         COMMITTEE ON COMMERCE,
                      SCIENCE, AND TRANSPORTATION
                          UNITED STATES SENATE

                      ONE HUNDRED SEVENTH CONGRESS

                             FIRST SESSION

                               __________

                             MARCH 28, 2001

                               __________

    Printed for the use of the Committee on Commerce, Science, and 
                             Transportation





                        U.S. GOVERNMENT PRINTING OFFICE

88-017                        WASHINGTON : 2004
_____________________________________________________________________
For sale by the Superintendent of Documents, U.S. Government Printing
Office Internet: bookstore.gpo.gov  Phone: toll free (866) 512-1800
Fax: (202) 512-2250  Mail: Stop SSOP, Washington, DC  20402-0001







       SENATE COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION

                      ONE HUNDRED SEVENTH CONGRESS

                             FIRST SESSION

                     JOHN McCAIN, Arizona, Chairman
TED STEVENS, Alaska                  ERNEST F. HOLLINGS, South Carolina
CONRAD BURNS, Montana                DANIEL K. INOUYE, Hawaii
TRENT LOTT, Mississippi              JOHN D. ROCKEFELLER IV, West 
KAY BAILEY HUTCHISON, Texas              Virginia
OLYMPIA J. SNOWE, Maine              JOHN F. KERRY, Massachusetts
SAM BROWNBACK, Kansas                JOHN B. BREAUX, Louisiana
GORDON SMITH, Oregon                 BYRON L. DORGAN, North Dakota
PETER G. FITZGERALD, Illinois        RON WYDEN, Oregon
JOHN ENSIGN, Nevada                  MAX CLELAND, Georgia
GEORGE ALLEN, Virginia               BARBARA BOXER, California
                                     JOHN EDWARDS, North Carolina
                                     JEAN CARNAHAN, Missouri
                  Mark Buse, Republican Staff Director
               Ann Choiniere, Republican General Counsel
               Kevin D. Kayes, Democratic Staff Director
                  Moses Boyd, Democratic Chief Counsel




                            C O N T E N T S

                              ----------                              
                                                                   Page
Hearing held on March 28, 2001...................................     1
Statement of Senator Brownback...................................    17
Statement of Senator Hollings....................................     3
    Prepared statement...........................................     3
Statement of Senator Inouye......................................     4
Statement of Senator Kerry.......................................     1
    Prepared statement...........................................     2

                               Witnesses

Clay, Hon. William, U.S. Representative from Missouri............    37
Gonzalez, Hon. Charles A., U.S. Representative from Texas........    38
    Prepared statement...........................................    40
Ericksen, Eugene P., Professor of Sociology and Statistics, 
  Department of Sociology, Temple University.....................    54
    Prepared statement...........................................    56
Evans, Hon. Donald L., Secretary of Commerce; accompanied by Mr. 
  William Barron, Acting Census Director.........................     4
    Prepared statement...........................................     7
Maloney, Hon. Carolyn, U.S. Representative from New York.........    50
    Prepared statement...........................................    52
Miller, Hon. Dan, U.S. Representative from Florida...............    41
    Prepared statement...........................................    43
Murray, Dr. David W., Director, Statistical Assessment Service 
  and Census Monitoring Board, Congressionally Appointed Member..    23
    Prepared statement...........................................    25
Vargas, Arturo, Executive Director, National Director, National 
  Association of Latino Elected and Appointed Officials (NALEO) 
  Educational Fund...............................................    33
    Prepared statement...........................................    35
Wachter, Kenneth W., Professor, Chair, Department of Demography, 
  University of California.......................................    64
    Prepared statement...........................................    65

                                Appendix

Article from USA TODAY...........................................    87
Response to written questions submitted by Hon. Barbara Boxer to 
  Donald L. Evans................................................    87
Response to written questions submitted by Hon. Sam Brownback to 
  Donald L. Evans................................................    75
Response to written questions submitted by Hon. John McCain to:
    Eugene P. Ericksen...........................................    81
    Donald L. Evans..............................................    75
    David W. Murray..............................................    77
    Arturo Vargas................................................    79
    Kenneth W. Wachter and David A. Freedman.....................    76


                       YEAR 2000 DECENNIAL CENSUS

                              ----------                              


                       WEDNESDAY, MARCH 28, 2001

                               U.S. Senate,
        Committee on Commerce, Science, and Transportation,
                                                    Washington, DC.
    The Committee met, pursuant to notice, at 9:28 a.m., in 
room SR-253, Russell Senate Office Building, Hon. John F. 
Kerry, presiding

           OPENING STATEMENT OF HON. JOHN F. KERRY, 
                U.S. SENATOR FROM MASSACHUSETTS

    Senator Kerry. The hearing will come to order. We are going 
to start just literally a moment early, because we are under 
the time gun here. A number of our colleagues from the House 
have requested to testify early, and I appreciate that.
    I wanted to try to accommodate it. The problem is, we also 
have the Secretary under the same time pressure so we are going 
to try to move as expeditiously as possible and not lose part 
of the hearing in doing so.
    Mr. Secretary, thank you very much for taking the time to 
come up. We appreciate it. As the Members of the Committee 
know, the census was something that a lot of us have been 
following very closely, and I am grateful to Secretary Evans, 
who said during his confirmation hearing in January that he 
would follow this closely and be happy to appear before the 
Committee again on that, and I appreciate his honoring that 
commitment today.
    I am not going to go through my full statement.
    Obviously, what is at stake here is the accuracy of the 
census, which really is at the heart of a number of issues, not 
just redistricting and the adequacy of representation in the 
country, but also the distribution of funds, and whether or not 
there is a sense of fairness and inclusivity in the process by 
which we not only represent America but see that America 
receives its fair share of the funds distributed federally.
    There are roughly $185 billion a year that is distributed 
based on population counts that come from the census over a 10-
year period. You are obviously, therefore, looking at over $1 
trillion to $2 trillion. That makes a difference in the lives 
of our fellow Americans, and so it is particularly important 
for people to have a sense that this was done properly, fairly, 
and in a way that every American can have confidence that they 
are properly represented.
    There are comments I would make with specificity about 
Demographic Analysis and other things, but I do not want to 
lose the time now that we need desperately to move forward in 
terms of testimony and questions, and I will try to do it in 
the course of that.
    [The prepared statement of Senator Kerry follows:]
               Prepared Statement of Hon. John F. Kerry, 
                    U.S. Senator from Massachussetts
    Thank you, Mr. Chairman. I would like to thank the Chairman for 
calling this hearing, and especially for managing to schedule it during 
what we all know is an incredibly busy time for him. His willingness to 
arrange this hearing is reflective of his excellent leadership and of 
his understanding of the importance of this issue.
    As the Committee knows, the census is an issue that I have been 
following very closely and that I am extremely interested in. I thank 
Secretary Evans for coming again before this Committee to address our 
questions and concerns. When Secretary Evans was before us for his 
confirmation hearing in January, the census and the possibility of 
using corrected data to compensate for people missed by the census were 
issues raised by several senators. During his confirmation hearing, the 
Secretary assured us that he would be happy to appear before this 
Committee again on this matter, and I appreciate his willingness to 
honor that commitment and testify here today.
    I know that we have many distinguished witnesses here, and I don't 
want to delay the Secretary or the House Members, who I know have very 
busy schedules to maintain. But I would, just briefly, like to make a 
few comments about what we'll be discussing this morning. First of all, 
I believe very strongly that it is critically important to understand 
the consequences of missing people in the decennial census. I am quite 
pleased that the Census Bureau believes it has possibly achieved the 
most accurate census ever. I know that it took the tireless efforts of 
over 800,000 people and significant resources to reach this goal, and I 
commend all of those involved for their excellent work.
    More good news: earlier this month, the Census Bureau reported that 
it had dramatically reduced both the net and the differential 
undercounts from the last census. Again, I commend all of those who 
dedicated themselves to improving the census. But in the midst of this 
progress, we must not lose sight of the fact that at least 6.4 million 
people were not counted in Census 2000. Blacks were missed at twice the 
rate of Whites, Hispanics were missed three times as often, and 
American Indians were missed five times as often.
    The Bureau has spent $400 million on the Accuracy and Coverage 
Evaluation, or A.C.E., which evaluates accuracy and corrects for any 
undercount. As we know, earlier this month the Bureau did not recommend 
releasing the A.C.E. for the purpose of redistricting, and therefore, 
the official data remains the raw count, short over 6 million people.
    It is my understanding that the Census Bureau has not yet released 
the complete data that would tell us who the people are who were not 
counted, but if the 1990 census is any indicator, then we can be sure 
that they were predominantly poor and predominantly minority. What can 
be gleaned from the 1990 data is that unless the 2000 census data is 
corrected to account for those missed by census takers, vast numbers of 
poor, minorities in this country will be denied hundreds of millions of 
dollars in federal assistance.
    President Bush has often repeated that he is committed to leaving 
no child behind. I can think of no greater or clearer example of 
millions of children--literally--getting left behind than not being 
included in the census and potentially being denied federal funding. 
Roughly $185 billion per year in federal funding is distributed based 
on population counts derived from the 1990 census, to say nothing of 
the policy decisions that were made in states, cities, and counties 
around the nation based on this data. We must do everything in our 
power to ensure that this situation does not repeat itself. Indeed, the 
Census Bureau has worked hard to develop a methodology to respond to 
this situation. I sincerely hope that they are able to solve the 
mysteries of this issue before the fall to ensure that no child is left 
behind.
    My hope for this hearing today is that Secretary Evans will help to 
clarify the Census Bureau's March 1 decision not to use the corrected 
number for redistricting. I am concerned that there is misinformation 
swirling about, there is an impression that the Bureau ruled against 
the use of modern statistical methods in any census data. I have read 
their recommendation, and I do not believe that to be the case. My 
interpretation of their report--which I think is relatively clear--is 
that the Census Bureau needed more time to verify the accuracy of 
A.C.E. Further, the Bureau stated that there is considerable evidence 
to support the use of corrected data and that the majority of the 
evidence indicates the superior accuracy of the corrected data.
    The reason that both clarifying this point is so important and 
understanding where the Secretary stands with respect to this process 
is so important, is that in the fall the Bureau will release numbers 
for use in determining federal funding allocations. It is unclear to me 
who will make the decision about which numbers to use and how that 
decision will be made. It is my sincere hope that the Secretary will 
elucidate this matter for us.

    Senator Kerry. Mr. Ranking Member, do you have any 
comments?

             STATEMENT OF HON. ERNEST F. HOLLINGS, 
                U.S. SENATOR FROM SOUTH CAROLINA

    Senator Hollings. Mr. Chairman, I thank Senator McCain and 
yourself and, of course, the distinguished Secretary for being 
here. My statement will go in the record. I have another 
hearing, as we always do, so I will move to that and I will cut 
it short, because we have got House Members who are trying to 
make a roll call.
    But Secretary Evans, it is sort of good to harken the 
history of the census count and the problems we have had in the 
1990 census. President George Bush, Senior was in at that 
particular time and, as to the count, the lawsuits broke out 
like the measles with all of the undercounts and everything 
else. So we got the National Academy of Sciences to see how 
best to get the most complete, the most full, and the most 
accurate count possible. The Academy belabored it and came up 
with the sampling solution, which of course we have avoided 
thus far, but I feel very strongly in favor of that approach.
    Everybody knows the politician stands in the well and says 
the Constitution says count everybody, which is totally 
impossible, yet on the other hand you want to make the best 
endeavor that you possibly can. Right to the point, what we 
want to avoid are headlines like today's headline: ``Cities' 
Minorities Losing Census Undercount.'' But here we go again. In 
the famous words of our leader, Ronald Reagan. Here we go 
again.
    So that is what concerns this Committee. We thought we had 
the right approach as developed, and that is why Secretary 
Mineta did not want to make the decision on sampling. He left 
it to the professionals because he did not think he had 
expertise to analyze the scientific studies which showed should 
be done, that is, use the sampling approach.
    So I wish you would consider that background and history, 
and the present-day cries right now with respect to the problem 
we are having, because nobody's trying to get an overcount of 
anything. There is no chance of that, but the best system that 
I know of that has been devised so far, unless the Secretary 
has got a better one, is that sampling approach. We do 
appreciate your appearance here before the Committee, and I 
apologize for having to get to this other Committee meeting.
    Thank you, Mr. Chairman.
    [The prepared statement of Senator Hollings follows:]
            Prepared Statement of Hon. Ernest F. Hollings, 
                    U.S. Senator from South Carolina
    I would like to thank the Chairman for holding this hearing today. 
In addition, I would like to thank Secretary of Commerce Donald Evans 
for appearing before us today to honor a commitment that he made during 
my all-too-brief Chairmanship of this Committee in January to return 
and have a frank discussion with us about the Census.
    Mr. Secretary, as I have told you before, I am a strong supporter 
of statistical sampling. Even though some are calling the 2000 Census 
``the best Census ever,'' I am of the opinion that no Census is good 
enough if it leaves more than 3 million people uncounted.
    I also believe that decisions regarding the use of sampling should 
be made by the professionals not by the politicians. It is convenient 
for the Administration that the professionals recommended against 
sampling, but I wonder whether the threat of being overruled influenced 
that recommendation.
    However, I do not want to play the blame game that we so often play 
here in Washington. I want to look to the future. First of all, I would 
like to see the Census fully release its Accuracy and Coverage 
Evaluation (A.C.E.), upon which any adjustment to the Census numbers--
any sampling--would be based. If the A.C.E. truly is flawed, what 
better way to know than to allow independent statisticians to take a 
look at the full data?
    Second, I would like to encourage the Census to keep working to 
improve the 2000 data. Simply because the preliminary redistricting 
data has been released, the issue of accuracy in the Census is not 
resolved. We use these data, adjusted from year to year, as the basis 
for allocation of government funding. To that end, I applaud the Bureau 
for continuing to work on finding answers to the questions that they 
have about the adjusted data.
    The issue of ensuring that the entire population is counted is very 
real, not just for large cities like New York and Los Angeles, but also 
for my State of South Carolina. Nearly 2 in 5 households in my State 
did not mail back a Census form--the Nation's lowest response rate 
except for Alaska. I am worried that thousands of South Carolinians 
were not counted. The State Budget and Control Board estimates that 
each uncounted resident could cost the State $2,000 per person in 
Federal funding. Overall, an inaccurate count could cost my State 
millions--and South Carolina is only the tip of the iceberg.
    In conclusion, when groups and areas are undercounted, the strength 
of their vote is diluted, they do not get their fair share of Federal 
funding, and they suffer economically when banks and insurance 
companies base their decisions on demographic data. So I hope that the 
Census Bureau and the Department will continue to improve its data in 
order to achieve the most accurate count. That is the only way to 
ensure justice and fairness are the guiding principles in this vital 
matter.

    Senator Kerry. Thank you very much, Senator Hollings.
    Senator Inouye.

              STATEMENT OF HON. DANIEL K. INOUYE, 
                    U.S. SENATOR FROM HAWAII

    Senator Inouye. Like my colleague, I will have to be 
leaving soon, Mr. Chairman. If I may, I would like to hear the 
Secretary's testimony.
    Senator Kerry. Mr. Secretary, let us, if we can, invite 
your testimony, and then we will try to get into questions and 
see if we can incorporate some of the issues I talked about 
earlier.

        STATEMENT OF HON. DONALD L. EVANS, SECRETARY OF

              COMMERCE; ACCOMPANIED BY MR. WILLIAM BARRON,

                         ACTING CENSUS DIRECTOR

    Secretary Evans. Thank you very much. It is a pleasure to 
appear before this Committee again to discuss Census 2000. You 
have my prepared statement for the record, but I would like to 
just briefly touch on some of the points from that statement 
here today and then answer any questions you might have.
    Senator Kerry. Thank you. Your full statement will appear 
in the record.
    Secretary Evans. Thank you very much, Senator. I appreciate 
that. I have said Census 2000 is the most accurate this nation 
conducted in the past 60 years, perhaps ever. Its success is a 
direct result of the hard work and dedication of the employees 
of the Census Bureau and of hundreds of thousands of people, 
including thousands of your constituents who worked on Census 
2000.
    Census 2000 was an operational success. The Census Bureau 
met the deadline for releasing data, counted more residents 
than ever before, reduced the estimated undercounts, and 
reversed the trends toward smaller percentages of the 
population responding to the census.
    One of the major reasons for the operational success of 
Census 2000 was Congress' commitment to provide full funding 
that brought a number of improvements to bear on Census 2000. 
The Census Bureau carried out an unprecedented outreach program 
specifically targeted to groups that historically have had the 
highest undercounts. They reached out to families, 
neighborhoods, and communities to encourage participation.
    First, the Census Bureau developed an aggressive marketing 
plan and partnership program. Partnerships with a variety of 
governmental community and educational groups, 140,000 in all, 
were key to building support for census participation all the 
way down to the neighborhood level. These helped the Census 
Bureau to knock down obstacles that might get in the way of 
participation.
    The bureau also greatly expanded its outreach into 
communities by using paid advertising for the first time, 
placing more than $100 million in ads to educate and motivate 
the public to respond, especially those who had been 
undercounted at higher rates in the past censuses.
    Second, the Census Bureau took its educational outreach 
program directly into schools, providing materials to classroom 
teachers so that they could teach lessons on the census.
    Third, the Census Bureau made questionnaires more user-
friendly, so they were easier to read and to fill out. They 
published questionnaires in six languages and provided 
materials to help individuals fill them out in 49 languages.
    Another improvement that went a long way to Census 2000's 
success was the hiring of highly skilled staff on a temporary 
basis to get the job done on time. Through its ``Quality 
Counts'' operation, the Census Bureau redoubled its efforts to 
ensure quality and completeness in the count, identifying homes 
that should be visited again to review and verify information 
or to fill in the blanks.
    In using the advanced technology, the Census Bureau was 
able to process the data faster and to introduce quality 
assurance steps to be sure that data were captured accurately. 
Along with other results, due to these efforts the Census 
Bureau achieved a higher response rate through the mail than 
was expected. In fact, people in about 65 percent of the homes 
that were mailed questionnaires had already responded by mail 
before it came time for the Census Bureau to follow up with its 
door-to-door field operation.
    As the Members of this Committee know, the Census Bureau 
conducted an independent survey of approximately 314,000 homes. 
The Accuracy and Coverage, Evaluation Survey, or A.C.E., that 
was designed to measure net census coverage according to the 
current estimates from the survey, Census 2000, achieved net 
coverage rates for the total population of about 99 percent. 
The estimated coverage rates for individual ethnic groups were 
also extremely high. A.C.E. estimates support the conclusion 
that Census 2000 achieved both reduced net and differential 
undercoverage from the 1990 Census levels.
    Of particular interest, the national undercount for race 
and Hispanic origin groups was reduced from the 1990 census 
rate of 1.61 percent to 1.18 percent for Census 2000, a 
reduction of about 25 percent.
    Throughout the planning for Census 2000, a major issue of 
concern to the Census Bureau was whether the results of the 
Accuracy and Coverage Evaluation Survey could be used to make 
the census counts more accurate. A committee of Census Bureau 
professionals was formed to evaluate whether using the A.C.E. 
to adjust the census figures would improve the results for 
using redistricting.
    The professionals serving on the committee collectively 
have 390 years experience as statisticians and demographers. 
Bill Barron, the Acting Director of the Census Bureau, 
presented me with that committee's report and recommendation, 
along with his own recommendation, on March 1, 2001.
    Mr. Barron agreed with and approved the committee's 
recommendation that the unadjusted census data should be 
released as the Census Bureau's official redistricting data. 
The committee of Census Bureau professionals reached its 
recommendation because it was unable, based on the data and 
other information it had at the time, to conclude that the 
adjusted data were more accurate for use in redistricting.
    The report of the committee is being made available for the 
record. In a nutshell, the A.C.E. results could not be squared 
with the population estimates derived from Demographic 
Analysis, a long-accepted method of judging census quality. 
There were other significant potential errors that might affect 
the accuracy of the A.C.E. estimates.
    After receiving the Census Bureau's recommendation, I 
thoroughly reviewed the report and supporting materials, and I 
consulted with a diverse group of respected, non-Government 
statisticians and demographers in addition to Census Bureau 
professionals. On March 6, I announced my decision to release 
the unadjusted data for use in the redistricting process. In 
making my decision, I followed a process that was transparent, 
responsible, reasonable, and fair, and took full account of the 
view of experts, both career Government professionals, and 
professionals from outside Government.
    As you know, the Census Bureau is obligated to complete 
release of the redistricting data to the states by April 1, and 
it will do so. I should emphasize that the committee of Census 
professionals could not have resolved critical questions about 
use of adjusted data prior to that deadline, or even soon 
thereafter, I am confident that the committee did all that it 
could, and that it reached the only reasonable conclusion.
    Some have requested that the bureau release A.C.E. adjusted 
data notwithstanding the committee's recommendation to release 
unadjusted numbers as official redistricting data. The 
Department will not do so, because the A.C.E. numbers simply 
are not fit for use. From everything we currently know, the 
actual head count was extremely accurate. The A.C.E. projection 
would not be an improvement, and those estimates very well may 
change as a result of the bureau's ongoing analysis.
    It would be irresponsible for the Department to release 
redistricting data for which it cannot gauge the accuracy. Over 
the coming months, the bureau will gather additional data and 
continue its analysis of Census 2000 results. Before the end of 
2001, the bureau will determine whether or not to recommend 
statistically adjusting the Census 2000 population estimates 
for use in other Census Bureau products, and as the basis for 
annual adjustments that will commence in December 2001. Any 
such adjustment data may also be used in the allocation of 
funds under Federal programs and for other purposes. At the 
moment, there is really no basis on which to predict what the 
outcome of the bureau's further work will be.
    The Census Bureau is determined to build on the success of 
Census 2000, and the Bush Administration supports that effort, 
as reflected in the President's budget, details of which will 
be forthcoming. It funds continuing efforts for the Census 
Bureau to improve the census as we progress toward 2010.
    I look forward to continuing to work with Members of this 
Committee and other interested committees and Members of 
Congress to define and provide appropriate support for a total 
Census 2010 effort.
    Mr. Senator, that concludes my testimony, and I will be 
pleased to answer any questions you might have.
    [The prepared statement of Secretary Evans follows:]
   Prepared Statement of Hon. Donald L. Evans, Secretary of Commerce
    Mr. Chairman, Senator Hollings, and Members of the Committee:
    It is a pleasure to appear before this Committee again and, on this 
occasion, to discuss with you Census 2000.
                 census 2000: an invaluable achievement
    I have said that the 2000 Census is the most accurate census this 
nation has ever conducted. A Constitutional mandate, censuses have been 
conducted every 10 years since 1790--22 times in all. So the success of 
Census 2000 is a remarkable achievement. It is attributable to the hard 
work and dedication of the professional staff at the Census Bureau and 
all the hundreds of thousands of people, including thousands of your 
constituents, who worked on Census 2000. I commend them all. We are 
indebted also to the American public, whose response exceeded 
expectations; to the thousands of Census partner organizations; and to 
the Congress, for your oversight, support, and vision in providing 
sufficient resources to conduct Census 2000.
    This is an exciting period for those who want facts to bolster 
their understanding of our nation's people. The Census Bureau began to 
roll out the results of Census 2000 just three months ago, with the 
release of the state population totals used for apportionment. Again 
meeting the schedule set by the Congress, by the end of this week, the 
Census Bureau will have released for all 50 states, the District of 
Columbia, and Puerto Rico, population data--by age, race, and Hispanic 
Origin--that will be used to redraw legislative districts.
    Not only do these current data releases allow the redistricting 
process to begin, but these are the first data from Census 2000 for 
counties, cities, towns, townships, and smaller geographic areas. They 
are the first race and ethnicity data from Census 2000, and the first 
to show the effect of multiple race reporting. Not a day has gone by in 
the last month without numerous news stories about the increasing 
diversity of our population and about which areas are growing in 
population and by how much. I share the fascination: These data tell us 
so much about ourselves, our neighbors, our great country. This gold 
mine of information will continue to yield a wealth of information as 
the Census Bureau prepares and releases much more data over the next 
two years. All of this attention focused on population data reminds us 
what an important national resource we have in the census.
                  census 2000: an operational success
    Census 2000 was an operational success. The Census Bureau met or 
exceeded its goals, including meeting the mandated deadlines for 
releasing data for use in apportionment and redistricting. This success 
can be attributed to the Congress' commitment to provide full funding 
for a number of improvements, including unprecedented outreach programs 
to groups that historically had the highest undercounts:
     Marketing and Partnerships: First, the Census Bureau 
implemented a multi-faceted, aggressive marketing and partnership 
program to encourage householders to include themselves in the census, 
by completing and mailing back their census forms. Based on the 
experience of declining response rates over the preceding three 
censuses, the Bureau had anticipated that fewer households would return 
forms by mail in Census 2000. Partnerships--140,000 in all--with state, 
local, and tribal governments; community and advocacy groups; the 
private sector; religious organizations; educational institutions; and 
the Congress were key to building support and removing obstacles to 
participation in the census. The Bureau successfully implemented paid 
advertising for the first time in Census 2000, placing over $100 
million in media buys designed to educate and motivate the public to 
respond. Paid advertising also allowed the Census Bureau to target ads 
to groups that had been undercounted at higher rates in past censuses.
     Educating Families: As part of the Census in Schools 
program, the Census Bureau provided lesson plans, wall maps, and take-
home materials to classroom teachers so they could teach lessons on the 
census.
     User Friendly Questionnaires: The Census Bureau designed 
the questionnaires so that they would be easier to read and fill out. 
The Bureau also sent advance letters and reminder cards before and 
after the questionnaires were mailed out to increase response. The 
Bureau further offered multiple ways to respond, to ensure everyone had 
a chance to include themselves in the census. These included printing 
questionnaires in six languages and making available upon request 
materials in 49 languages to assist people in completing the 
questionnaire.
    These cumulative outreach efforts were successful. The expected 
mail response rate of 61 percent was significantly exceeded, reaching 
about 65 percent by the start of the field operation to follow up on 
homes for which a questionnaire was not returned.
     Staffing: The Census Bureau hired and retained enough 
highly skilled temporary staff, throughout the course of the census, to 
complete all operations on time. Because of a resourceful recruiting 
plan, research on pay rates and recruiting, and the attractive wages 
that the Census Bureau could offer because of the full census funding 
that the Congress provided, the Census Bureau was able to recruit some 
3.7 million job candidates and eventually hire 960,000 people over the 
course of the census. Over 500,000 worked on the operation to follow up 
on those homes for which a questionnaire was not returned, and, through 
their hard work, the Census Bureau was able to complete the enormous 
task of personally visiting 42 million homes slightly ahead of 
schedule.
     Quality Checks: Because of the timely completion of the 
follow-up operation, the Census Bureau had the time and resources to 
conduct other operations designed to improve coverage, including 
additional re-enumeration efforts in selected areas. The Census Bureau 
called these operations ``Quality Counts.'' Based on Census Bureau 
experience and using various quality indicators, the Census Bureau 
identified about 10 percent of the Nation's homes that it believed 
should be visited again in these review, verification, and clean-up 
operations. If it had not conducted these additional operations, the 
Census Bureau would have provided an incomplete enumeration of the 
population. The ``Quality Counts'' operations helped improve coverage 
and the census count.
     Technology: For Census 2000, the Census Bureau used 
digital imaging and optical-character recognition technology for the 
first time to recognize handwritten answers in addition to marked 
circles or boxes. This was a vast improvement over previous computer 
systems and allowed the Census Bureau to process the data faster and 
introduce quality assurance steps to be sure they had captured the data 
accurately. During the peak of questionnaire receipts, the Census 
Bureau's data capture centers processed 3.3 million forms a day. Each 
bit of information on the captured census forms was transmitted over 
secured lines to the Census Bureau headquarters, where staff performed 
quality control checks to ensure they had complete data. The improved 
data capture systems, with the ability to capture names, also meant 
that the Census Bureau could offer multiple options for responding to 
the census with confidence that it could find and remove duplicate 
responses.
                the result: a highly accurate headcount
    The operational improvements not only contributed to the ability to 
meet legal deadlines, but more importantly they also produced an 
improved count. The Census Bureau conducted an independent survey of 
approximately 314,000 housing units--called the Accuracy and Coverage 
Evaluation (A.C.E.)--that was designed to measure net census coverage. 
It was also designed to measure differences in coverage rates for key 
groups.
    The first chart attached to this testimony illustrates the 
remarkable job the Census Bureau did in counting people in Census 2000. 
According to current estimates from the A.C.E., Census 2000 achieved a 
net coverage rate for the total population of 98.82 percent. Even 
better, the estimated coverage rates for individual groups were also 
very high. The coverage rate for Non-Hispanic Blacks was 97.83 percent; 
for Hispanics, 97.15 percent; for American Indians and Alaska Natives 
on Reservations, 95.26 percent; for American Indians and Alaska Natives 
off Reservations, 96.72 percent; for Native Hawaiians and Other Pacific 
Islanders, 95.40 percent; for Non-Hispanic Asians, 99.04 percent; and 
for Non-Hispanic Whites, 99.33 percent. The A.C.E. results thus support 
the conclusion that Census 2000 achieved both reduced net and 
differential undercoverage from 1990 census levels.
    Attached are two additional charts showing estimated net undercount 
rates for key groupings of the population for 1990 and 2000. Chart 2 
shows net undercount rates for the total population and race and 
Hispanic-origin groups. Chart 3 shows net undercount rates for age and 
sex groups, owners, and renters.
    The A.C.E. estimates that the net national undercount was reduced 
from the 1990 census rate of 1.61 percent to 1.18 percent for Census 
2000, a reduction of about one-fourth. This reduction is substantial 
and reflects high census quality. The A.C.E. further found that not 
only was the net undercount reduced, but there was a substantial 
reduction in the undercount rates for certain groups and in the 
differential undercount. In 1990, minorities, renters, and children 
were differentially undercounted, that is, undercounted at higher rates 
than the population as a whole. While these groups still have higher 
undercount rates, the differential has dropped considerably.
    The estimated undercount rate for Non-Hispanic Blacks was cut by 
about half--it dropped from 4.57 percent in 1990 to 2.17 percent in 
2000; and the estimated undercount rate for Hispanics dropped by about 
40 percent from 4.99 percent to 2.85 percent. The undercount rate for 
American Indians and Alaska Natives on Reservations in Census 2000 was 
4.74 percent, a reduction of about 60 percent from the 12.22 percent 
published for 1990. For American Indians and Alaska Natives off 
Reservations, Native Hawaiians and Other Pacific Islanders, and Non-
Hispanic Asians, Census 2000 showed undercount rates of 3.28 percent, 
4.60 percent, and 0.96 percent, respectively. The undercount rate for 
renters has dropped from 4.51 percent to 2.75 percent and for children 
has been reduced by about half from 3.18 percent to 1.54 percent.
                 the question of statistical adjustment
    Throughout the planning for Census 2000, a major issue of concern 
to the Census Bureau was whether the results of the A.C.E. could be 
used to make the census counts more accurate. In June 2000, the Census 
Bureau Director preliminarily decided that using the A.C.E. for this 
purpose was generally feasible, but to reach a final decision, it would 
be necessary to consider operational data to validate the successful 
conduct of the A.C.E., to assess whether the A.C.E. measurements of 
undercount are consistent with historical patterns of undercount and 
independent demographic analysis benchmarks, and to review measures of 
quality.\1\ The Bureau has long used demographic analysis as an 
independent check on the quality of the count. Unlike the A.C.E., which 
is a sample survey, demographic analysis uses records and estimates of 
births, deaths, legal immigration, and Medicare enrollments, and 
estimates of emigration and net undocumented immigration to estimate 
the national population, separately from the census.
---------------------------------------------------------------------------
    \1\ Accuracy and Coverage Evaluation: Statement on the Feasibility 
of Using Statistical Methods to Improve the Accuracy of Census 2000.
---------------------------------------------------------------------------
    A team of Census Bureau professionals--called the Executive 
Steering Committee for A.C.E. Policy or ESCAP-was formed to conduct the 
evaluation to determine whether using the A.C.E. to adjust the census 
figures would improve the results for use in redistricting. After 
extensive meetings and staff work and the review of many analytic 
reports, the ESCAP completed its report\2\, and Acting Director Barron 
submitted that report and recommendation, along with his 
recommendation, to me on March 1, 2001.
---------------------------------------------------------------------------
    \2\ Report of the Executive Steering Committee for Accuracy and 
Coverage Evaluation Policy: Recommendation Concerning the Methodology 
to be Used in Producing the Tabulations of Population Reported to 
States and Localities Pursuant to 13 U.S.C. 141(c).
---------------------------------------------------------------------------
    As a member of the ESCAP and as Acting Director, Mr. Barron 
concurred with and approved the ESCAP's recommendation that unadjusted 
census data be released as the Census Bureau's official redistricting 
data. The ESCAP reached its recommendation because it was unable, based 
on the data and other information it had at the time, to conclude that 
the adjusted data were more accurate for use in redistricting.
    The ESCAP found that both the census and the A.C.E. were of very 
high quality. The primary reason for arriving at its conclusion that 
unadjusted data should be released was the apparent inconsistency 
between A.C.E. and demographic analysis. The demographic analysis 
estimates are significantly lower than both Census 2000 and the A.C.E. 
estimates for important population groups. The ESCAP investigated this 
inconsistency extensively, but in the time available could not 
adequately explain it. The ESCAP noted that the inconsistency between 
the demographic analysis estimates and the A.C.E. estimates is most 
likely the result of one or more of three scenarios:
     First, that the 1990 census and the associated coverage 
measurement methodologies together undercounted the population by a 
significantly greater amount and degree than previously believed, but 
that Census 2000 included portions of this previously unenumerated 
population.
     Second, that demographic analysis may not have accounted 
for the full population growth between 1990 and 2000.
     Third, that Census 2000, as corrected by the A.C.E., 
overestimates the Nation's population.
    The Census Bureau must further investigate these concerns before it 
can recommend that adjustment would improve accuracy of data for 
purposes other than redistricting. It is also investigating other 
potential errors that could affect the accuracy of the adjusted 
numbers. All of these issues are discussed in detail in the ESCAP's 
report, which we are making available for the record.
    After receiving the Census Bureau's recommendation, I thoroughly 
reviewed the ESCAP's report and supporting materials, and I obtained 
advice from a diverse group of prominent, non-government statisticians 
and demographers, in addition to the advice of the Census Bureau 
professionals. On March 7, I announced my decision to release the 
unadjusted data for use in the redistricting process. In making my 
decision, I followed a process that was transparent, reasonable and 
fair, and took full account of the view of career professionals and 
outside experts.
    I should emphasize that ESCAP could not have resolved the critical 
questions about use of adjusted data prior to the April 1 deadline for 
completing release of redistricting data to the states, or even soon 
thereafter. I am confident that the Committee did all that it could, 
and that it reached the only reasonable conclusion.
                     the road ahead to census 2010
    As I have said many times, Census 2000 is the most accurate in our 
nation's history. But we cannot rest on our laurels. The Census Bureau 
has already begun looking toward 2010.
    While Census 2000 was an operational success and produced data of 
high quality, the process was costly, many people felt burdened by 
having to answer the long form questions, and the census was constantly 
at risk due to insufficient early planning and development, and 
disagreement on the design. If the Census Bureau has adequate resources 
early to build upon the successes of Census 2000, then it can reduce 
operational risks for the 2010 census and explore ways to further 
reduce the undercount.
    In a letter of January 17, 2001, from Mr. Chris Mihm of the General 
Accounting Office to my predecessor, Secretary of Commerce Norman 
Mineta, Mr. Mihm announced that Census 2000 had been removed from the 
GAO's list of high-risk Federal government programs. That Census 2000 
was on this list is a reminder of the great challenges the Census 
Bureau faced and overcame in conducting a successful census. In his 
letter, Mr. Mihm stated:

          As the Bureau plans for the 2010 Census, it will be important 
        for the Department of Commerce to ensure that the Bureau 
        completes its evaluations of key census operations as planned, 
        and in a timely manner, explores innovative options that could 
        help ensure a cost-effective headcount in 2010.

    Completing Census 2000 evaluations will shed further light on what 
worked well or did not work in this census. To build on the success of 
Census 2000, to reduce operational risk, and to reduce the undercount 
even further, the Census Bureau must improve the accuracy of its 
geographic database and Master Address File, eliminate the long form 
from the decennial census by collecting those data in the American 
Community Survey, and reengineer the census process through early 
planning. The improved geographic systems will ensure that there is a 
complete and unduplicated address list and will facilitate automation 
and electronic data collection.
    In this regard, the American Community Survey will provide more 
frequent detailed data for small geographic areas and allow the Federal 
statistical system to keep pace with ever increasing demands for timely 
and relevant data. And it will revolutionize the way we take the 
decennial census by simplifying the 2010 census requirements and 
allowing the Census Bureau to focus exclusively on the basic count. 
However, early 2010 planning and development is necessary for a re-
engineered process for the 2010 census, taking advantage of 
opportunities provided by having improved geographic systems and the 
American Community Survey.
    As reflected in the President's budget, details of which will be 
released shortly, the Administration supports the Census Bureau's 2010 
efforts. I look forward to working with Members of this Committee, 
other interested Committees and Members of Congress, to define and 
provide appropriate support for the total Census 2010 effort. We cannot 
delay, as every day brings us closer to what will be an even greater 
challenge to capture our increasingly diverse, vibrant population.
    Mr. Chairman, that concludes my testimony. I will be pleased to 
answer any questions you may have.
[GRAPHIC] [TIFF OMITTED] 88017.001


    Chart 2.--Estimated Percent Net Undercount for Race and Hispanic Origin Groups: 2000 A.C.E.  and 1990 PES
----------------------------------------------------------------------------------------------------------------
                          2000 A.C.E.                                               1990 PES
----------------------------------------------------------------------------------------------------------------
                                             Net                    Net
          Estimation grouping            Undercount   Standard  Undercount   Standard     Estimation grouping
                                             (%)     error (%)      (%)     error (%)
----------------------------------------------------------------------------------------------------------------
Total Population in Households.........        1.18       0.13        1.61       0.20  Total population\1\
Race and Hispanic Origin:                ..........  .........  ..........  .........  Race and Hispanic Origin:
  American Indian and Alaska Native            4.74       1.20       12.22       5.29    American Indian and
 (on reservation).                                                                      Alaska Native (on
                                                                                        reservation)
  American Indian and Alaska Native            3.28       1.33  ..........  .........  .........................
 (off reservation).
  Hispanic Origin (of any race)........        2.85       0.38        4.99       0.82    Hispanic Origin\2\
  Black or African American (not               2.17       0.35        4.57       0.55    Black or African
 Hispanic).                                                                             American
  Native Hawaiian and Other Pacific            4.60       2.77  ..........  .........  .........................
 Islander (not Hispanic).
  Asian (not Hispanic).................        0.96       0.64        2.36       1.39    Asian and Pacific
                                                                                        Islander
  White or Some Other Race (not                0.67       0.14        0.68       0.22    White or Some Other
 Hispanic).                                                                             Race (not Hispanic)\3\
----------------------------------------------------------------------------------------------------------------
Notes: The race and Hispanic categories shown on the left side of this chart represent estimation groupings used
  in developing estimates based on the A.C.E. Survey and do not conform with race and Hispanic categories that
  will appear in the redistricting (P.L. 94-171) files and other Census 2000 data products. In developing the
  estimation groupings used to evaluate the coverage of Census 2000, the principal consideration was to combine
  people who were expected to have the same probability of being counted in Census 2000. Consequently, the race
  and Hispanic origin groupings used to create the A.C.E. estimates of coverage are exceedingly complex. For a
  complete description of the estimation groups, see DSSD Memorandum Q-37, which will be provided on request.
In general, American Indians and Alaska Natives (AIAN) are included in that category, regardless of whether they
  marked another race or are Hispanic. A few exceptions apply, especially for those who do not live on a
  reservation, on trust lands, or in an AIAN statistical area.
Similarly, Native Hawaiians and Other Pacific Islanders (NHPI) generally are included in that category, unless
  they lived outside of Hawaii and marked more than one race or marked Hispanic.
Hispanics are mostly in that category, unless they marked AIAN and lived on a reservation, on trust lands, or in
  an AIAN statistical area, or marked NHPI and lived in Hawaii.
People who marked Black or African American are generally in that category unless they fell in the categories
  described above; similarly those who marked Asian are generally in that category, unless they fell in the
  categories described above.
The final category includes most people who marked only White or only Some Other Race or marked three or more
  races but did not fall into the categories described above
The race and Hispanic categories shown on the right side of this chart represent selected population groupings
  used in conducting the PES and do not conform exactly with race and Hispanic tabulations that were released
  from the 1990 Census.
The data in this chart contain sampling and non-sampling error.
\1\ Includes household population and some Group Quarters; excludes institutions, military group quarters
\2\ Excludes Blacks or African Americans, Asian and Pacific Islanders, and American Indians on reservations.
\3\ Includes American Indians off reservations.


      Chart 3.--Estimated Percent Net Undercount for Age, Sex, and Tenure Groups: 2000 A.C.E.  and 1990 PES
----------------------------------------------------------------------------------------------------------------
                          2000 A.C.E.                                               1990 PES
----------------------------------------------------------------------------------------------------------------
                                              Net     Standard      Net     Standard
         Estimation Grouping (%)          Undercount    Error   Undercount    Error       Estimation Grouping
                                              (%)        (%)        (%)        (%)
----------------------------------------------------------------------------------------------------------------
Total population in Households..........       1.18       0.13       1.61       0.20  Total Population\1\
Age and Sex:                              ..........  ........  ..........  ........  Age and Sex:
  Under 18 years........................       1.54       0.19       3.18       0.29    Under 18 years
  18 to 29 years........................  ..........  ........  ..........  ........    18 to 29 years
    Male................................       3.77       0.32       3.30       0.54      Male
    Female..............................       2.23       0.29       2.83       0.47      Female
  30 to 49 years........................  ..........  ........  ..........  ........    30 to 49 years
    Male................................       1.86       0.19       1.89       0.32      Male
    Female..............................       0.96       0.17       0.88       0.25      Female
  50 years and over.....................  ..........  ........  ..........  ........    50 years and over
    Male................................      -0.25       0.18      -0.59       0.34      Male
    Female..............................      -0.79       0.17      -1.24       0.29      Female
Tenure:                                   ..........  ........  ..........  ........  Tenure:
  In owner-occupied housing units.......       0.44       0.14       0.04       0.21    In owner-occupied
                                                                                       housing units
  In nonowner-occupied housing units....       2.75       0.26       4.51       0.43    In nonowner-occupied
                                                                                       housing    units
----------------------------------------------------------------------------------------------------------------
Notes: The data in this chart contain sampling and non-sampling error; a minus sign denotes a net overcount.
\1\ Includes household population and some Group Quarters; excludes institutions, military group quarters.


    Senator Kerry. Mr. Secretary, thank you very much. We 
appreciate the testimony. Your testimony, incidentally, I read 
completely last night just to get a sense of it and, indeed, 
you appropriately point out the degree to which the census this 
year improved in many ways. I mean, there was a very 
significant effort to do things better, and I certainly want to 
acknowledge that.
    I don't think we should approach this in a contentious way 
that suggests there was not a bona fide effort to try to do the 
best census possible. The problem is that even those who 
engaged in the census, even those who undertook to do the best 
census possible, acknowledge, as I am sure you do, that there 
still is an overcount and an undercount. I mean, I assume you 
accept that even your current census is flawed to some degree. 
The question is, to what degree, is that correct?
    Secretary Evans. Right.
    Senator Kerry. So we know there is some overcount, the 
overcount generally being, according to most historical 
analysis, individuals who are generally white and affluent, 
because it winds up being kids counted twice who are away at 
college, people who have second homes, and those are the kinds 
of situations that lead to double counting, whereas minorities, 
poor people, new immigrants, people in the inner cities, tend 
by and large to be those historically have been undercounted, 
and I assume you would agree with that.
    Secretary Evans. That is what the data seem to reflect.
    Senator Kerry. Now, that said, Mr. Secretary, let me just 
kind of throw the sort of historical background of this on the 
table and ask you how we can proceed from here. This is a 
charged political issue, and I think the effort ought to be by 
all of us to try to defuse that tension, to uncharge it, but it 
has broken down largely along partisan lines. I mean, I just 
want to sort of lay it out on the table the way it is.
    Democrats have tended to support the use of statistical 
sampling, because it appears to be a method of more accurately 
counting the minority population, those undercounted, and I 
guess if I were telling the truth about it, by and large, most 
people would say, ``Well, those folks tend to be of sort of 
Democratic proclivity,'' whereas the people double-counted and 
the people on the upside who would just as soon see the formula 
not be as generous toward underserved areas have tended to be 
Republicans. Republicans oppose, Democrats have fought for it.
    Now, this started in the Reagan-Bush administrations. The 
issue was precipitated by the Census Bureau decision of 1987 to 
use statistical sampling in the 1990 census in order to correct 
for the undercount, and that is when the A.C.E. was first 
supposed to be incorporated in the Census, but President 
Reagan's Department of Commerce overruled the bureau and 
attempted to cancel the funding for sampling. The funds were 
eventually restored, and plans for use of the method proceeded.
    Then in 1991, President George Bush's administration 
appointed a Census Director, Barbara Bryant, with the support 
of a 7-2 decision by the Census board, recommended to then-
Commerce Secretary Mosbacher that statistical sampling be used 
to correct the 1990 census. That recommendation was rejected, 
and the 1990 census was the first census to be less accurate, 
as a result, than the previous one.
    The matter was then taken up by Congress, resulting in the 
1991 passage of legislation directing the National Academy of 
Sciences and Census Bureau to design a plan with higher 
accuracy, and they came up with the extensive statistical 
sampling plan.
    I might add that many Republicans, including Senator John 
McCain and Congressman Gingrich, strongly supported statistical 
sampling, and when the Republicans took over Congress after 
1994, that is when you began to see a shift in their position.
    Now, I say all of this as background, Mr. Secretary, simply 
to try to see if there is not some way in this new year for a 
new administration and this Congress to diffuse some of the 
politics of this, and to try to present America with a real 
bipartisan approach that has the confidence of everybody that 
it is not one side trying to protect its interest versus 
another, but rather, as Ranking Member Hollings said, we are 
trying to get the best count we can and the most available data 
that gives people the best confidence in the country, and I 
wonder if you might share with us, therefore, your sense of how 
you proceed from here.
    While I am not going to argue with you that you received a 
recommendation from an Acting Director to proceed in a certain 
way, the rationale was that they felt they did not have time to 
resolve the issues, and I wonder if you would therefore now say 
to the Committee that you are prepared to give them the time, 
get this analysis and discrepancy between the data between the 
A.C.E. and the Demographic Analysis resolved so that we could 
see if we can find a bipartisan acceptable understanding of how 
we proceed from here.
    Secretary Evans. Senator, thanks for putting it in 
perspective, and your points are well-taken. We have the most 
accurate census in history, as I said in my remarks, vast 
improvement from the 1990 census, and I congratulate and salute 
all of those that were involved in the effort for their 
efforts. I look at--and as you acknowledged, I think, as I said 
in my remarks when I was here in January, I would look at this 
with an open mind, transparently, would listen to the bureau 
and what their recommendations would be, and that is what we 
did. We listened to the professionals, and the professionals 
made their recommendation to us and to me and I accepted that.
    Senator Kerry. The recommendation is temporary.
    Secretary Evans. I have said, and I said in my testimony 
that they will continue to evaluate the data, and they will 
continue to make assessments as to whether or not statistical 
adjustment will make the data, could make the data more 
accurate.
    My point was going to be that the lower the undercount is, 
the more difficult it is to say that you can statistically 
adjust the data to a more accurate level. But will we take more 
time to evaluate the data? Yes, we will. I have instructed Bill 
to do that, and he has indicated to me that it will be sometime 
this summer before they will be willing to be ready to come 
back to me with a recommendation as to whether or not the data 
should be statistically adjusted.
    Senator Kerry. Well, let me try to put that in perspective. 
Number 1, you do agree that ESCAP did not decide that the 
uncorrected data is more accurate than the corrected data. That 
decision was not made, correct?
    Secretary Evans. Repeat that, Senator.
    Senator Kerry. That the uncorrected data is more accurate 
than the corrected data. That decision has not been made.
    Secretary Evans. Well, what ESCAP said was that at the 
point that we made the decision, that they could not conclude 
that the A.C.E. data would be more accurate than the data.
    Senator Kerry. Well, quoting from the ESCAP report, it 
says,

          ``While the majority of the evidence indicates both a 
        continued existence of a differential undercount of the 
        population and the superior accuracy of the corrected numbers, 
        the Committee has concerns, and until these concerns are fully 
        investigated and addressed, we cannot recommend using it.''

    Secretary Evans. They could not recommend to me that it 
would be more accurate to use the A.C.E. data at that point.
    Senator Kerry. But at this moment we are, by everybody's 
understanding, living with an undercounted number of Americans. 
We know that. At this moment we know there are an undercounted 
number of Americans, and the question is, are we going to find 
a way to adjust for that, or are we prepared to live for the 
next 10 years knowing we have undercounted a number of 
Americans.
    Secretary Evans. Right, and that is what we are going to 
continue to work on and see if we can come up with a more 
accurate number.
    Senator Kerry. Would you make the following commitment to 
the Committee: Would you make the commitment that we will have 
a resolution of whether or not your department will use 
corrected data for the purpose of allocating Federal funds? I 
mean, that is really what this fight is about, and the question 
is going to be whether or not we are going to have the 
opportunity for those Federal funds to be distributed in the 
most fair basis possible. I am not asking for an overcount.
    Secretary Evans. I understand.
    Senator Kerry. No one here is asking you to overcount, but 
we simply know we cannot live with an undercount.
    Now, can we have a commitment that this Committee will have 
a chance to review with you that decision before those Federal 
funds are allocated?
    Secretary Evans. Yes, you do.
    Senator Kerry. And you would come back to us?
    Secretary Evans. Yes, I will.
    Senator Kerry. I think that is very important. I appreciate 
that enormously. I mean, look, it is in all of our interests to 
do this as fairly as possible, and resolve the difference 
between the statistical results. There are some people arguing 
the Demographic Analysis may be in error, and the A.C.E. may, 
in fact, be more accurate, and the question is, if that is 
true, there may be a way to resolve that, and I suppose--my 
colleagues are here and I want to give them an opportunity 
also, but let me just ask you this. Would you be prepared to 
release the A.C.E. data down to the block level so that 
scientists across the country can help make this judgment in 
resolving any undetected problems?
    Secretary Evans. Let me turn to Bill and let him respond to 
that.
    Mr. Barron. Senator, we think at this juncture to release 
the data down to the block level would be misleading to folks. 
In fact, from the experts I have talked to, I do not believe 
they think they need data at the block level to make this 
decision, but if someone believes that to be the case we would 
be happy to entertain that request. We provide a lot of data to 
the National Academy of Sciences and others.
    Senator Kerry. Well, I assume someone could FOIA it.
    Mr. Barron. I suppose they could, Senator.
    Senator Kerry. Would it not be better to do this in a 
cooperative, open way? I mean, if the data in fact supports the 
conclusion that the quality was generally good, I would think 
you would want that to be thoroughly analyzed and support your 
conclusion.
    Mr. Barron. Block-level data are notoriously noisy data 
sets. Throughout most of this process I have had to defend 
every block-level adjustment, and we have had to say at that 
level the A.C.E. does not improve data.
    You have to start aggregating data above blocks to see the 
improvement that dual system estimation, or sampling, as it has 
become known popularly, brings to the process, so by 
emphasizing block-level data we are really taking the worst 
part of the whole process and putting it out on the street, and 
at this juncture, since we have no confidence in those numbers, 
we do not want to do that.
    Senator Kerry. Well, let me come back to that. I want to be 
fair to my colleagues.
    Senator Brownback.

               STATEMENT OF HON. SAM BROWNBACK, 
                    U.S. SENATOR FROM KANSAS

    Senator Brownback. Thank you very much, Senator Kerry.
    Mr. Secretary, thank you for being here. First, thanks for 
the great picture that you are showing America. It is on a 
regular basis the people who are reviewing this data, and it 
really helps us to look and see who we are today, and it 
provides the data and the background. I think that is a very 
valuable thing. We can discuss political issues here, and those 
are important to discuss, but the picture you are showing to 
America is far more important, what the people are seeing, and 
I think you have done a very nice job of that.
    One thing I want to ask you about in looking forward on the 
census and looking at the next census, and on your American 
community survey, was this year, this census for the first time 
since I think we have conducted, or at least since 1880, we did 
not ask on the short form about marital status, the first time 
since 1880 that we have not asked that, and the reason I want 
to draw to your attention is that family structure has such an 
important impact on what takes place to our children, what 
environment our children are working in.
    This is really very important and useful information for 
people to be able to have, to study, and to look forward to 
what is going to happen to our families and what is going to 
happen to our children. I am hopeful that as you look forward 
to the 2010 survey on the short form you will reinstate that 
marital question. There was a vote in the Senate last year 
unanimously in favor of doing that and reinstating that marital 
status question, and I am hopeful that you will do that.
    Any thoughts? Have you had a chance to look at that?
    Mr. Barron. Senator, I was not at the Census Bureau when 
the final decisions on the short form were made, but certainly 
what you are asking for on a go-forward basis is exactly what 
we intend to do. We will be working with the Congress, and I 
think the Office of Management and Budget is coordinating that, 
but we have heard you, and we will be working with the Congress 
to decide what the questionnaire is going to look like for 2010 
and in the A.C.E.
    Senator Brownback. Good. I hope you will look at that 
issue, because it really does have a huge impact. The second 
item is on your American community survey, which as I 
understand it is an ongoing, almost rolling survey that 
continues to be working. I have been particularly concerned 
that the census long form did not do an adequate job of 
gathering the family structure in the country, such that we 
have lost valuable data on what our families look like across 
the country.
    I want to draw your attention--I would ask you again here 
what steps are being taken to ensure the American communities 
survey improves and enhances data collection on family 
structure in America, and I would not be so pointed on this, 
except for this impact on what is happening to our children and 
what happens on our school system and what happens in crime, 
within all of the various things that impact our society and 
impact our Government. We need that information in a solid 
form.
    Any thoughts on what you are looking at on the family data 
either, Mr. Secretary, or Mr. Barron?
    Mr. Barron. The answer would be the same, Senator. In the 
2000 census there was a very narrowly applied rule that if it 
was not explicitly required by law, or to support a legal 
purpose, then the questions were not asked, but this time we 
will be happy to work with the Congress to see if a broader 
approach could be taken, and we will obviously keep your 
thoughts in mind as we go through that.
    Senator Brownback. If you could. Senator Moynihan, before 
he left the Senate, and I worked closely on statistical 
gathering, and its impact on the country, and his point was 
that until we figured out a way to measure economic data like 
unemployment, inflation rates--we didn't know how to change 
things.
    To be able to change something you have go to be able to 
measure it, and probably one of the most valuable things we can 
do at the Federal level is to be able to provide accurate 
measurements, whether it is on economic factors or social 
factors, and you probably have the lead tool for us on social 
factors of what is taking place across the country, and for us 
to be able to move in a positive direction socially we have got 
to have better numbers here and be able to provide it on a 
state-by-state, major geographic basis, and then let 
communities react to those numbers.
    That is probably the most valuable thing we could do, and 
he and I were very strong on pushing that point. Thank you. I 
know we are under a time line, and we are under a vote, so I 
would yield to my colleague from California, or the chairman, 
and I will look forward to working with you on these two 
issues.
    Senator Kerry. I am going to go vote, and I will be right 
back, Mr. Secretary.
    Senator Boxer. Senator, do you want to put us in a recess?
    Senator Kerry. If necessary.
    Senator Boxer. Well, I do not want to miss the vote.
    Mr. Secretary, welcome. It is nice to see you, and nice to 
see you, Mr. Barron, and I look forward to working with you on 
this and many other issues. I will be brief, because we do have 
this vote, and Senator Brownback, please do not feel obliged to 
listen if you have to leave. It may not be your favorite point 
of view here.
    Well, my overriding principle has been from day 1 that we 
need the most accurate count, and I know we all feel that way. 
I have reasons to feel that way that go beyond just the 
fairness questions to real pragmatics, which is what happens to 
California when there is an undercount. With a population of 34 
million people, we get the worst of it. We usually do.
    I will explain what happened in the last census and how 
many dollars were shorted to my state for children, old people, 
transportation, and all the things we need. I do want to 
commend the bureau for producing a more accurate census in 2000 
than was produced in 1990. However, I do share the concerns of 
many of my colleagues, Congresswoman Maloney and others, that 
the accuracy in coverage evaluation (A.C.E.) used to measure 
the accuracy of the raw counts shows there still remains a 
significant differential undercount for our minority groups 
compared to others.
    That means that in my state, which has large minority 
populations, many urban communities will not receive the 
Federal funds they need desperately for basic programs, and 
their residents will lack the proportionate representation that 
is their basic right under the Constitution.
    The Census Bureau has admitted that in the 2000 census at 
least 6.4 million people were missed, mostly in the cities, and 
3.1 million others were double-counted, mostly in the suburbs. 
According to a report in the New York Times, those numbers 
could be higher, with the bureau having missed as many as 7.6 
million people, and double-counting 4.3 million others.
    Now, for my state, that is just not an acceptable result. 
As a result of the 1990 undercount, California lost nearly $214 
million in Federal funds, and the city of L.A. alone lost $120 
million in Federal and state funds. Millions of dollars are at 
stake for educating our children, for title I, for health 
services, for the poor through Medicaid, and for a multitude of 
other services like child care and day care that allow parents 
to go to work and have peace of mind.
    For the 2000 census, $400 million was spent on improving 
the A.C.E. and $700 million spent on improving the raw count. 
That is a lot of money, and Iam troubled by the fact, Mr. 
Barron, that you are not interested in releasing these numbers. 
They are not the secret property of the Census Bureau. They 
belong to the American people. The American people spent a lot 
of money, and I think these numbers belong to the American 
people. I would like to see you reconsider your decision. I 
think it is going to happen anyway. It is a democracy. It is 
going to happen.
    Second, I would love to see the Census Bureau move with 
immediate haste to resolve the issues that have distorted the 
numbers and provide Congress with adjusted numbers well before 
the fall deadline for determining the distribution of Federal 
funds. As I say, these are not luxury programs. They are very 
necessary services to people who are trying to grab the 
American dream. They need these funds.
    I certainly look forward to working with Secretary Evans. 
He and I have a nice working relationship. I want to keep it 
that way, and with you, Director Barron, and with my colleagues 
in Congress, and with organizations like the National 
Association of Latino Elected and Appointed Officials, the 
Mexican American Legal Defense Fund, so that we can reach a 
consensus on an adjustment to the census that reflects an 
accurate count.
    I have questions. What I will do is place them in the 
record, because immediately following the vote, I am due at a 
conference on the environment. I have a few disagreements on 
little things like arsenic and lead, so I will not be 
returning. However, I will be submitting these questions for 
the record, and again, I make these requests in friendship, in 
a spirit of cooperation.
    I think we still have time to work together to put 
everything out on the table and try to do our best to get an 
accurate count. Our people deserve it. I think you know that. 
USA Today reports that 350,000 people in Texas could have been 
undercounted. In New York, 300,000 may have been ubdercounted; 
in California a \1/2\ million. These are not little side bars. 
These are living human beings who, for the most part, deserve 
the services that we provide.
    So thank you very, very much, and we will stand in recess 
until Senator Kerry returns, which should be any minute now. 
Thank you very much.
    [Recess.]
    Senator Kerry. The Committee will come back to order, 
please.
    Mr. Secretary, thank you. You have been very kind to hang 
in there. I am not going to keep you more than just a couple of 
minutes, because I know you are already late for your meeting.
    Just one question. Mr. Secretary, I think we spent 
something like $7 billion on the census itself, and some $400 
million was spent on the accuracy coverage evaluation. What 
does the bureau allocate for the Demographic Analysis model?
    Secretary Evans. I do not know what that number is, do you?
    Mr. Barron. I do not. It is clearly smaller.
    Senator Kerry. It is considerably smaller.
    Mr. Barron. I think it is $250 million for the A.C.E.
    Senator Kerry. Not that the accuracy is obviously measured 
by the amount of money we spend, is it not? I mean, is there 
some reason to believe that, in fact, we may have a more 
accurate A.C.E. model now than we do Demographic Analysis at 
this point? Is that conceivable?
    Mr. Barron. I think the value of Demographic Analysis, 
Senator, in part is that it is an independent check on the 
quality of the A.C.E. It is not a system that has changed a lot 
over time, but except for one very important omission that I 
can talk about, the base system of adding births and deaths is, 
we think, a pretty solid system. There is an issue with exactly 
what the solution is to improve that in terms of money. We are 
looking at that, but I do not know that on a go-forward basis 
we are going to have a proposal that we are seeing, if money 
would improve this problem. We are not sure it would.
    Senator Kerry. Do you see any problems in terms of the 
timeframe we are looking at before we have to make judgments 
about the distribution of funds that would be able to come to a 
public conclusion that can be sort of put to the test with 
respect to trying to resolve the discrepancies between data 
sets?
    Mr. Barron. Senator, I think unlike the period we just went 
through, where in essence we had to make a decision by March 1, 
I think that by fall we are going to be able to go to the 
Secretary with an analysis of the issues we encountered, not 
just on Demographic Analysis, but also with the accuracy and 
coverage evaluation survey, and make a recommendation. I am 
confident we are going to be able to do that. If not, then I am 
going to have to come up and tell you that I am not, and why I 
am not.
    Senator Kerry. Well, Mr. Secretary, can I make a suggestion 
that is made in completely good faith to try to resolve this? I 
can understand why you made the decision you made, given the 
input you had, given the time issue and you obviously had 
pressure on you with respect to redistricting, needless to say, 
you have got to make a decision. But on the other hand, I would 
hope you could also understand the measure of concern and even 
skepticism that a lot of people have about what may happen now, 
as we proceed down the road.
    It seems to me it is very much in your interest--I mean, to 
look at the California potential lawsuit now, you look at the 
questions raised, it just does not serve any of us well to have 
the country at odds over the methodology by which we determine 
how many folks are living here.
    There ought to be some way of people acting in good faith 
to come up with an agreement that the data is trustworthy, that 
the methodology is trustworthy, and all we are talking about is 
a fair count. I can assure you there is no way for me to be 
advantaged or for the folks you are arguing for to be, quote, 
advantaged in this because we are not going to get an 
overcount.
    The only issue is, will the overcount issue be addressed in 
a way that gives confidence to people this has been done fairly 
for the purpose of the next 10-years distribution. I would 
think you would want it to be done to the greatest capacity 
possible, and I sense you do, but I just want to encourage you 
to consider the notion that if the data is out there, subject 
to analysis, it will be subject to different people's analysis. 
It is better, probably, to suffer that than to have people 
saying you are hiding something, or you are unwilling to submit 
it to scrutiny, so I would strongly urge you to let folks kind 
of pore over it.
    I do not think that is going to resolve the issue 
completely, but hopefully we can work together, Mr. Secretary, 
in the next months to come to some conclusion about how we can 
avoid any controversy over this.
    Secretary Evans. Senator, we will give that consideration 
and give it some thought. We share the same ultimate goal. We 
both want the American people to believe that the data that we 
have released are fair and are accurate. But there is another 
component to just the count, and that is the location, where 
these people live. When you get into the statistical analysis 
and that portion of adjustment, it gets complicated and 
complex, and those are some of the issues that I know that the 
professionals struggle with.
    It is not only, having to say how many people, but you have 
to determine where they live, all the way down to the block 
level. So know that you and I share the same goal. We both want 
the American people to have the fairest and most accurate data 
that we can provide them, and that is what I intend to do.
    Senator Kerry. Well, I appreciate that commitment, and 
obviously, I mean, this is something people have struggled with 
for long periods of time. I think the statistical sampling has 
become remarkably sophisticated. It is subject to huge amounts 
of scientific statistical analysis and tends to be objected to, 
frankly, on the partisan lines I have described That is its own 
message, and I think we need to be particularly sensitive to 
that message as we go forward here.
    We have the scientific know-how about how to make up for 
this undercount without at all abusing the other side of the 
coin, if you will. There are ways even to weight that 
statistical sampling to guarantee that you don't abuse it, so 
it seems to me that the test here is the test of good faith, 
and the effort to try to not live another 10 years with an 
undercount that everybody acknowledges exists.
    There must be some way to make up that undercount fairly 
and thoughtfully. That said, Mr. Secretary, we have taken more 
of your time than you had originally agreed to, and I am very, 
very appreciative for your doing so, and I apologize to whoever 
it is you are keeping waiting.
    Thank you very much.
    Secretary Evans. Thank you, Senator, very much.
    Senator Kerry. What we will do is go to the third panel, 
and we will just announce ahead of time we will interrupt that 
panel the minute our Members of Congress come back. They have a 
vote, and obviously have a distance to travel, so if we could 
invite Dr. Ericksen, Dr. Murray, Mr. Vargas, and Dr. Wachter, 
we will commence, and then we will interrupt, as I said, at the 
appropriate time.
    Gentlemen, thank you very much for coming. I appreciate 
your participation in this hearing very much. I notice the 
hearing room half-emptied as the Secretary left with members of 
the Commerce Department. I do not know if we had the whole 
Census Bureau here or what.
    Dr. Murray, why don't we begin with you, and we will just 
run right across the table, and I appreciate your being here.

          STATEMENT OF DR. DAVID W. MURRAY, DIRECTOR,

               STATISTICAL ASSESSMENT SERVICE AND CENSUS

                   MONITORING BOARD, CONGRESSIONALLY

                            APPOINTED MEMBER

    Dr. Murray. Thank you, Senator. I am Dr. David Murray. It 
is a pleasure and honor to be in front of you, Senator Kerry 
and Members of the Committee. I am honored to appear before you 
to discuss the 2000 decennial census.
    I serve as director of the Statistical Assessment Service 
here in town. I am also here in my capacity as a 
congressionally appointed member of the U.S. Census Monitoring 
Board. As you know, Congress created this bipartisan panel to 
observe and report on preparation and implementation of Census 
2000, and I have written testimony which I will submit, and I 
also request one thing, Senator. In my written testimony, I 
alluded to a report that our monitoring board put out. I 
brought a copy of that, and if that is acceptable I would like 
to also include that as part of my testimony.
    Senator Kerry. Absolutely. We will make the full report a 
part of the record.
    Dr. Murray. Coming to grips with technical issues in the 
census is obviously a challenge to all of us, and I think in 
particular for Senators who have to quickly come up to speed on 
the issue. I tried to prepare in my mind an analogy as to where 
we stand, and I hope this is a reasonable and homiletic 
expression of what our difficulty is.
    Every 10 years, as it were, the country has to balance its 
checkbook.
    Senator Kerry. Pull the mike up a little bit.
    Dr. Murray. Every 10 years, sir, the country has to balance 
its checkbook, has to bring its accounts into rectification. It 
has three ways of measuring and looking at what our balance 
might be, and we have to do that with the census. Our 
fundamental problem is, we cannot balance our checkbook this 
time. The three different measures we have give us a different 
set of results.
    Now, we can count our dollars. That's the enumeration. That 
gives us a sense. We can go back and look at the bank records 
and our check register. That is what the Demographic Analysis 
(DA) is roughly comparable to. It gives us a very different 
picture of what our true balance is. Or we can take a sample of 
about \3/10\ths of 1 percent of the households of our account 
and try to correct it based on that.
    We have three fundamentally different measures that in some 
measure all of them have strengths and weaknesses. Each one has 
a weak point. Each one has relative strengths, but they differ 
from each other in a fundamental and inconsistent way that we 
found could not reconcile, and as such we do not necessarily 
have one number that is better than another.
    We do not necessarily know that we can correct the census 
by using, for instance, the A.C.E. What we have is different 
numbers. The A.C.E. gives us a different number than the 
enumeration does. The DA gives us a different number than the 
enumeration does. Given that circumstance it seemed the most 
prudent course of action was to stick with the one, the devil 
we knew, rather than the devil that we don't, and it was a 
prudent and wise decision to go with the enumerated numbers as 
the most accurate that occasion the fewest number of costs to 
the system.
    In some sense the A.C.E. began to be problematic, as is the 
DA, and the sense that the cure may have been worse than the 
disease, and let me just say a couple of formal things here. 
Where do we stand with an estimated undercount of only 1.18 
percent? That is, we have a census that at the moment is 98.82 
percent accurate. We should realize that the cost-benefit ratio 
of our respective choices begins to shift.
    Given all the attendant legal and political difficulties 
that the A.C.E. engenders, its saving grace was that it might 
be a good technical fix, but now that is in question. The 
A.C.E. appears to have levels of statistical noise in its 
probability fluctuations that are greater in magnitude than 
this signal it was designed to detect and correct, that is the 
1.18 percent undercount is smaller in magnitude than the error 
range that we use out of the statistical sample. The 
probability fluctuation is greater than that.
    Applying the adjustment to a census as accurate as the one 
we have just completed begins to slide down the slope of 
diminishing returns, technically, as well as in terms of 
governance consequences. We face a genuine dilemma which the 
ESCAP report issued at the end of February well captured. The 
actual enumerated Census 2000 count placed us in a new and 
perplexing landscape, one unanticipated by the designers of the 
A.C.E.
    The count of 281.4 million surpassed the best reckoning of 
the population given by the Demographic Analysis by nearly 2 
million people. This is an anomalous outcome, since the DA has 
traditionally stood above the enumerated count and told us 
about the magnitude of the undercount in the enumeration.
    Now, we have broken through the DA measurement. By the 
reckoning of the DA we actually have an overcount as the 
measurement that we get from the DA and the enumerated census. 
Perhaps, on the other hand, the enumerated census was, in fact, 
a huge achievement, diminishing the undercount substantially, 
so much so that it is the most accurate number we have.
    Various scenarios to explain the irreconcilability have 
been proposed by the ESCAP report and by others, but no 
scenario has achieved a reconciliation without introducing yet 
other anomalies. We are somewhat like the person who has a 
sheet over us. We pull the sheet up to cover our chin, and find 
our feet exposed. We pull the sheet down to cover our feet, we 
find our chin exposed. Every statistical adjustment that we try 
to bring to bear produces new anomalies and new problems.
    The emerging and tentative adjustment from the A.C.E. took 
us even further away from solid ground, with an estimated 
result for approximately 284.7 million. The gap with the 
enumeration count is substantial, while the gap between A.C.E. 
and DA is fundamentally problematic, too large, in fact, to be 
reconciled by any scenario yet deployed. Moreover, the A.C.E. 
itself contains internal anomalies that are difficult to engage 
with.
    A.C.E. may be a valuable tool that tells us things about 
the undercount, that tells us things about the quality of the 
census, but the A.C.E. data cannot be shown to be more accurate 
than the numbers which they would replace, the enumerated 
numbers. The fundamental requirement of the A.C.E. was that it 
be superior to the data which they would supplant. The deepest 
fear is that using the A.C.E. numbers introduces error and 
irreconcilability, particularly at the local level, where 
apportionment decisions must be made in redistricting.
    As such, it may actually be a cure worse than the disease, 
producing more problems and more anomalies than we already 
have, hence the prudent idea that we go with the unadjusted 
data as the best course of action, and learn the lesson that 
basically an intensified census at the local level has a 
stronger chance to correct the fundamental problem we address, 
the undercount.
    We all care about the undercounted. We want to attain 
incorporating them into the system. It looks as if the best 
alternative way of doing that is to intensify and strengthen 
the census count itself, and not resort to an unproven method 
of probability that introduces more problems than it solves.
    Thank you.
    [The prepared statement of Dr. Murray follows:]
   Prepared Statement of Dr. David W. Murray, Director, Statistical 
    Assessment Service and Census Monitoring Board, Congressionally 
                            Appointed Member
    Mr. Chairman, Senator Hollings, Members of the Committee, I am 
honored to appear before you today to discuss the 2000 decennial 
census. I serve as the Director of the Statistical Assessment Service 
and I am also here in my capacity as a Congressionally Appointed Member 
of the U.S. Census Monitoring Board. As you know, Congress created this 
bipartisan panel to observe and report on the preparation and 
implementation of Census 2000.
    Rarely has a policy dispute generated a ratio of heat-to-light 
greater than did the just-concluded census dispute over the use of 
statistically adjusted data for apportionment of political power. 
Because the matter is dauntingly technical, the press faced a serious 
challenge in telling the story, no matter how many explanatory data 
charts and expert consultations were available. Nevertheless, even 
given the difficulty of the task, the media's performance over the last 
three years was (except for a couple of stories from the Washington 
Post and the Los Angeles Times) uniformly disappointing.
    Most succumbed to the temptation to cast the story as purely 
political, with potential winners and losers resorting to raw clout as 
they disdained scientific accuracy. In general, those who favored 
adjusted numbers were characterized as seeking to ``count every 
American'' (and incidentally aid their party's representation), even 
though their proposal was in fact to estimate, rather than actually 
count, missing people. Alternately, those who expressed doubts were 
cast as opponents of ``modern scientific methods'' seeking to preserve 
political advantage by deliberately ignoring missing people, 
principally members of minority communities and children.
    As the Houston Chronicle (Feb. 22) editorialized, ``Some ideologues 
oppose correcting the numbers . . . The opponents of statistical 
analysis are mistaken, of course . . . Those who oppose adjustments . . 
. either do not understand arithmetic, or they understand it all too 
well.'' The ``whose ox is gored'' story line built around putative 
political motives nearly always won out over real engagement with the 
technical complexities. As a consequence, numerous myths about the 
census adjustment process and its supposed consequences were introduced 
into the media bloodstream.
    In the resultant morality play, adjustment advocates usually came 
off as earnest advocates for the poor, who could be aided by a simple 
application of statistical justice. Those who favored an enumerated 
count, on the other hand, were often cast as stubbornly refusing to use 
a readily available technical means to solve a social problem--
``correcting'' the undercount by statistics. Lost in the fracas were 
genuine arguments about the feasibility and advisability of supplanting 
the standard enumeration with these technical means--a position 
ultimately validated not only by the Supreme Court decision of January, 
1999, but as well on February 28, 2001 by the decision of the Census 
Bureau itself. The enumerated count prevailed for good technical, not 
political, reasons.
    At first glance, the undercount problem should have a simple 
solution. In the 1990 census, the net undercount was roughly 4 million 
people, about 1.6 percent of a population of 248 million. That is, 98.4 
percent were properly enumerated (in contrast, the 2000 census missed 
approximately 3 million, which represented just 1.18 percent of 281 
million people--a 25 percent improvement). We knew about the undercount 
because we could compare the enumeration to higher figures from 
Demographic Analysis, which were regarded as more accurate. We could 
have saved a lot of money (the 2000 census cost over $6 billion, much 
of which goes to finding that last percent) by simply adding a 1.6 
percent ``correction'' to the overall population and calling it quits. 
But the census, unlike other government data, needs to know two things 
about Americans--how many in the aggregate, and also, how many in 
geographical (and demographic) distribution, in the smallest geography/
detail.
    That is, accuracy means not only getting the total count right, but 
positioning people where they actually reside, so that apportionment of 
political power can be congruent with their actual presence. And now 
the social problem gets tricky, because the undercount (either the1990 
1.6 percent or the 2000 1.18 percent) is not evenly distributed 
geographically (it tends to cluster in a handful of counties 
nationwide--mostly dense urban ones), nor is it evenly distributed 
demographically. This becomes the heart of the challenge.
    In general, the likelihood of being undercounted is thought to be 
related to being identified in various racial/ethnic groups (among 
other factors, such as home ownership). Members of minority communities 
are more at risk for undercount, other things being equal, than are 
non-Hispanic Whites. This means that the 1990 undercount of 1.6 percent 
was actually composed of a 0.7 percent undercount rate for non-Hispanic 
Whites, a 4.6 percent rate for Blacks, a 5.0 percent rate for 
Hispanics, all the way up to an estimated 12.2 percent rate for 
American Indians on reservations. Because there are legal triggers 
involved in these disproportions--there has to be a differential 
adjustment distributed proportionally, not a uniform add-back. Doing 
this correctly, so that one actually improves accuracy rather than 
introducing more problems into the count, is an enormous mathematical 
challenge.
    Here we encounter the first media-generated myth--that the 
statistical adjustment was based on a proper sampling methodology, like 
we find in political poll taking, which could then be used to 
``correct'' the undercount. This is only partly true. The actual 
process of determining who was likely missed in the census derives from 
wildlife biology, where it is known as a ``capture/re-capture'' form of 
``dual system estimation'' (DSE). Want to know the number (and the 
species proportions) of fish in a lake? One could drain the lake and 
count the bodies, but a more viable process is to cast a net, capture 
and count a sample of the fish, and then tag them. After the fish are 
released, one makes another cast and re-captures some of the tagged 
fish in another sample. By comparing the two catches, we can figure out 
the ratios of those caught in this ``dual'' system and make good 
estimates about the real population of the lake.
    Of course, the system isn't perfect. Some types of fish are likely 
to be missed in both the first and the second net cast. These are 
termed the ``wily trout'' about which we can only make indirect 
guesses, if we already suspect that they're ``really there.'' So why do 
we suspect that they are really there in the population? Largely by 
comparing our enumeration to another measurement, vital statistics 
records such as birth and death certificates, which tend to give us a 
higher count of the population than those enumerated. These record 
provide what was introduced above, the Demographic Analysis (DA--about 
which more in a moment), which serves to indicate our likely shortfall. 
(Some have wondered why we do not just rely on a Demographic Analysis-
type census in the first place, based on a variety of administrative 
records. The thought deserves consideration, but there are problems. 
For instance, the DA itself depends upon an estimate, since the number 
of those immigrating and emigrating must be modeled.)
    Granting that there is an undercount, what is the best response? 
Some advocated intensifying the enumeration, trying to reach all 
quarters (or at least substantially reducing those missed). They have 
been substantially vindicated by the 2000 outcome, which saw the 
straightforward enumeration actually cut each of the differential 
undercounts by half or more--a genuine triumph. Others decided to try 
an experiment. The DSE methodology was thought sound enough that it 
could be incorporated into the census design, which would first take 
one cast of the net (the actual count, which is in reality a sample of 
the population, since we know some were missed), and then return to 
take another sample of 314,000 households, the denizens of which were 
``caught'' again. This process was called the Accuracy and Coverage 
Evaluation (A.C.E.). By comparing the records for an address on the two 
captures, one can find correct ``matches'' (a person found both in the 
enumeration and the re-capture), find overcounts (those found in the 
enumeration, not found in the follow-up), and ``find'' undercounts 
(those appearing in the more intensive search of the A.C.E., but not 
recorded in the enumeration). So far, so good. Matches and mis-matches 
form the basis for a statistical model of how to adjust the whole 
population, both upward and downward, for various groups.
    But what about those ``wily trout'' that evade being caught in 
either net? That remains a genuine dilemma, the technical name for 
which is ``correlation bias.'' There are people who are nearly 
impossible to reach no matter the methodology, and they make up an 
unknown proportion of the undercount. Basically, you cannot know what 
you cannot find. An attempted solution is to ``model'' those people 
based on those you did find who were likewise hard to count, such as 
using anomalies in the sex ratios of those found in certain demographic 
groups. For instance, if we assume women are easier to catch than the 
men who correspond to them in age and race, and we find proportionally 
more women in our counts, we can estimate the number of men who 
``should'' be there as well. We can only hope that the sex ratios 
provide a good model of the unknown; there is no way to demonstrate it.
    Measurement errors in either of the two ``samples'' are a real 
threat (as is the fact that some data are not based on actually 
touching someone's nose, but are derived from information provided by 
proxy--a neighbor or even projected onto a household from the 
characteristics of nearby community members with ``similar'' 
characteristics--this is known as ``imputation''). After all, even the 
perfect plan is being implemented on the ground by an army of recently 
hired part-time census workers who are as prone to mistakes and fatigue 
as any of us.
    Even when all goes well in the field the greatest problem is the 
matching process. Remember that we tagged the fish, presumably on the 
fin. This is not a popular thing for the American government to do to 
the people who happen to be residing here (the census counts citizens 
and anyone else present as well), no matter how efficient it could make 
the census. Moreover, fish rarely all on their own suddenly pull up 
stakes, as it were, and depart for another lake, nearby or across the 
country, without letting us know. Hence, we're never really sure that 
we're catching and matching the same fish when we make our DSE 
comparison, which we will then project onto the whole population. (In 
an earlier incarnation of the A.C.E. design attempted in 1990, a single 
mis-matched family of five led to nearly 45,000 people being 
erroneously added into the adjusted population. This problem and other 
very consequential mistakes in the earlier version--dubbed the PES, for 
Post Enumeration Survey--were only discovered two years after the 1990 
census by a panel of expert reviewers. Fortunately, the PES adjustment 
was not applied.)
    But let's put all those concerns aside for a moment, and presume 
that all went well in the measurement process. Have we really taken a 
sample with the A.C.E. that can then be used to correct the count? The 
answer is still, um, not exactly. There are several remaining steps to 
go. There is first the problem of selecting the 314,000 A.C.E. 
households. Every pollster knows that a proper sample, which will be 
used to extrapolate opinions ``upward,'' as it were, onto the entire 
population (not, as has been noted, what the A.C.E. actually does), 
must guard against being skewed or biased in its selection of 
participants. The best guard is to have a systematically random 
probability process for the selection. But this can not be done exactly 
for the A.C.E. design, since there are constraints on the sample that 
result in trade-offs. For instance, there must be some households 
allocated to every one of the 50 states, and further, we must ensure 
that the households are ``distributed'' in such a way that they 
represent demographic groups of interest.
    When the households are selected, some weighting formulas have to 
be devised to make sure that their members adequately represent the 
groups in question in the overall sample. Skeptics of the process 
wondered just how much the selection of the A.C.E. households and the 
weightings applied could pre-shape the kind of answers that the sample 
was inclined to provide. (Because of correlation bias, we believe that 
the A.C.E. adjustment is prone to report that certain demographic 
groups were ``undercounted'' virtually no matter how good the initial 
count turned out to be. That is, the A.C.E. process may itself be 
biased to ``discover'' an undercount for certain types of people, 
perhaps even in conditions when the initial count itself was, in the 
aggregate, already too high.) At any rate, thinking that the sample 
selection and weighting was based on scientific grounds alone became a 
matter of (by all indications, properly granted) trust.
    But the most important step was to divide up the 314,000 households 
into what were termed ``post-strata.'' That is, the population in the 
sample was stratified and assigned into multiple cells that, according 
to the A.C.E.'s sociology, represented appropriate ``types'' of 
Americans. The appropriateness of a type was related to the probability 
that an individual in one of those cells would be missed in the 
enumerated census. The post-strata (think of the cells in an Excel 
spreadsheet) represented the intersection of variables like race/
ethnicity, sex, age, and tenure (homeownership or not), the whole 
apparatus further divided by four regions of the country and by type of 
community (a range of larger-to-smaller metropolitan areas continuing 
to rural). As it turns out, only the largest post-stratified type, non-
Hispanic Whites, were subdivided into the full set of post-strata 
distinctions. For most other types, the cells had to be conflated 
because of small absolute numbers, meaning that Asian Americans, for 
instance, were placed into two national cells (owner and non-owner) 
without regional breakdowns.
    That is, not all groups were post-stratified by the same criteria. 
Proportionally larger demographic types (whites) could be subdivided 
more finely without seriously affecting data quality, while other 
smaller demographic groups had to be treated as broad bands across the 
whole country (that is, the data were national in the first instance, 
and no effort was made to subdivide them by finer-grained 
distinctions). This decision would have later consequences, such as 
being forced to ``adjust'' the population of one state based on data 
actually derived from nearby, or even relatively distant, states. (This 
fact led some adjustment critics to argue that the A.C.E. design seemed 
more consequential in shifting demographic shares within the population 
rather than prioritizing the need for accurate state-by-state counts of 
all demographic groups.) Other post-strata subdivisions, such as the 
number of age-group breakdowns or the degrees of community density, 
were likewise collapsed for some demographic groups where numbers were 
small, while for those under age 18, the male-female distinction itself 
was dropped. (Representative examples of post-strata would be non-
Hispanic white male homeowner between the ages of 18-29 living in the 
northeast in a large metropolitan area; Hispanic female renter between 
30 and 49 living in a rural area anywhere in the country).
    While at first glance a sample of 314,000 households (close to a 
million people) is a huge number, providing reassurances about likely 
margins of sample error (the larger the sample, the smaller the likely 
probability spread, ideally), the actual population of each post-
stratum cell becomes mathematically problematic. There were initially 
448 post-strata in the A.C.E. (later conflated to 416). Hence, the 
314,000 households divided by 416 post-strata actually yields only a 
little over 700 households per cell. That's not a reassuring number for 
sampling margin of error purposes, especially when we realize that the 
total number of households were not evenly divided among the post-
strata. Because there are so few Americans in some of the assumed 
demographic ``types,'' (example: Hawaiian or Pacific Islander female 
aged 50+ renting a house-trailer in rural Wyoming), the cells 
representing them, even when distributed regionally or nationally, are 
dangerously sparse.
    Further, all of these purely quantitative concerns must be coupled 
with the apparent arbitrariness and uncertainty about the sociological 
assumptions underlying the choice of American ``types.'' Were the 
assumptions actually legitimate models of the probability of being 
enumerated? We simply don't know. Overall, we must realize that the 
census represents the intersection of sophisticated quantification 
(assumptions about numbers) with real human beings (assumptions about 
which are, unfortunately, anything but a sophisticated science). The 
A.C.E. design represents the place where two sets of very complicated 
models of the world derived from two very different disciplines 
interact, with any errors (in theory or in implementation) compounding 
each other. The results are then magnified by becoming the basis for 
adjusting the data on 281 million other people, the A.C.E. being 
considered the last word in accuracy, and hence, the benchmark standard 
for calibrating the entire US data collection system. Suffice it to 
say, the stakes are high for such a probability mechanism of unproven 
reliability.
    And all that has been discussed above transpires before the results 
are released to the public, and ultimately encounters the requirements 
of the legal and constitutional system, fundamental provisions of which 
contradict the A.C.E. activity on the face of it. Finally, in a 
development beyond the scope of this discussion, we must remember the 
indeterminacy added to this census by the first-ever multiple-race 
selection, cross-cutting the whole system with 126 possible choices of 
racial/ethnic self-identification (which choices have themselves been 
acknowledged to be completely arbitrary governmental categories with no 
basis in scientific fact; moreover, the choices are unstable even in 
single individuals at different times).
    What are the particular quantitative dangers of the post-strata? 
Demographers realize that they are caught in their own statistical 
version of Heisenberg's ``uncertainty principle'' when it comes to 
dividing samples into strata. You can pursue one piece of information, 
but only at the expense of its counterpart. The twin problems that must 
be balanced are ``variance'' and ``homogeneity.'' Let us start with the 
second one. If we were devising a statistical model to subdivide 
inanimate objects, such as steel washers coming from an assembly line 
which we wanted to quality-check, our sample need not worry too much 
about homogeneity. We can look for variation in defects, let us say, 
while being reassured that most fundamentals would remain relatively 
constant (the washers wouldn't suddenly form into quartets and start 
singing, for instance).
    With humans (and somewhat less so for fish), that is not so clear. 
Homogeneity is assumed whenever we expect a given cluster of people to 
react the same way to some variable (in this case, getting counted). 
The larger the group of people chosen, the less assured we are that 
they are reliably homogenous. Let us say we were interested in the 
likelihood of being missed in the census and we treated as alike all 
Hispanic females nationwide older than 29 but younger than 50. 
Unfortunately, we would be led to believe thereby that a migrant worker 
who did not finish high school living in a colonia in rural New Mexico 
is as likely to have been missed in the census as a Member of Congress 
living in suburban New York City. That is bad sociology (moreover, the 
thinking is suspiciously akin to what in other contexts is termed 
``racial profiling'').
    Rather obviously, the way to avoid over-homogenizing is to have the 
group to which the assumptions apply be fairly narrow. The smaller the 
cluster of people, the greater the likelihood that they genuinely share 
characteristics of importance. But now we are settling onto the other 
horn of the dilemma. Groups small enough to be reliably similar are 
also small enough to produce large variance (the statistical ``spread'' 
of the data) when their results are applied beyond the group. Hence, 
the design problem for the A.C.E.: develop sufficient post-strata that 
every cell is composed of reasonably homogeneous members, but do not 
make so many post-strata that small cells produce inherently unreliable 
sample data. Once again, a trade-off is faced, trying to optimize a 
response to the twin challenges.
    As it turns out, the National Academy of Sciences, which was 
routinely characterized by the press as having ``endorsed'' census 
adjustment, in reality only agreed to the  principle  of statistical 
adjustment as quantitatively sound. Some members were never 
enthusiastic about some particulars of the actual A.C.E. plan (much 
less its field implementation), especially given that it was hurriedly 
developed in response to a 1999 Supreme Court decision ruling against a 
much more ambitious version of statistical adjustment, an effort to 
create a ``one number census'' based on an Integrated Coverage 
Measurement.
    Let us grant for the moment that the A.C.E. design was adequate for 
our purposes (and you must not forget that our purposes include 
political apportionment and redistricting, as well as the proper 
distribution of federal funds, over and above the need to tabulate the 
aggregate numbers in the census). For all 416 (collapsed) post-strata, 
the matching process between the enumeration and the A.C.E. begins to 
tell us about which types were overcounted, which were undercounted, 
and which are ``just right'' (once again, this must be a somewhat 
simplified description; there are other complicated process to cause 
concern such as the unduplication of records or imputations).
    Now comes the adjustment activity. Based on the ``signal'' derived 
from the A.C.E., we develop another weighting, regarded as a 
``correction factor,'' which we export back into the total population 
count after it has been likewise stratified to match the A.C.E. types. 
(Again, critics argue over terminology. To term the factor a 
``correction'' appears to prejudice the case that the result is somehow 
more accurate than the original number to which it is applied. 
Accordingly, it may be more valid to simply term the factor an 
``adjustment,'' acknowledging that the A.C.E. doesn't necessarily 
produce a ``better'' number, just a different one.) Adjustment factors 
can be positive (we are adjusting an undercount by using a number 
higher than one) or negative (we are adjusting an overcount by using a 
number less than one). We then multiply the count for each group in the 
enumeration by their respective adjustment factor, the product being 
what we record as their actual (adjusted) count.
    This latter process led to some surprise when it was realized that 
the effect was to ``delete'' from the census actual people who had 
bothered to do their civic duty and fill out a form. Being 
understandably sensitive about appearances, the Bureau denies that 
anyone is deleted. They prefer to note that what happens is merely that 
a negative record is imputed to the census count, in effect nullifying 
the count of a real person that chose to participate. Whatever 
terminology we accept in this issue, a study of the 1990 PES identified 
no fewer than 1.48 million such ``nullifications'' based on overcount 
assumptions. Preliminary data from the Census 2000 A.C.E. indicate 
approximately one million such nullifications would take place.
    At any rate, we are now at a point in our analysis where we can 
adjust the census statistically to ``correct'' the count. If a post-
stratum has a positive number, such as 1.08, that means that we found 
more people in the A.C.E. survey in that stratum than the enumeration 
had recorded. Rather than 100 people, let us say, the A.C.E. is telling 
us there likely are 108 people. Hence, every time we find a record back 
in the enumeration national census for some other one in that post-
stratum, we do not record just a one; we instead write down 1.08 for 
each one found. That means for every group of 100 people we find 
anywhere in the country who fit this profile, we ``add'' 8 more people 
of that type (where, exactly, do we put them? More in a minute . . .). 
Now we are rolling at last. For every 10,000 found, we write down 
10,800. For every ten million, well, let us see here, the model tells 
us we have got 800,000 more people just like that, which we have to 
place somewhere on the map, even though we've never actually met nor 
counted them directly.
    Moreover, some actual post-strata receive corrections that are 
hefty indeed. Hispanic males aged 30-49 rural non-owners in low 
enumeration districts, for instance, receive a correction factor of 
1.19; that's nearly twenty percent, 120 for every 100, 60,000 for every 
50,000. These are, in an important sense, virtual people, who must 
nevertheless be awarded their ``fair share'' of very real political 
power and funding (which are, by the way, zero-sum entities; if I give 
this finite resource to someone, it can only happen at the expense of 
someone else in direct proportion). It follows, of course, that for 
those receiving a negative correction factor, we write down for every 
one we encounter a number less than one. So for the presumably 
overcounted types, whenever we find them, we write down .92, for 
instance, and then add them together. For every 10 million of these 
losers, of course, we only record 9,200,000.
    Even more remarkable, we have just engaged in a process that is not 
really ``sampling'' at all, but rather another (and less supportable) 
statistical maneuver known as ``synthetic estimation.'' Recall that the 
adjustment consists of comparing one sample (the enumeration) with 
another sample (the A.C.E. population) and seeking matches. Based on 
the assumption that the A.C.E. results are always to be considered 
superior to the actual count (which may not be true, especially if the 
enumeration, which made a greater effort to activate local community 
outreach, was more successful at coverage of the recalcitrant than was 
the more ``professional'' A.C.E. re-contact, undertaken without the 
intensified community efforts), a set of ``adjustment factors'' are 
computed for each post-stratum. So far, so good, as far as statistical 
probability goes.
    But then the adjusted numbers are applied to the entire national 
population with each post-stratum receiving its proportional adjustment 
higher or lower. The overall effect is a movement that goes in two 
directions. First the sample adjustments are adduced upward, as it 
were, to the national totals, and then brought back down, as it were, 
to the local level when the count is adjusted block-by-block.
    It is this second movement back down from the aggregated total and 
distributed onto the smallest components of the population groupings 
that causes statistical concern. When we bring the totals from the 
national level back into the local aggregations we are engaged in what 
is no longer ``sampling extrapolation'' by any means, but rather a 
different maneuver--the ``synthetic estimation.'' The fundamental (and 
contested) assumption behind the ``synthetic'' part is that because a 
certain proportion or ratio of a population can be asserted about a 
whole group (the US population), therefore each distinct component of 
the aggregated whole likewise must mirror those proportions or ratios 
in equal manner.
    But what is true of a statistical whole is not necessarily true of 
each individual component (statistics, after all, representing a summed 
average of many measures). Imagine for a moment that I discover a ratio 
of females to males at a university of thirty thousand students--
females are 55 percent, males 45 percent. At the aggregate level, that 
is, the whole university, this can be accurate, without necessarily 
implying that each classroom in the university replicates this exact 
proportion. French classes, for instance, may not show the same ratios 
of female to male as chemistry courses, even though when taken together 
they ``average'' the overall ratio. To likewise expect every table in 
the cafeteria to exactly mirror the overall ratio quickly leads to 
absurdity--we should expect, under the principles of synthetic 
estimation, to find exactly 5.5 females and 4.5 males at every table of 
ten. Clearly something is wrong.
    And yet this is just what the census adjustment process leads us to 
formulate. The adjustment factors for each post-stratum population 
found in the A.C.E. sample are ``nationalized,'' as it were, and then 
applied down to the local level of neighborhoods, expecting the same 
ratios of under- and over-counted to apply at every level of the 
population hierarchy--state, county, congressional district, census 
tract, local block. As it turns out, the A.C.E. plan did, in fact, run 
into difficulty with this ``synthetic'' assumption, which further 
reflects the problems noted above in the discussion of assumed 
``homogeneity'' of the post-stratum.
    In actuality, we begin to see many difficulties with the operation 
of the A.C.E. conceptually, over and above those concerns linked to 
problems of measurement error and implementation issues. One of the 
central conceptual difficulties is that the statistical estimation, 
incorporating, as it must, a certain probability margin of error that 
is ineradicable (it being inherent in the operation of probability), 
only begins to ``even out'' its errors at certain levels of 
aggregation. That is, for the gross level of the total population (the 
aggregate count of the total number of the population, roughly 281.4 
million persons), the probability errors (the inevitable pluses and 
minuses wavering from the actual target) do ``average out.'' For 
instance, for every 100,000 measurement ``pluses'' that are too high 
there will also occur about 100,000 corresponding measurement 
``minuses'' that cancel each other out. But a lower levels of 
aggregation (state, county, district) the possibility starts to magnify 
that they do not all ``average out,'' and we may well be left with 
residual error--a less accurate count than we began with in the 
enumeration.
    There is considerable dispute as to what level, exactly, we begin 
to lose ground with that adjustment, and actually start introducing 
error by adjusting. It may happen at the state or congressional 
district level, especially for selected demographic groups in the post-
stratum (and perhaps worsened by the realization that we are using 
gross regional or even national data to adjust populations within a 
state--that is, we are not directly adjusting a state's population 
based on data derived only from that state). Whatever the eventual 
resolution of that dispute about the accuracy/inaccuracy threshold, 
everyone now agrees that at smaller levels of aggregation (in counties 
with less than 100,000 people it becomes clearly problematic), on down 
to the block level, we can no longer assure ourselves that the 
adjustment is superior to the unadjusted numbers, and we begin to 
seriously suspect that the adjustment is actually distorting our 
understanding by introducing error into the count.
    Yet it is at the block level that politically important decisions 
must be made--such as the boundary of an electoral district. Moreover, 
the hoped-for randomness of the pluses and minuses canceling each other 
out is further belied by the practice of the re-districters, who tend 
to accumulate together blocks of people who share certain demographic 
characteristics, if for no other reason than their physical 
propinquity. Hence, if a particular demographic post-stratum is off 
within the margin of error in one consistent direction, there will be 
no balancing out of the error because they will be grouped together 
with similar blocks likewise erroneous in the same direction. The 
effect is to amplify the error in the redistricting result, rather than 
having randomness producing a canceling-out effect.
    Problems abound. By virtue of the adjustment design, we have 
generated estimated people (virtual people) who have never been 
contacted nor identified, yet must be placed in some concrete location 
in an actual census block. The principle for assigning them a ``local 
habitation and a name'' is arbitrary and based on unproven assumptions. 
Yet their presence can have consequence in the apportioning of 
political power and funding. Moreover, the Congressional-appointed 
members of the Board further demonstrated in our report of September, 
1999 that the effect of adjustment is to fail to position the 
undercounted correctly and proportionally in the communities where they 
were actually missed. By applying a ``blanket'' adjustment to every 
sector across the country, the adjustment gives the illusion of a 
remedy, because the actual undercounted are not uniformly distributed 
across the country. Those communities that ``lose'' in the undercount 
do not receive a commensurate adjustment.
    Further, it could be argued that the adjustment design, an effort 
to statistically ``model'' the population and then reformulate it, 
could have the effect of introducing more political features into the 
census than are found in the actual enumeration. Let it be noted that 
the ``political'' aspect of the census adjustment does not have to 
necessarily imply the active intervention of partisan concerns. As with 
budgetary or income tax battles, any process that is ``assumption-
dependent'' is thereby open for political debate. Whoever sets the 
assumptions, or establishes the criteria for which factors are 
considered important (and in which order), can largely constrain the 
possible outcomes of the strictly quantitative process. All census 
activities are, of course, assumption-dependent, in this sense--witness 
the dispute between Utah and North Carolina over the allocation of the 
last House seat based on population. The issue hinged on whether or not 
overseas missionaries were assumed to be equivalent to overseas 
military in terms of their state assignment. It follows that every 
aspect of the census has political implications, in that it constructs 
political definitions and quantifies what are properly political 
entities--human beings in groups. Nevertheless, even given these 
caveats, census enumeration is relatively more assumption-independent 
than is the alternative--modeling the population for statistical 
adjustment purposes, where changed assumptions have the power to 
radically alter the entire nature of our national self-portrait.
    A corollary of this reasoning is that the enumeration count will 
likely prove more accountable to democratic processes in the long run, 
as well. Witness the difficulties already encountered by policy makers 
and courts trying to understand and evaluate the highly technical 
nature of the adjustment's probability models. Who can truly grasp them 
and interrogate them but a very restricted group of technical experts? 
In this sense, an enumeration process, being relatively more 
transparent in its assumptions and enactment, may be not only more 
accessible and hence accountable but also more prudently consistent 
with the spirit of self-government.
    Finally, where do we now stand? In the first place, with an 
estimated undercount of only 1.18 percent (that is, a census that is 
98.82 percent accurate), we should realize that the cost/benefit ratio 
of our respective choices begins to shift. Given all of the attendant 
legal and political difficulties that the A.C.E. engenders, its saving 
grace was that it might be a good technical fix. But now that is in 
question. The A.C.E. appears to have levels of ``statistical noise'' in 
its probability fluctuations that are greater in magnitude than the 
``signal'' it was designed to detect and correct (the 1.18 undercount 
is smaller than the margins of error range of the A.C.E. at certain 
levels of application). Applying an adjustment to a census as accurate 
(by all the evidence to date) as the one just completed begins to slide 
down the slope of diminishing returns, technically as well as in terms 
of governance consequences.
    We face a genuine dilemma, which the ESCAP report issued at the end 
of February well captured. The actual enumerated count from Census 2000 
placed us in a new and perplexing landscape, one unanticipated by the 
designers of the A.C.E. (and many other parties as well). The count of 
281.4 million surpassed the best reckoning of the population provided 
by the Demographic Analysis (DA) by nearly 2 million people. This is an 
anomalous outcome, since the DA has traditionally stood above the 
enumerated count and told us the magnitude of the undercount in the 
enumeration. But now we have broken through the DA measurement. By the 
reckoning of the DA, we actually have an OVERCOUNT in the enumerated 
census. Perhaps, on the other hand, the enumerated census is right on 
the mark, having itself nearly eliminated the heretofore undercount. 
Various scenarios to explain this have already been proposed by the 
ESCAP report and others, but none has achieved a reconciliation without 
introducing yet other anomalies.
    Moreover, the emerging and tentative adjustment count from the 
A.C.E. took us even further away from solid ground, with an estimated 
result of approximately 284.7 million. The gap with the enumeration 
count is substantial, while the gap between the A.C.E. and the DA 
(about 5 million) is fundamentally problematic--too large, in fact, to 
be reconciled by any scenario yet deployed. Moreover, the A.C.E. 
results themselves contain internal anomalies and inconsistencies, in 
addition to the incapacity to be reconciled with either of the two 
other measurements.
    It should be apparent to anyone seriously engaged with this problem 
that while many specific details remain to be resolved, and further 
while the A.C.E. design has a valuable contribution to make in helping 
us understand what transpired in the census count, the inadequacies in 
concept and in practice preclude use of the numbers derived from the 
A.C.E. for the critical purposes of apportionment and redistricting. 
The A.C.E. methodology simply cannot met its primary obligation--being 
demonstrably more accurate than the data which they might supplant.
    We must acknowledge the wisdom in the ESCAP recommendation to the 
acting Bureau Director (a recommendation accepted by Secretary Evans, 
and which moreover was endorsed by the previous Bureau Director Dr. 
Kenneth Prewitt) to regard the unadjusted numbers from the Census 2000 
enumeration as the accurate numbers, that are the most appropriate for 
the Constitutional uses to which they are put. It was a decision made 
on the merits of the case as it was examined.
    Let us not shun the larger lesson from this overall undertaking. 
The undercount is a genuine American difficulty, to which we need 
genuine solutions. No one should in principle be uncounted, and we must 
develop more effective remedies to ensure that the principle of the 
census is fulfilled. By all that we now know about the enumeration 
process, we should recognize a striking achievement, which was to 
reduce the differential undercount. The promise of that outcome is that 
we can close it yet more by intensifying the enumeration, by forming 
local partnerships to accomplish it, and by motivating people to find 
their way into full participation in the American system. ``All 
politics is local,'' was wisely said. All censuses may likewise be 
local. Let us properly invest in what works best.
    Thank You.

    Senator Edwards. Thank you, Dr. Murray.
    Mr. Vargas.

        STATEMENT OF ARTURO VARGAS, EXECUTIVE DIRECTOR,

           NATIONAL DIRECTOR, NATIONAL ASSOCIATION OF LATINO

                ELECTED AND APPOINTED OFFICIALS (NALEO)

                            EDUCATIONAL FUND

    Mr. Vargas. Thank you, Senator. My name is Arturo Vargas. I 
am the executive director of the National Association of Latino 
Elected and Appointed Officials Educational Fund, and thank you 
for the invitation to appear before you at this Senate hearing 
on the decision to release the adjusted census data for Census 
2000.
    The NALEO Educational Fund is the nation's leading 
organization that enables Latinos to participate fully in the 
political process. We are a 501(c)(3) nonprofit bipartisan 
organization with a bipartisan board of directors that took a 
strong position in support of the most accurate census 
possible, and in support of the Bureau's decision to use 
statistical sampling.
    We were one of the organizations that worked with the 
Census Bureau through a partnership to promote a full count, 
and we were particularly proud of the mail-back response rates 
that we saw in Latino communities throughout the country. We 
commend the Census Bureau for the many elements of the Census 
2000 which made it such an operational success, and we will be 
happy to share with this Committee at an appropriate time our 
views on the elements of the census that were particularly 
successful as well as the areas where we believe that we could 
have improvement for the 2010 census.
    Now, as we all know, the preliminary estimates of the 
under-
count by the bureau indicate that there was, in fact, a 
differential undercount, and that differential undercount does 
not fall equally among all Americans. Latinos, African 
Americans, Asians, children, immigrants bear the brunt of the 
undercount.
    When the career statisticians at the bureau announced the 
recommendation against the release of the adjusted data, they 
based their decision on three different methodologies that have 
been discussed by the Secretary and the Acting Director. These 
professionals unequivocally concluded that there is 
considerable evidence to support the use of adjusted data. 
However, as has been indicated, they were troubled by 
discrepancies between adjusted data and the results obtained by 
the Demographic Analysis.
    Senator, Members of the Committee, I think the issue before 
us is, we must permit the bureau to take the time it needs and 
the resources it needs to resolve this issue. The ESCAP 
committee did not conclude that the A.C.E. data were less 
accurate than the Census. The ESCAP committee merely concluded 
that at the time the recommendation was required, they could 
not resolve the discrepancies. They did not say that the census 
data are, in fact, more accurate than the A.C.E. They did not 
say that the A.C.E. was more accurate than the census data. 
They indicated that at the time of the decision of their 
deadline, that they could not make that conclusion yet.
    We should not be forced to live with a 10-year error 
because of a 3-month deadline. As we heard Secretary Evans 
express here today, the Census Bureau intends to continue its 
evaluations. I think it is the major responsibility of this 
Committee to support the professionals at the Census Bureau and 
ensure that they have all the resources and all the time 
necessary to complete the analysis of the Demographic Analysis, 
Census 2000, and the A.C.E. so that in the fall they can make 
this recommendation as to whether or not, in fact, Census 2000 
can be corrected to compensate for the 3.3 million Americans 
who have been excluded from Census 2000.
    Now, as an organization that works hard to ensure that 
Latino Americans are able to fully participate in our political 
process and in our society, the differential undercount has a 
strong, harmful impact on that goal of Hispanic Americans. It 
has an impact on our political representation, on the ability 
of institutions such as schools to provide appropriate services 
and classrooms, resources to children.
    We cannot live with a 10-year error, Mr. Chairman, and I 
would ask the Committee at this time to hold accountable 
Secretary Evans for the commitment he made here today to this 
Committee that in the fall Census Bureau professionals will be 
recommending to him a full recommendation as to whether or not 
to adjust the census data.
    Again, I think the issue here is not that the A.C.E. was 
less accurate, or that the census itself was more accurate. The 
issue here is that the ESCAP committee at the time they were 
required to make their recommendations, did not have the time 
sufficient to conclude one way or the other.
    I think it is most important for the interests of this 
country that we allow the Census Bureau professionals to do its 
job with the full support of Congress, the full support of the 
administration, to ensure that when they make this new 
recommendation as to whether or not the A.C.E. is more accurate 
or not, that this country can be allowed to go forward for the 
next decade with the most accurate decade available to it.
    Thank you, and I have summarized my comments here, and we 
have submitted our comments for the record.
    [The prepared statement of Mr. Vargas follows:]
   Prepared Statement of Arturo Vargas, Executive Director, National 
    Director, National Association of Latino Elected and Appointed 
                   Officials (NALEO) Educational Fund
    Chairman McCain, Ranking Member Senator Hollings and Members of the 
Committee: I am Arturo Vargas, Executive Director of the National 
Association of Latino Elected and Appointed Officials (NALEO) 
Educational Fund. Thank you for the invitation to appear before you 
today on behalf of the NALEO Educational Fund to discuss the full 
impact on the Latino community of the recent decision by Commerce 
Secretary Don Evans to release Census 2000 data for redistricting that 
has not been adjusted to correct for the differential undercount.
    The NALEO Educational Fund is the leading national organization 
that empowers Latinos to participate fully in the American political 
process, from citizenship to public service. The NALEO Educational Fund 
carries out this mission by developing and implementing programs that 
promote the integration of Latino immigrants into American society, 
developing future leaders among Latino youth, providing assistance and 
training to the nation's Latino elected and appointed officials; and by 
conducting research on issues important to the Latino population. The 
NALEO Educational Fund is a 501(c)(3) non-profit, non-partisan 
organization. Our constituency includes the more than 5,400 Latino 
elected and appointed officials nationwide.
    As a member of the Commerce Secretary's Decennial Census Advisory 
Committee, I am pleased to be able to discuss with you the decision to 
release unadjusted Census 2000 data as the official data for the 
purposes of redistricting.
    The NALEO Educational Fund is committed to ensuring that our nation 
will be able to rely on the most accurate data possible from the 2000 
Census. Our organization, like hundreds of others across the country, 
mobilized to encourage all U.S. residents to answer the census. We are 
particularly proud of the mail back response rates in several Latino 
majority communities which demonstrated the sincere desire among 
millions of Latinos to make themselves count in 2000. We commend the 
Census Bureau for the many elements of Census 2000 which made it such 
an operational success, including its partnership program and 
commitment to work closely with community institutions, its high 
quality outreach and advertising program, and its efforts to hire an 
enumeration force that had the skills and capacity to carry out this 
monumental task. We would be happy to share with this Committee at 
another appropriate time our views on the elements of the census which 
were particularly successful and those areas in which we would 
recommend improvements for 2010. Our focus today, however, concerns the 
most basic element of the census, the accuracy of the data on which we 
will rely upon for an entire decade.
    As we all now know, the preliminary estimates released from the 
Bureau indicate that the differential undercount was not eliminated. 
While the Census 2000 was an operational success, there was a net 
undercount of 3.3 million Americans. And many of those missed were 
Latinos--over one million. That, Mr. Chairman, is more than the entire 
state of Wyoming.
    When the career statisticians at the Bureau initially announced 
their recommendation against release of the adjusted data, they based 
that decision on their examination of three different methodologies 
used to determine our nation's population: the traditional 
``headcount,'' the statistically-adjusted data based on the Accuracy 
and Coverage Evaluation (A.C.E.), and the Bureau's separate demographic 
analysis. These professionals unequivocally concluded that ``there is 
considerable evidence to support the use of adjusted data;'' however, 
they were troubled by discrepancies between the adjusted data and 
results obtained by the demographic analysis. They had to meet a 
deadline to make a recommendation regarding the release of the adjusted 
data, and they simply ran out of time to examine and explain those 
inconsistencies.
    It is critical that we permit the Bureau to take the time it needs 
to resolve this issue. We should not be forced to live with a 10-year 
error because of a three-month deadline. If the Bureau determines that 
the adjusted numbers are more accurate, the Bureau should release them 
for redistricting and other purposes. The connection between 
redistricting and the Census goes back to the founding of our nation. 
The redistricting process plays a key role in ensuring that our 
democratic process provides fair representation for our nation's 
residents. The use of unadjusted data for this process will result in 
inherently mal-apportioned districts. Because the undercount occurred 
predominately among minority populations, Congressional and state 
legislative districts with substantial numbers of minority residents 
will in fact contain a much larger population than what the unadjusted 
data indicate. Thus, those districts would in reality be comprised of a 
larger number of residents than districts which are predominately non-
minority. The differences between the size of the actual population in 
such districts could exceed the deviation permitted under the ``one 
person, one vote'' principles of current law.
    We are also concerned about the negative impact unadjusted data 
could have on voter participation in communities with language 
barriers. Section 203 of the Voting Rights Act requires jurisdictions 
that meet certain criteria to provide language voting assistance to 
their residents. Jurisdictions qualify if (a) they include at least 
10,000 voting-age citizens who belong to a single language community 
with limited English-language abilities, or (b) such citizens comprise 
more than 5% of their voting-age citizen population. This is determined 
by census data.
    For the Latino community and the nation as a whole, the 
repercussions of not releasing data adjusted to correct the undercount 
will extend far beyond our political system. In general, accurate, 
corrected data are vital for all types of programs and services. As you 
know, Mr. Chairman, there has been much discussion about the dramatic 
growth of the Latino community, and its implications for this country's 
economic, social and political institutions. This is an important 
discussion, because as a result of this growth, our community and 
nation will face many challenges. Moreover, community providers, urban 
and rural planners and policy makers must be equipped with the most 
accurate baseline data available to make the comparisons and 
assessments that are critical for their work.
    A census undercount also drastically undermines access to quality 
education, a particularly important issue for Latino families. The 
Census Bureau's most recent Current Population Survey data reveal that 
36% of the Latino population is under the age of 18. Decisions about 
the allocation of resources in school districts are based on census 
data. We know very well who was actually missed in the 2000 census. In 
low-income communities, it was immigrants and children. What this means 
to many Latino communities across this nation is that when school 
administrators are determining where to build new facilities, the 
number of teachers they need, or the number of school books to buy, 
they may mistakenly plan for 10,000 children, instead of the 12,000 who 
actually reside and attend school in the district. Given the 
extraordinary crisis in our public schools today, and their inability 
to adequately educate the nation's Latino children, this is an 
extremely critical juncture for our nation's future success.
    So there is much at stake for the Latino community, not just 
politically, but also economically. If the Latino population is not 
fully counted, the communities in which they reside will likely lose 
funding for schools, hospitals and other vital social programs. These 
communities will, in effect, be disenfranchised for the next ten years.
    What is even more pressing now, Mr. Chairman, is the recent 
revelation that the Census Bureau has, using scientifically approved 
methods to correct the undercount, produced a corrected set of numbers 
down to the block level for the 2000 census in all 50 states. Today 
perhaps, Mr. Chairman, in this committee which has a tradition of 
openness and full disclosure, we can receive a commitment from the 
Census Bureau and the Commerce Secretary to release the corrected data 
if the Bureau determines they are indeed more accurate than the 
traditional ``headcount.'' Our government may have spent as much as 
$400 million to pay for the A.C.E.
    If the A.C.E. has produced the best numbers available, Congress and 
the American taxpayers should be entitled to this important 
information.
    Mr. Chairman, we urge Secretary Evans to direct the Bureau to 
complete an analysis of the accuracy of the adjusted data as quickly as 
possible. If that analysis reveals that the adjusted numbers are more 
accurate than the unadjusted count, the Bureau should immediately 
release the data for redistricting and other purposes. If the analysis 
is completed after the data can be feasiblely used for redistricting, 
the adjusted numbers should still be released, for public policy 
planning purposes and to assist us in improving the way we conduct 
future census efforts. If the analysis reveals the adjusted numbers are 
less accurate than the unadjusted count, the adjusted data should still 
be released to enhance our understanding of census enumeration 
methodology.
    I thank the Chairman, the Ranking Member, and the Committee once 
again for providing the NALEO Educational Fund with the opportunity to 
share our views today on the release of the Census 2000 data.

    Senator Edwards. Thank you very much, Mr. Vargas. As the 
chairman indicated earlier, when the House Members got back 
from their vote, we were going to interrupt this panel so that 
they could present their testimony. Dr. Ericksen and Dr. 
Wachter, we need to interrupt you and proceed with the House 
Members.
    I do not know who wants to go first. Mr. Clay, would you 
like to begin?

                STATEMENT OF HON. WILLIAM CLAY, 
               U.S. REPRESENTATIVE FROM MISSOURI

    Mr. Clay. Mr. Chairman, thank you for inviting us to speak 
before your committee today. As you know, the census is an 
issue of great importance to every Member of the House of 
Representatives every 10 years, this country has peacefully 
redistributed congressional power between the states based upon 
the census and remarkably only once in our nation's history has 
this process failed us. In 1920, Congress would not accept that 
the population change had shifted the majority from rural areas 
to urban areas. This disbelief led to a decade of inactivity.
    As a result, in 1929 Congress put an end to the decennial 
battles in Congress over how the census would be taken and how 
seats would be apportioned by approving legislation that 
established permanent authorization for the census, and a 
standing formula for the distribution of seats among the 
states. In other words, Congress put an end to the political 
wrangling over the census.
    Unfortunately, that legislation enabled another unsavory 
tradition, congressional districts of unequal size. By the time 
the Supreme Court ruled that these districts should be equal, 
the largest congressional district was 10 times the size of the 
smallest. In this decision, the court said such variances were 
blatantly unfair. The Supreme Court, in righting the wrongs of 
redistricting, put this census clearly in the middle of the 
political battles once again.
    Voting rights and the census are subjects near and dear to 
the hearts of African Americans. Historically, we have been 
short-changed by both. The first census counted African 
Americans as three-fifths of a person. That ugly tradition of 
part person, part property, continued until the passage of the 
Fourteenth Amendment in 1868.
    In 1940, African Americans first made us aware of the 
errors in the census when more African American young men 
registered for the draft than the Census Bureau thought existed 
in the entire nation.
    The realization that African Americans are routinely 
overlooked at a much higher rate than whites has been the basis 
of fixing the errors in the census. Even the Census Bureau's 
own numbers show that over 2 million African Americans were 
missed in the 2000 census.
    Today, we find ourselves at the end of a 5-year political 
dog fight over the census in that same unsavory tradition of 
the 1920's. It is time for this Congress to act as responsibly 
as the 71st Congress did in 1929.
    There is a very easy way to put an end to the arguments 
over the quality of the 2000 census. Release all the numbers 
for public scrutiny. If the 2000 census is as accurate as the 
Secretary says it is, the numbers will show it. If there are 
flaws in the census, the numbers will show that, too, and if 
there are flaws in the dual system estimates created by the 
Census Bureau, public scrutiny will reveal those flaws as well. 
One thing is for sure, hiding the numbers will only increase 
the public suspicion that the process has been rigged.
    Congress has a long and honorable tradition of standing on 
the side of public disclosure of government information. It was 
Congress that broadcast the first Census results to the public, 
and it was Congress that established the Government Printing 
Office and the depository library program to keep the public 
informed. It was also Congress that passed the Freedom of 
Information Act and the Presidential Records Act, and it was 
Congress that said groups advising the executive branch should 
do so in the light of day, and pass the Federal Advisory 
Committee Act.
    Congress has stood again and again for the right of the 
public to know what its government is doing, and the 107th 
Congress is obligated to carry on that great tradition. That is 
why I urge you to join with me in calling for the release of 
all of the data from the 2000 census to do any less would deny 
the American public its right to an open and honest government.
    As a note of interest, representing the state of Missouri, 
we see that according to Gene Ericksen, Missouri was 
undercounted by approximately 31,000 people. I think that it is 
only right that the Census Bureau release the adjusted numbers 
for the states, because it is going to have a negative impact 
on the state of Missouri as well as all of the other states 
where an undercount exists, and I represent an urban-suburban 
district, and I am certain that those 31,000, the majority of 
them come from the area that I represent.
    So I thank you, Mr. Chairman, for this opportunity.
    Senator Edwards. Thank you, Congressman. Thank you for 
taking the time to be with us and sharing your testimony with 
us.
    Mr. Gonzalez.

            STATEMENT OF HON. CHARLES A. GONZALEZ, 
                 U.S. REPRESENTATIVE FROM TEXAS

    Mr. Gonzalez. Mr. Chairman, thank you very much, and to the 
other Members of the Committee for this opportunity to appear 
before you today. I am here today as chair of the Congressional 
Hispanic Caucus Census and Civil Rights Task Force to share our 
thoughts on recent actions and decisions affecting the 2000 
census and census data to be utilized over the next 10 years.
    As we are aware, the 2000 census has shown the nation's 
Hispanic population to be the largest and fastest-growing 
minority population in the entire country. Unfortunately, it is 
also among the highest undercounted populations. As such, I 
would argue the nation's Hispanic community has the most to 
lose if we continue down the road in utilizing census data that 
by all accounts can and should be greatly improved.
    We can all agree that the 2000 census was vastly improved 
from the past decennial censuses and that it was an operational 
success. However, there remains much room for improvement, and 
we should support a concerted effort aimed at identifying 
shortcomings in those areas need of improvement.
    On March 1, 2001, Acting Census Bureau Director William G. 
Barron, Jr., who was here today--I am not sure if he is still 
here--recommended to Secretary of Commerce Don Evans that 
unadjusted data be released to the Census Bureau's official 
redistricting figures to the states. Mr. Barron's 
recommendation was based on the report of the Executive 
Steering Committee for Accuracy and Coverage Evaluation Policy, 
referred to as ESCAP.
    While that report did, in fact, recommend that unadjusted 
numbers be released as the Census Bureau's official 
redistricting information, it also stated that the Committee 
believed that (1) unadjusted census totals would reflect a 
national undercount, which has been brought up by each speaker 
here today, (2) that there was considerable evidence to support 
the use of adjusted data, and (3) that further research may 
establish that adjusted data would result in improved accuracy.
    It is critically important to note that the Committee cited 
time constraints and lack of sufficient information at that 
point in time as key to its recommendation. In addition, in its 
report, the Committee was clear that although its 
recommendation was to release unadjusted data, that decision 
did not mean or even suggest that unadjusted data was superior 
in any way.
    In a meeting on March 2, 2001, with Secretary Evans, just 
days before the Secretary announced its decision to release the 
less accurate, unadjusted data as the official redistricting 
figures, the Congressional Hispanic Caucus requested that the 
Secretary require the Census Bureau to complete its work on 
Census 2000, and to subsequently make a recommendation 
regarding the use of adjusted data for other purposes, 
including the Intracensal Census.
    Specifically, the Hispanic Caucus requested that the Census 
Bureau be afforded any additional time and resources necessary 
to conduct further research into inconsistencies between 
estimates derived from outmoded Demographic Analysis tools and 
the modern scientifically enhanced and supported accuracy 
coverage evaluation estimates.
    We believe that such an evaluation is critical to 
establishing the modern day reliability of Demographic Analysis 
as a tool and subsequently to determine whether or not adjusted 
data would result in improved census accuracy.
    We further requested, and it has been requested here today 
by the witnesses, in keeping with the Census Bureau's 
longstanding policy of openness, that the adjusted or corrected 
block level data be made available.
    I am going to make reference now to the ESCAP report on 
page 1 of 28, where it states, this report is also being 
released to the public at the same time that it is being 
forwarded to the Secretary of Commerce. Footnote 2, in addition 
to the requirement to make the report public, the Census Bureau 
firmly believes that full disclosure in a vigorous and informed 
debate will improve both the Census Bureau's internal processes 
and the public's understanding of statistical adjustment.
    So in keeping with the very spirit of the bureau and its 
policy, and with that I would imagine the Secretary of Commerce 
would also follow that spirit, we are asking today that that 
unadjusted data be made available to anyone that would request 
it, and not wait for a Freedom of Information Act or any other 
type of request. Just put the information out there.
    The Secretary in his testimony earlier this morning, sir, 
indicated that he did not believe that the information would be 
fit for use. I beg to differ. It is all in the eye of the 
beholder. There are statisticians and scientists outside of the 
bureau that may find that information very useful for whatever 
purpose they choose. If there is disagreement as to its 
utilization, then we will have an open debate in the 
marketplace of ideas, which is the best type of dialog.
    So we are here today, and I am representing the caucus 
again asking this Committee to please join us in the request of 
Secretary Evans regarding that full resources immediately be 
applied to reaching some conclusion on the discrepancy between 
the democratic analysis and the A.C.E. so that we can have the 
adjusted figures for all other purposes.
    In the state of Texas, Senator, this impacts the Hispanic 
community like no other segment of the population. In Texas, it 
is 350,000. In my district, it could be 35,000, it could be 
50,000, but when it comes to education, when it comes to roads, 
when it comes to libraries, when it comes to health care, all 
of it, we put a human face on it, and all we are asking is for 
accuracy, with the tried and proven scientific method.
    This fight is still not over. We have characterized it as a 
civil rights debate and issue of the decade, and we still 
believe it is, and again, thank you for your courtesies.
    [The prepared statement of Mr. Gonzalez follows:]
            Prepared Statement of Hon. Charles A. Gonzalez, 
                     U.S. Representative from Texas
    My thanks to Chairman McCain, Ranking Member Senator Hollings and 
the entire Senate Committee on Commerce, Science and Transportation for 
inviting me to testify today on an issue in which I, along with every 
person residing in the United States of America, have a vested 
interest--the Census.
    I am here today as the Chair of the Congressional Hispanic Caucus 
Census and Civil Rights Task Force to share our thoughts on recent 
actions and decisions affecting the 2000 Census and census data to be 
utilized over the next decade. As we are aware, the 2000 Census has 
shown the nation's Hispanic population to be the largest and fastest 
growing minority population in the country. Unfortunately, it is also 
among the highest undercounted population. As such, I would argue that 
the nation's Hispanic community has the most to lose if we continue 
down this road of utilizing census data that by all accounts can and 
should be greatly improved.
    We can all agree that the 2000 Census was vastly improved from past 
decennial censuses, and that it was an operational success with higher 
than expected numbers of people participating by returning their census 
forms. I particularly credit Census Bureau employees nationwide, the 
thousands of communities and organizations across the country that 
partnered with the Census Bureau to promote participation, and the 
Census Bureau for devising and implementing its comprehensive plan, all 
of which contributed to a better census.
    However, there remains much room for improvement and we should 
support a concerted effort aimed at identifying shortcomings and those 
areas for improvement. It is only through this process that we will be 
able to ensure improved accuracy and greater success in future 
censuses. Through this process we will also be afforded the opportunity 
to establish the validity of the Accuracy and Coverage Evaluation 
(A.C.E.).
    In fact, it is my understanding that using the A.C.E. population 
analysis tool the Census Bureau has now completed the process of 
producing adjusted census data down to the block level for all fifty 
states. That data may be more accurate than data derived using only 
traditional methods, but we simply do not know because that corrected 
data has not been made public.
    On March 1, 2001, Acting Census Bureau Director, William G. Barron, 
Jr. recommended to the U.S. Secretary of Commerce, Donald Evans, that 
unadjusted data be released as the Census Bureau's official 
redistricting data. Mr. Barron's recommendation was based on the 
recommendation included in the Report of the Executive Steering 
Committee for Accuracy and Coverage Evaluation Policy (E.S.C.A.P.).
    While that Report did in fact recommend that unadjusted data be 
released as the Census Bureau's official redistricting data, it also 
stated that the Committee believed that (1) unadjusted census totals 
would reflect a national undercount, (2) that there was considerable 
evidence to support the use of adjusted data, and (3) that further 
research may establish that adjusted data would result in improved 
accuracy. It is critically important to note that the Committee cited 
time constraints and a lack of sufficient information as key to its 
recommendation.
    In addition, in its Report, the Committee was clear that although 
its recommendation was to release unadjusted data, that decision did 
not mean or even suggest that unadjusted data was superior in any way. 
In fact, I believe that given the appropriate time and necessary 
resources to thoroughly and completely investigate the inconsistencies 
that led to what I consider to be a preliminary decision on the 
accuracy of adjusted data, the Committee would conclude, as it has 
already intoned, that adjusting census data would in fact result in a 
more accurate, true, and reflective evaluation of the nation's 
population.
    In a meeting with Secretary Evans, just days before the Secretary 
announced his decision to release less accurate unadjusted data as 
official redistricting data, the Congressional Hispanic Caucus 
requested that the Secretary require the Census Bureau to complete its 
work on Census 2000, and to subsequently make a recommendation 
regarding the use of adjusted data for other purposes, including the 
Intracensal Census.
    Specifically, the Hispanic Caucus requested that the Census Bureau 
be afforded any additional time and resources necessary to conduct 
further research into inconsistencies between estimates derived from an 
outmoded demographic analysis tool and the modern, scientifically 
enhanced and supported Accuracy Coverage Evaluation estimates. We 
believe that such an evaluation is critical to establishing the modern-
day reliability of the demographic analysis tool and subsequently to 
determine whether or not adjusted data would result in improved Census 
accuracy.
    While Secretary Evans stated support for moving full steam ahead, 
it appears that little if anything has been done by the Department of 
Commerce or at the Census Bureau to continue that evaluation process. 
The Congressional Hispanic Caucus is deeply concerned by this 
inactivity, particularly given the importance of determining the final 
accuracy or inaccuracy of the census and the impact that will have on 
Hispanic and other minority communities that were once again 
disproportionally undercounted.
    For them, the estimated 3.4 million Americans, including over a 
million Hispanics, who the 2000 census failed to count using 
traditional methods, this is the civil rights issue of the decade. If 
we fail to do all that we can to correct the problems, then those 3.4 
million people will simply not count for the next 10 years.
    We should all take a step back and consider what this would mean 
for people living in each and every community where minorities and 
children were disproportionally undercounted. It means a real and 
increased potential for another decade of unjustly lost federal dollars 
to build schools, roads, day care centers, and to fund countless other 
programs that rely on census data. It also means diminished political 
power for each of those communities. That is hardly the level of 
justice and fairness that we as a revered Democratic nation should 
aspire to.
    Therefore, in the spirit of fairness to all, regardless of age, 
ethnicity, race, or economic status, the Congressional Hispanic Caucus 
fully supports and encourages the Chair, the Committee, and the entire 
Congress to seek the release of the Census Bureau's adjusted census 
data so that it may be reviewed and where deemed appropriate applied.

    Senator Edwards. Thank you, Mr. Gonzalez. Thank you for 
your leadership of the Hispanic Caucus, and thank you for 
taking the time to be with us here today.
    Mr. Miller.

                 STATEMENT OF HON. DAN MILLER, 
                U.S. REPRESENTATIVE FROM FLORIDA

    Mr. Miller. Thank you, Senator, and thank the Committee for 
allowing me to testify today. As you may know, I have been the 
Chairman of the Subcommittee on the Census since 1988, and I 
have been deeply involved in the oversight of the 2000 Census 
from planning through its execution.
    Our public position on adjustment should not be a surprise 
to anyone. I have always been concerned that adjustment 
introduces more error into the census than it has the ability 
to correct, particularly at lower levels of geography critical 
for the redistricting process.
    However, even if sampling errors were not the issue, there 
are larger legal and public policy implications that should be 
giving all of us very serious concern. Acting Director Barron 
outlined many reasons for the success of the 2000 census 
enumeration when he testified before my subcommittee several 
weeks ago. Among those reasons, congressional support allowed 
the Census Bureau to hire \1/2\ million enumerators at 
competitive wages during a time of record low unemployment. 
This support also allowed for a first-time-ever paid 
advertising campaign, more than 40,000 local partnerships, and 
an unprecedented effort to provide multilingual assistance.
    In the end, it was emphasis on counting people, not making 
estimates, that made this census a success and reduced the 
differential undercount of minority communities. We need to 
give a great deal of gratitude for the great work the Census 
Bureau did on this, and we should commend the efforts of the 
Census Bureau.
    Supporters of adjustment say that success is not enough. 
They will argue for the release of these inaccurate numbers in 
the name of fairness and justice. I share their desire for 
fairness, but good intentions do not justify bad public policy, 
and a statistically adjusted census is bad policy on many 
different levels.
    Let me explain. Statistical adjustment is less accurate 
where it is critically important to be accurate, at small 
geographic levels. We have known for some time that 
statistically adjusted numbers do not give us a more accurate 
picture of the population when describing smaller aggregations 
of the population such as small towns, rural areas, and blocks. 
The bureau's recommendation against sampling reads, ``analysis 
for counties with populations below 100,000 people indicated 
that the unadjusted census was more accurate,''.
    Since apportionment and redistricting go hand in hand, it 
is arguably illegal for adjusted numbers to be used for 
redistricting, based on the 1999 Supreme Court ruling. Use of 
adjustment would reduce people's incentives to participate in 
the census, which would degrade data quality and reduce 
accuracy. Why stand up and be counted when you can sit down and 
be sampled?
    It is also disturbing to me that under adjustment, some 
people are counted as less than a whole person. Under 
adjustment, everyone is assigned to a category. One of these 
categories, for example, is Native Hawaiian and Pacific 
Islander women over 18 who own their own home. If adjustment 
were to go forward, everyone in that category would be assigned 
a value of .95. In other words, every Hawaiian adult woman 
owning a home anywhere in the United States who answered the 
census would be counted as little more than \9/10\ths of a 
person.
    Adjustment assumes that all people in a certain category 
act alike, or have the same likelihood of filling out their 
census forms. These types of assumptions are something we 
should be moving away from and not embracing.
    Finally, I worry about the subjective assumptions that are 
inherent to the statistical adjustment process. A professor at 
Harvard University put it very well in a commentary in the Wall 
Street Journal on February 15. He said,

          ``Unfortunately, statistical adjustment also gives much 
        greater discretion to the Census Bureau. The correction 
        procedure is based on populations of groups, and choosing them 
        is very subjective. Do we treat all young, urban black males as 
        a subgroup, or do we separate them by region? How many ethnic 
        groups do we want to treat as distinct? This leads to a general 
        point. As you allow for more statistical sophistication, you 
        put more discretion in the hands of the statistician.''

    This census and the Census Bureau have proven we can 
achieve nearly 100 percent accuracy through a strong 
congressional commitment, strong Bureau management, improved 
technology, and expanded local partnerships. By all means, 
there is still room for improvement, but this and future 
Congresses should put a priority on the methods that are legal, 
accurate, and ethical.
    Finally, on the question of releasing adjusted numbers, 
census data for Federal funding and other purposes, because of 
the errors and the problems with their adjustment, I think it 
would be highly irresponsible to release adjusted data for any 
official purpose.
    One of the things we have heard for the past several years 
in the Committee by my colleagues on the other side of the 
aisle is, listen to the experts. Let us trust the experts. 
Well, we have this ESCAP committee of Census Bureau experts, 
and the experts said there is a problem with the data, and so 
let us listen to them. The data should not be released at this 
time. Let them continue to work on that data and if at some 
stage they feel they have accurate data, then it can be 
released, but it would be irresponsible, in my opinion, to 
release that data now.
    Thank you, Mr. Chairman.
    [The prepared statement of Mr. Miller follows:]
                Prepared Statement of Hon. Dan Miller, 
                    U.S. Representative from Florida
    Thank you Mr. Chairman and Members of the Committee, for granting 
my request to testify before you today regarding the 2000 Census.
    As you know, as Chairman of the Census Subcommittee since 1998, I 
have been deeply involved in the oversight of Census 2000--from 
planning through its execution. My public position on adjustment should 
not be a surprise to anyone. I have always been concerned that 
adjustment introduces more error into the census than it has the 
ability to correct, particularly at lower levels of geography critical 
for the redistricting process. However, even if the sampling errors 
were not the issue, there are larger legal and public policy 
implications that that should give all of us serious concern.
    Acting Director Barron outlined the many reasons for the success of 
the 2000 census enumeration when he testified before my subcommittee 
several weeks ago. Among those reasons, Congressional support allowed 
the Bureau to hire a half million enumerators, at competitive wages, 
during a time of record low unemployment. This support also allowed for 
a first-time-ever paid advertising campaign, more than 140,000 local 
partnerships, and an unprecedented effort to provide multilingual 
assistance. In the end, it was the emphasis on counting people--not 
making estimates--that made this census a success, and reduced the 
differential undercount of minority communities.
    Supporters of adjustment say this success is not enough and they 
will argue for the release of these inaccurate numbers in the name of 
fairness and justice.
    I share their desire for fairness. But good intentions do not 
justify bad public policy. And a statistically adjusted census is bad 
policy on many different levels.
     Statistical adjustment is less accurate where it's 
critically important to be accurate--at small geographic levels. We 
have known for some time that statistically adjusted numbers do NOT 
give us a more accurate picture of the population when describing 
smaller aggregations of the population, such as small towns, rural 
areas and blocks. The Bureau's recommendation against sampling reads, 
``The analysis for counties with populations below 100,000 people 
indicated that the unadjusted census was more accurate.''
     Since apportionment and redistricting go hand-in-hand, 
it's arguably illegal for adjusted numbers to be used for 
redistricting, based on the Supreme Court's 1999 ruling.
     Use of adjustment would reduce people's incentive to 
participate in the actual census, which would degrade data quality and 
reduce accuracy. Why stand up and be counted when you can sit down and 
be sampled?
     It is also disturbing to me that under adjustment, some 
people are counted as less than a whole person. Under adjustment, 
everyone is assigned to a category. One of these categories is ``Native 
Hawaiian and Pacific Islander Women over 18 who own their home.'' If 
adjustment were to go forward, everyone in that category would be 
assigned a value of 0.95. In other words, every Hawaiian adult woman 
owning a home anywhere in the United States, who answered the census, 
would be counted as little more than nine-tenths of a person. 
Adjustment assumes that all people in a certain category act alike, or 
have the same likelihood of filling out their census forms. These types 
of assumptions are something we should be moving away from, and not 
embracing.
     Finally, I worry about the subjective assumptions that are 
inherent to the statistical adjustment process.
    A professor at Harvard University put it very well in his 
commentary in the Wall Street Journal on February 15. He said,

          ``Unfortunately, statistical adjustment also gives much 
        greater discretion to the Census Bureau. The correction 
        procedure is based on population subgroups, and choosing them 
        is very subjective. Do we treat all young urban black males as 
        a subgroup or do we separate them by region? How many ethnic 
        groups do we want to treat as distinct? This leads to a general 
        point: As you allow for more statistical sophistication, you 
        put more discretion in the hands of the statistician.''

    This census and this Census Bureau have proven that we can achieve 
nearly one hundred percent accuracy through a strong congressional 
commitment, strong Bureau management, improved technology, and expanded 
local partnerships. By all means, there is still room for improvement. 
But this and future Congresses should put a priority on the methods 
that are legal, accurate and ethical. We should not continue the 
pursuit of a costly, unlawful, inaccurate, and racially biased 
adjustment of our constitutionally mandated decennial census.
    Finally, regarding the question of releasing adjusted census data 
for federal funding and other purposes, because of the errors and 
problems with the adjustment, I agree with Bureau officials that it 
would be highly irresponsible to release adjusted data for any official 
purpose.

    Senator Edwards. Thank you, Mr. Miller.
    Mr. Gonzalez, in the 1990 census, over 6 percent of the 
Hispanic community in North Carolina was under-counted. During 
the last 10 years, according to the 2000 census, we have had 
almost a 400-percent increase in the Hispanic population. I 
cannot imagine what the undercount must be in this census.
    You and your family, your dad, has been involved in the 
Hispanic community and served as leaders in that community for 
a long time. I wonder if you would talk a little bit about what 
impact undercounting has, because these are real people. It 
affects their lives. They are not numbers on a piece of paper. 
I wonder if you would talk about the impact undercounting has 
on the lives of the people who are not being counted.
    Mr. Gonzalez. I think in terms of city planners, Senator 
and in terms of how you build, you know, smart growth and 
communities, your school administrators and trying to 
anticipate the number of students, and especially in a state 
like Texas, a city like San Antonio, 150 miles from the border, 
but all of the southwest states with some of the greatest 
growth, your four largest growing cities, really, it is because 
of the Hispanic populations, and that is not even to say what 
is going on in states such as yours.
    But there is a practical application. These numbers are 
used to plan ahead, not just for funding purposes, but to get 
accurate head counts on what you need as far as medical 
facilities, our county hospital, which has a tremendous burden 
of taking care of many people that are not insured. Everything 
goes hand in hand here. The uninsured problem, the number of 
people, we have to get an accurate number.
    How do you plan how many students are gong to be in class? 
How many need Head Start? How many need meals? But it goes 
beyond that, in roads and highways.
    I was thinking the other day, even Meals on Wheels is 
predicated on this. I had a meeting the other day with the 
Small Business Development Center in San Antonio. They service 
79 counties in South Texas with a heavy Hispanic population, 
and with the cuts on the SBA, you can imagine how that is 
basically almost--the effect is multiplied, when we do not have 
accurate numbers.
    We have reduced amounts in the budget. We are all talking 
about keeping within budget caps, there is going to be reduced 
spending. Every dollar becomes more precious. How do we apply 
these dollars in the best way, the most economical and 
effective way, if we do not have accurate numbers, and that is 
what the cities are faced with. That is what my district is 
faced with.
    In talking to the University of Texas in San Antonio and 
their community development center, they are telling me any 
cuts will be drastic in trying to take care of the small 
businessman and woman in 79 counties. It is important to have 
this information, and have it now, so that we can have our 
experts look at this information.
    And getting back to a very important point that my 
colleague in Florida has pointed out, their big concern has 
been, is the scientific method loses some of its accuracy at 
the block level.
    If you believe in ESCAP, which I do, on page 27, block 
level accuracy is not an important criterion to evaluate either 
Census 2000 or A.C.E., so they made that conclusion. They have 
made that finding. They have addressed that concern.
    This is all really important. I sit down with my city 
council, this is what they are bringing to my attention.
    Senator Edwards. Thank you, Mr. Gonzalez. It is so 
important that we put a human face on this. These are not just 
numbers.
    Mr. Clay, I wonder if you could comment, because I suspect 
some these are some of the same concerns you might have in the 
African American community.
    Mr. Clay. Mr. Chairman having spent 17 years in the 
Missouri General Assembly, and having some experience with 
pass-through Federal dollars that we pass on to local 
communities in Missouri, it is so important, such as Head 
Start, we know how important it is to get an accurate count of 
people throughout the state of Missouri.
    It seems to me that in urban areas, especially the area 
that I represent, that the needs are so great that you rob 
those communities of needed resources when you do not 
accurately count those communities, so what happens is that 
those communities where overcounts occur are usually the ones 
that get extra resources, and the ones where the need is the 
most get the least of those resources.
    Senator Edwards. For the reason that both of you have 
described so eloquently, I think it is critically important 
that we have the census be as accurate as possible. I also have 
great difficulty understanding why the adjusted numbers are not 
being released and not subject to public scrutiny, which I 
believe they should be.
    Senator Kerry, did you have questions? Mr. Miller, I think, 
had something he wanted to say.
    Mr. Miller. Let me add a couple of comments. The question 
as to the relief is whether they are accurate enough, and there 
is very serious concern by the statisticians at the bureau, the 
professionals, not politicians, that at this stage it is not 
accurate enough, so if we are going to trust the professionals, 
we need to do that.
    One of the questions comes up about money. First of all, 
everybody agrees we have had a great census this time, and we 
really should celebrate the census. The differential undercount 
for Hispanics and African Americans has been cut in half so we 
have really had a great effort, and Mr. Gonzalez has worked 
hard in his district, working with the Catholic Church, the 
Hispanic Church and all of that, so we have done a good job in 
executing a better census.
    With respect to the money issue, most of the money we are 
talking about is a zero sum game, and so everybody is saying we 
are going to lose all this money. Well, if, for example, Meals 
on Wheels, it is important in my district, lots of seniors in 
my district. There is a set pot of money for Meals on Wheels.
    The question is, you are not going to get more money. It is 
just the total amount of money. That is a budget fight rather 
than a census fight, so I just think sometimes it is distorting 
the thing to say we are missing all this money. Well, if you 
add more money in his district, you are taking money from my 
district, so it is really a zero sum game when we are talking 
about money.
    Yes, we need to have the most accurate census possible, and 
I think we did a good one, and I think we should really be 
proud of what the bureau did under the leadership of Dr. 
Prewitt and John Thompson and the others at the bureau.
    Thank you.
    Senator Edwards. Mr. Gonzalez, Mr. Clay, did you have any 
comment in response to that?
    Mr. Gonzalez. Well, I understand we all have competing 
interests. We represent our constituencies. But what if I just 
happen to have a certain segment of the population that a 
certain program is attempting to address their needs more so 
than another colleague. I can make that argument with census 
figures. I cannot make that argument if I do not have accurate 
numbers.
    And I know this debate really goes almost to a political 
philosophy, or whatever, and it should not be. It is really one 
about sound science, and we are willing to debate that. If you 
want to go to court we are willing to debate it in a courtroom. 
I can guarantee you now that this will pass legal muster when 
it comes to scientific methodology and what is accepted out 
there in the scientific community.
    Thank you.
    Mr. Clay. Mr. Chairman, just to add to that, I would think 
that the Census Bureau as well as the U.S. Government would 
want to get it right, would want to be as accurate as possible.
    I mean, as I stated in my testimony, we can go back to 1940 
when the undercount was so drastic among African American males 
that they did not realize that they had made an error until 
these African American males showed up in greater numbers for 
the draft for World War II, and then the U.S. Government 
realized how severe the undercount was in that community, so 
that needs to be corrected, and that is something that happened 
more than 60 years ago.
    Senator Edwards. Mr. Clay, Mr. Gonzalez, I share your 
concern. Mr. Kerry is now back and may have some questions for 
you.
    Senator Kerry. Thank you very much, Senator Edwards. Thank 
you for relieving me for that period of time. I appreciate it.
    Mr. Miller, let me just comment very quickly on your 
comment about sort of the zero sum game. In one sense, yes, 
there is a specific amount of money, and it depends where it 
goes, but it really runs a little deeper than that, because 
when you talk about where it may go from, you are talking about 
the difference between affluent communities and communities 
that are struggling with resources, and you are talking about 
the difference of a community that has no tax base for property 
taxes, and yet their schools depend on it, and here you have an 
allocation that may be dependent on population as you do in 
certain school issues, or seniors, as Congressman Gonzalez has 
said, so it really is a question of fundamental fairness.
    Secretary Evans acknowledged there is an undercount. I 
assume if you accept the notion that there is an undercount, 
you accept the notion that there is something unfair. Do you 
accept that concept, if there is an undercount?
    Mr. Miller. The differential undercount is what we are all 
concerned about. Blacks, Hispanics, for example, are counted 
low, and not as accurately as whites.
    Senator Kerry. Do you think they should live with that for 
10 years?
    Mr. Miller. No. We should try to get the most accurate 
census we can, but if the adjustment is not accurate, if the 
statisticians and the professionals and the career people in 
the bureau say the numbers are not accurate, would we want to 
use inaccurate data.
    Senator Kerry. But they are not saying that they are 
inaccurate. They are trying to figure out where the 
distinctions may be. Now, why would we not want competent 
professionals to view the data--I mean, are you suggesting that 
some of these people sitting at this table are not 
professionals?
    Mr. Miller. No. I think we need to wait for the bureau to 
continue doing their work. They want to release the numbers 
before the bureau has certified the numbers.
    Senator Kerry. Well, the problem is, we could run out of 
time. We could run out of time to measure and come up with a 
solution in which we share some thinking about how you rectify 
some of these differences.
    I mean, I would like to have outside professional help me 
to say, well, Senator, here is how you could really do this in 
a fair-minded, thoughtful way, but if we allow this to be 
behind closed doors for the next months without any capacity to 
measure the data, then we are stuck with whatever judgment is 
made.
    Mr. Miller. But the professionals should have a chance to 
look at it, and they have reached an agreement between the 
bureau and the Republicans, the Democrats, and other outside 
groups to have a chance to look at it, and they have also 
contracted with the National Academy of Sciences, which is very 
independent, on doing an analysis of the same numbers.
    Unfortunately their analysis is not going to be ready until 
probably early next year, so it is a real mistake to try to use 
faulty data until the bureau, until the National Academy of 
Sciences, until the professionals sitting at this table and 
others have said they are accurate numbers. If you do not have 
accurate numbers, you are making a real mistake playing around 
with potentially faulty data. We do not know yet.
    Senator Kerry. Does it mean anything to you at all that 
this is so divided along partisan lines?
    Mr. Miller. It is unfortunate.
    Senator Kerry. Is it more than unfortunate? Is there a 
message there?
    Mr. Miller. I used to teach statistics. That is how I got 
to be Chairman of this particular subcommittee, but we have to 
base it, as we said all along, on science, and if the science 
says it is not right, why would we want to do something wrong--
it is like a medical test.
    Senator Kerry. I do not want to overrule science, believe 
me. If I started doing that for this, I would be in trouble on 
everything else. None of us are asking to overrule science, but 
there are three different methodologies here for how you arrive 
at the ball park. One, you have the Demographic Analysis, which 
they acknowledge has a problem with respect to immigration, You 
have the A.C.E., which we do not yet understand completely what 
the differential is, then you have the actual count.
    Now, if all of them are showing an undercount, it seems to 
me it is like a court of equity, where a judge sits there and 
makes some judgments. Well, how are we going to make up for 
that undercount level we have counted? You are shaking your 
head, Mr. Murray. You do not think we should do that? We are 
going to come back to you.
    But let me just ask Congressman Gonzalez and Congressman 
Clay a question, sort of following up on what Senator Edwards 
said. Your constituents, or the people in your districts who 
are cognizant already, or who have knowledge that there is not 
a count that is not fair, I mean, do they come up to you and 
talk to you about it?
    Mr. Gonzalez. Senator, as my colleague was indicating, we 
all came together, the city council, commissioners, the courts, 
the church, every social and civic organization, we came 
together to get people to make sure they return their forms, 
answer the door, that anybody who is not documented did not 
have fear of being arrested. It was an incredible effort, and 
we did end up with a more accurate result, but nevertheless one 
that is still inaccurate and leaves many undercounted, 
especially in the minority Hispanic communities. There is a 
very high awareness in my city that we are going to be 
basically short-changed.
    Senator Kerry. What is the impact of that? What is the 
effect of that on the body politic?
    Mr. Gonzalez. As I have indicated, of course just 
planning----
    Senator Kerry. Leaving out the planning.
    Mr. Gonzalez. The funds. We lost a substantial amount of 
Federal funds in the past 10 years because we missed 40,000 
people. We missed 16,000 children in San Antonio. We are part 
of a lawsuit The city council has authorized the city to go 
into the lawsuit--I guess it is out of Los Angeles, or 
whatever--regarding trying to get these numbers and having the 
adjusted figures utilized.
    All of this impacts services to the neediest of all 
citizens, and we are talking about, and some are not citizens 
but residents, and part of this, Senator, is trying to figure 
out why we have this discrepancy. Everyone is shocked at the 
number of immigrants in this country, of the number of 
undocumenteds. It has repercussions for Immigration and 
Naturalization Service and Social Services, as I have already 
indicated.
    Senator Kerry. Well, let me ask you a very hard-nosed 
question, if I may, and welcome, Congresswoman Maloney, thank 
you, and let me just say to all of you, I apologize profusely, 
we have had to go back and come back and vote, and I know how 
hard that is, and I apologize for the interruption. It is just 
the nature of the beast around here, and we all try to do the 
best we can.
    I want to ask one last question, then I am going to try 
to----
    Mr. Clay. Mr. Chairman, if I could just follow what Mr 
Gonzalez said, that also in the St. Louis community people are 
wondering just when will they be counted. They have witnessed 
the fiasco in Florida, where votes were not counted for the 
Presidential election, and now they are witnessing a back-
pedaling of sorts by the Census Bureau, and when will they 
count.
    Senator Kerry. This brings me to my question, and I am 
going to be sort of devil's advocate here. You know, some 
people sit on the other side of the fence, and they would say 
to uncounted people and to you, ``Well, wait a minute, they had 
a responsibility to fill out their census form''. We had the 
most broadbased outreach ever in history, and others filled it 
out and they got counted. Why do we have to make up for the 
seemingly unwillingness of these folks to participate? That is 
the question you get, and what is your response?
    Mr. Clay. Well, the system was quit a bit better, as Mr 
Miller says. I agree, the Census Bureau did a better job than 
in 1990. However, there for systematic reasons African 
Americans were missed in a disproportionate number, compared to 
the rest of the population, as well as Hispanics. They were 
missed also, because of system flaws, in just raw data, in 
counting people, so when we know that there is an undercount it 
is incumbent upon us, as a government, to go back and try to be 
as accurate as possible.
    Senator Kerry. Would you say that to a certain degree that 
some folks whose level of education may be higher, or whose 
communication skills may be better, or whose status and 
position in life are better for whatever reason, that there may 
be sort of a disadvantage in life that at this point plays out 
in how the census in fact is counted?
    Mr. Clay. Well, I would not go that far. What I would say 
is that people earnestly made an effort to be counted. If some 
were missed, we need to figure out why they were missed, and we 
need to go back and get as accurate a count as possible.
    Senator Kerry. Let me let your colleague have a chance.

              STATEMENT OF HON. CAROLYN MALONEY, 
               U.S. REPRESENTATIVE FROM NEW YORK

    Ms. T4Maloney. I just would like to add to the discussion 
that the census is one of the great civic ceremonies that we 
all participated in and saw our responsibility to fill out our 
forms and be counted, and it is important not only for our own 
families, but for our neighbors, because if we are not counted, 
and we are left out of Federal funding formulas to the tune of 
$185 billion a year that are distributed on census numbers 
representation, which is the true power in this country, and 
just plain good data to plan for the future.
    But we know that the people missed overwhelmingly are in 
urban and rural areas, and that they are oftentimes minorities, 
children, and the poor, and given the changing dynamics of 
America, the fact that some families are working two, three 
jobs, some families are backed up in apartments where two or 
three families are living there possibly illegally, and they do 
not want people to know, sometimes there is a language barrier.
    And given the extraordinary efforts of the Census 
Department and the many organizations that partnered with the 
Census to break through that barrier, it is still a barrier, 
and we know from modern scientific methods that there is an 
undercount, there's no question about it, so the question is, 
do we correct it or not, and we have the scientific ability to 
correct it, and why we would not go forward with the most 
accurate numbers is something that I cannot understand.
    I congratulate you, Senator, for your interest in this, and 
for your leadership in it. I consider it really the civil 
rights issue of the decade. It is representation from which all 
power flows, and if you are not counted, then the school is not 
where it is supposed to be, the health center is not there, and 
I would merely like to place in the record a letter that has 
been signed by 107 colleagues of mine in the House of 
Representatives calling for the release of the numbers.
    The American people paid for it, the Congress allocated--
over $400 million was invested in this information, and it 
belongs to the American people. Why are we not releasing it? It 
is absolutely unprecedented that such a valuable information 
should be withheld.
    Furthermore, no community in our country should be left 
short-changed. Every state, every city, every town, and every 
neighborhood is entitled to information about their area, and I 
do not understand any logic why they would withhold it. I would 
like really to place into the record--we are all in different 
committee meetings and voting on the floor--my comments, but I 
would like to make them very short, but first of all the census 
has always been very contested.
    The first census director was Thomas Jefferson, and he 
wrote Washington and said there was an undercount of a million 
people. Washington took him at his word and put a million more 
in the first census count. That is a true story.
    Unfortunately, it has been much more contested in this 
Congress. It has regrettably been the most partisan of issues. 
It held up two budgets. It held up disaster relief.
    But it is very important, because it goes to the absolute 
core of our democracy. If you are not counted, you are not 
represented, and I just want to say that the so-called sampling 
method is known in the scientific community as the dual system 
estimation, and it is a proven statistical method that has been 
perfected by scientists during the last half-century.
    The 1990 sampling results have been the foundation of 
nearly every major economic statistic of this country since the 
middle of the last decade they are using modern scientific 
methods, and in this A.C.E. follow-up we paid for a sample of 
314,000 households, which was then compared to the data 
collected from those same households during the census itself, 
and the so-called accuracy and evaluation, or A.C.E., is known 
as the census decennial report card, and we should really have 
these results.
    The census professionals in their report stated that they 
favor the use of modern scientific methods, and I quote, ``the 
committee believes it likely that further research may 
establish that a judgment based on A.C.E. would result in 
improved accuracy,'' and I just want to underscore that even 
the opponents of sampling have apparently blessed this 
methodology, although they may not really fully realize that 
they have done so.
    The administration has pronounced over and over again, we 
heard it today, that the 2000 census is the most accurate in 
history, and it came to this conclusion because of the results 
of the very scientific tool they oppose, the A.C.E. The A.C.E. 
program demonstrated that the 2000 census reduced the 
undercount, as compared to the 1990 census. As a result, they 
have decided there is no need to use the A.C.E. for its full 
intended purpose, to correct for the errors that remain in the 
Census 2000, and for the administration the A.C.E. is 
sufficient only to buttress their own political argument 
against a more complete count.
    Unless we have access to the A.C.E. data, we cannot prove 
that the Census 2000 is, in fact, more accurate, so they 
really, really should release it. Why the administration is 
keeping it under lock and key, I do not understand. It is 
harmful to government, and to government planners, and I might 
note, Mr. Chairman, that the Census Monitoring Board released 
the net undercount numbers, and I would like to know why they 
are releasing it before the bureau does, and they took this 
extraordinary step of presenting what they believe are the net 
numbers for each state that were missed in the census, because 
the Census bureau itself will not release them.
    I would just give the numbers for my own state. You can get 
them for all the states from the monitoring board, and these 
numbers show that there was a net number of 291,000 New Yorkers 
were missed in the census, and 188,000 in New York City alone, 
merely because the Census Bureau ran out of time, and according 
to an independent survey that was done by Pricewaterhouse this 
would result in approximately a loss of $2.3 billion over 10 
years.
    That is a lot of teachers, that is a lot of police 
officers, that is a lot of mass transit, and bridges, and 
roads, and to know that the accurate numbers are there and that 
they are not releasing them is just absolutely unconscionable. 
I, on behalf of my colleagues in the House, particularly the 
107 who signed the letter appealing for the release and the 
many people in America who would like to be counted but have 
been left out, we thank you, Mr. Chairman, for your attention 
to this and for this hearing, and for the time and 
understanding you have put in it, and I congratulate you and 
thank you.
    [The prepared statement of Ms. Maloney follows:]
              Prepared Statement of Hon. Carolyn Maloney, 
                   U.S. Representative from New York
    Thank you Chairman McCain and Ranking Member Hollings for inviting 
me to testify today on the Census. Mr. Chairman, welcome to the 
``sampling debate.'' By that, I mean the national dialogue which has 
been held during the last decade in search of ways to produce a better 
Census, one that would cure the national problem of the undercount and 
racial differential.
    It is a problem that has existed in every Census, beginning with 
the very first census conducted by Thomas Jefferson. The Secretary of 
State wrote to President Washington and our ambassadors overseas saying 
that he felt that the just-concluded census may have undercounted the 
fledgling nation's population by as much as a million. Perhaps, because 
he was Jefferson, he was taken at his word.
    In today's modern census, the Census Bureau has to do more than 
assert the quality of the census. It must demonstrate to Congress and 
the American public that it knows with some precision about the 
accuracy of the so-called ``head count.'' The modern solution is 
sometimes derisively called ``sampling,'' but is better known in the 
scientific community as ``dual system estimation.'' It is a proven 
statistical method that has been perfected by scientists during the 
last half century. The 1990 sampling results have been the foundation 
of nearly every major economic statistic of this country since the 
middle of the last decade.
    In Census 2000, the Congress funded the most robust scientific 
measurement of accuracy ever conducted. We paid for a follow-up sample 
of 314,000 households, which was then compared to the data collected 
from those same households during the Census itself. The Accuracy & 
Coverage Evaluation, or A.C.E., is known at the Census Bureau as their 
decennial report card. The results are in for Census 2000, and the 
bureau has announced that the A.C.E. worked extremely well and that the 
Census staff believes as they made clear in their report, that it still 
favored statistical sampling, ``The Committee believes it likely that 
further research may establish that adjustment based on the A.C.E. 
would result in improved accuracy.".
    Even the opponents of sampling have apparently blessed this 
methodology, although they may not fully realize what they have done. 
The Administration has pronounced the 2000 Census the most accurate in 
history. It came to this conclusion because the results of the very 
scientific tool they oppose, the A.C.E., demonstrate that the 2000 
Census reduced the undercount as compared to the 1990 Census. As a 
result, they have decided that there is no need to use the A.C.E. for 
its full intended purpose, to correct for the errors that remain in 
Census 2000. For the Administration, the A.C.E. is sufficient only to 
buttress their own political argument against a more complete count.
    Now that may be a legitimate call, but unless all of us--the 
Congress, outside experts, the scientific community, and the American 
public--unless we all have access to the A.C.E. data, we can not prove 
Census 2000 is in fact more accurate. And I tell you Mr. Chairman, I 
have yet to meet any Thomas Jefferson's in this Administration, so I am 
unwilling to just take their word for it.
    So here we are today, holding this hearing to ask why the A.C.E. 
results are under lock and key, known only to a handful of government 
employees, and when even Congress itself has not been provided the 
data. Why is the Census Monitoring Board releasing net undercount 
numbers before the bureau? They took the extraordinary step of 
presenting what they believe are the net numbers for each state that 
were missed in the census because the Census won't release them. Those 
numbers show us that a net of 291,000 New Yorkers were missed in the 
Census, 188,000 in New York City alone, missed because the Census ran 
out of time.
    Apparently, the Administration fears that we will not understand 
the data, or it may be used to advance political arguments the 
Administration opposes, or that we should simply have to trust them to 
look out for our best interests.
    Well Mr. Chairman, I think we understand the numbers release today 
very well--Americans were missed in the Census and we should not stop 
until we get them all included.
    For another perspective on this debate Mr. Chairman, I direct you 
to remarks made just yesterday by Chairman Greenspan before a meeting 
of business economists. His presentation was titled, ``The challenge of 
measuring and modeling the dynamic economy.'' His address was, in part, 
a call for more resources to enhance data collection methods. He said:

        ``. . . the experience of the last 40 years underscores a 
        fundamental dilemma of business economics. Should we endeavor 
        to continue to refine our techniques of deriving maximum 
        information from an existing body of data? Or should we find 
        ways to augment our data library to gain better insight into 
        how our economy is functioning? Obviously, we should do both, 
        but I suspect greater payoffs will come from more data than 
        from more technique.''

[Remarks by Chairman Alan Greenspan ``The challenge of measuring and 
modeling a dynamic economy'' At the Washington Economic Policy 
Conference of the National Association for Business Economics, 
Washington, D.C. March 27, 2001]

    What Mr. Greenspan knows and others often overlook, is that our key 
economic indicators, such as unemployment, poverty, inflation, and 
consumer confidence are all derived from ``samples'' of American 
households and businesses. This sample data is collected through 
interviews every month, conducted mostly by none other than the Census 
Bureau.
    I agree with Chairman Greenspan's main thesis. There is a greater 
payoff for our country when we in Washington have more data, better 
data, complete data . . . data like that produced by the A.C.E. 
Statistically sound data allows policy makers and experts to calculate 
with more precision the true stories of economic or social policies or 
as former Census Director Prewitt might say, allows us to get closer to 
the truth. Mr. Chairman, it is time now for the Census Bureau to 
release all the A.C.E. results. I urge you and both Houses of Congress 
to call upon the Director of the Census to release those numbers today. 
I would like to put into the record a letter sent yesterday to Acting 
Director Barron from 107 House Members urging just that.
    It is unprecedented that such valuable information should be 
withheld. Furthermore, no community in our country should be left 
shortchanged. Every state, every city, every town, and every 
neighborhood is entitled to information about the demography of our 
country.
    Together we made a four hundred million dollar investment in this 
information. It belongs to the people and it should be given to the 
people.

    Senator Kerry. Thank you very much, Congresswoman. None of 
the Members of Congress need to feel obligated--obviously, I 
know you will not, if you need to move on. I do not want to tie 
you down, because I know your time is critical and you have got 
other things going on. Is there any last comment any of you 
would like to make as we move back, because I want to get back 
to the panel, if we can.
    Mr. Miller. If I may make one last comment, Mrs. Maloney 
has been the Ranking Member for the past 3 years on the 
Subcommittee, and she has said rely on the professionals. Well, 
the professionals say the data is not accurate enough to be 
released at this time, and in 1990, when they tried to do an 
adjustment, there was a difference--it was not called A.C.E. 
back then, it had another name, but they went through several 
different iterations of different sets of numbers.
    So if you release them before you finalize the numbers you 
are creating a problem. That is what the argument is over, 
whether you release them now, or let the Academy of Sciences 
and let the professionals like these gentlemen sitting here 
have the chance to be sure they are accurate, so let us trust 
the professionals of the bureau. They are the ones who say we 
are not finished with this, we have got to keep working on it, 
and they have got the resources.
    We certainly have thrown all the money that they need to 
continue that job, so we need to continue having them work, and 
once they have made that decision, then we can decide what to 
do with that data.
    Ms. Maloney. I just might, since money was mentioned, I 
certainly would like the opportunity to respond, if I could, 
Mr. Chairman.
    Senator Kerry. It was mentioned with great credit.
    Ms. Maloney. The census professionals merely ran out of 
time, and they stated that they favored statistical sampling, 
and I quote, ``the committee believes it likely that further 
research may establish that adjustment based on the A.C.E. 
would result in improved accuracy,'' and what I have heard 
today from the very studied questions of the chairman, I 
believe the chairman asked, when will you release this, and it 
was not said when it would be released.
    They said they may get the numbers by the summertime, but 
never did I hear that they would be released, so what we are 
calling on is for the information to be released so that the 
scientists, the government planners, everyone can look at it.
    And then secondly, the chairman and others questioned 
whether or not the numbers could be used for the distribution 
of federal funds, and I have not heard a response on that, and 
silence is usually no, and if we know that we have a more 
accurate number, then it is only fair and just that it be used 
for the important distribution of federal funds.
    Senator Kerry. I think, Congresswoman, that--you may have 
just departed for the vote at that point. I did ask him if we 
would be able to get him back here to address their judgment in 
order that we be permitted to apply these to the allocation of 
funds, and he did consent to do that, and suggested they would 
be able to have those numbers by the early summer, and so 
hopefully--and we agreed to continue the discussion, so 
hopefully we will be able to have a dialog that will permit us 
to make those judgments.
    Now, you are correct, there was no consent to release, 
despite the request to do so. There was a statement about why 
they saw difficulties in doing it now. I think we need to 
continue to press that issue. I certainly intend to, and others 
will, and we want to explore that a little bit with the panel 
of experts we have here in front of us, and I intend to do that 
now a little bit, and so I am very grateful to all four of you. 
You have been extraordinarily gracious with your time, and with 
the inconvenience, and I thank you very, very much for doing 
so. Thank you for being part of this.
    Now, had you all had a chance to testify? Not yet, so you 
are waiting patiently. Dr. Ericksen, you are next in line, and 
then Dr. Wachter. Thank you, and thank you all for your 
patience, too.

         STATEMENT OF EUGENE P. ERICKSEN, PROFESSOR OF

                SOCIOLOGY AND STATISTICS, DEPARTMENT OF

                      SOCIOLOGY, TEMPLE UNIVERSITY

    Dr. Ericksen. First of all, thank you for the opportunity 
to discuss this issue. I think first we need to be clear about 
what we mean by census error, because that is what determines 
whether or not this is the most accurate census ever. The net 
undercount in the 2000 census is estimated to be 1.2 percent, 
or 3.3 million people. That is not a particularly good way to 
assess the census.
    A better way to assess the census accuracy is in terms of 
what we call gross coverage error, which is the sum of the 
number of omissions and the number of erroneous conclusions. 
The Census Bureau, working with the A.C.E. data, has released 
estimates that vary between 9.5 and 11.9 million.
    The third criterion, which may be even better, is a 
differential undercount, comparing the A.C.E. estimates in 2000 
with the post enumeration survey estimates in 1990. It appears 
that the bureau was able to reduce the differential undercounts 
of Hispanics and African Americans by more than half. All of 
these indicators place great credit to the Census Bureau. They 
have definitely made an improvement, but what we must be most 
concerned about is not so much the error at the national level, 
but whether or not it creates an uneven situation locally.
    In the 1990 census, the Census Bureau estimated there was a 
net undercount of nearly 8 percent among black renters living 
in New York City, and in the same census they estimated a small 
overcount of 0.23 percent among white homeowners also living in 
New York City. According to these estimates the most 
undercounted congressional district in the United States is 
District 16 in New York State, of 6.5 percent. The 3rd District 
in New York State had an overcount of 1.3 percent, and so it is 
the variations in the undercount and the overcount are what we 
should be concerned with.
    Now, the Census Bureau actually had tremendous discretion, 
and made several important decisions in the manner of taking 
the census. One very important decision is the program known as 
local update of census addresses, LUCA. Millions of addresses 
were added to the census, and as far as the research I have 
been able to do, and the reports that I have read, lead me to 
believe, it could very well be that this is the most important 
thing that the Census Bureau did to improve the accuracy of the 
2000 census.
    The problem with LUCA is that it was not applied evenly. 
LUCA was more effective in some areas than it was in others, 
and it could have contributed to error, although it improved on 
the national level.
    A second decision that is important is the decision about 
what we call whole person imputations. In the 2000 census the 
bureau ended data collection at a much earlier date than it did 
in the 2000 census, and it appears that one of the outcomes of 
this early decision is that there is a greater use made of the 
computer in filling out, I believe it is 5.5 million records. 
The comparable number in the 1990 census was about 2.1 million.
    Whether or not the whole person imputation, virtual people, 
if you will, improved the accuracy of the census remains to be 
studied, but if it did improve the accuracy or degraded the 
accuracy, it probably did that unevenly, because some areas had 
a greater incidence of these kinds of additions, and others, 
about half of the races of the people are other than non-
Hispanic white. In other words, they are minorities.
    Now, I have tried, using the data that are available to me, 
to evaluate the differential undercount among areas. Among the 
states, it appears that the deviations are not particularly 
great. Those states where the minority percentages are higher 
tend to have a higher undercount. Those states where the 
minority percentages are lower tend to have a lower undercount, 
but the differences between the higher states and the lower 
states are not great, but they do exist.
    There is another situation. I am from Philadelphia. Every 
year, particularly once we have had 5 or 6 years go by the 
census, I get phone calls from reporters. They want to know, 
why is the population of Philadelphia getting so small. Indeed, 
I believe that it was thought that Philadelphia had more 
population loss than any other city in the country.
    Now, because the 2000 census did a better job of reducing 
the differential undercount, it appears that Philadelphia did 
not lose as many people, and so now we are in the anomalous 
situation that the very improvement in the 2000 census creates 
errors in understanding the rate of population growth.
    In order to figure this out, we need block-level data, 
because we need not so much to estimate the growth of any 
particular block, but we need to have the flexibility to put 
different combinations of blocks together to understand and try 
to understand how much of the change in population is real 
growth, or real decline, and how much of it is simply changes 
due to the way they took the census.
    So I think we are in a situation to say that the Census 
Bureau did an excellent job. They had a lot more money to 
spend. I think that the issue of Demographic Analysis is an 
important one. Unfortunately, while we spent so much money on 
the collection of the data, it appears we have underfunded the 
Demographic Analysis research at the Census Bureau.
    My colleagues who have looked at this tell me that the 
Demographic Analysis agrees with the results of the A.C.E. 
pretty accurately for non-Hispanic blacks and non-Hispanic 
whites The difference lies in the Hispanic population, which is 
probably due to problems measuring immigration, which we know 
to be a very difficult thing to measure.
    I will conclude at that point and answer questions, if 
there are any.
    [The prepared statement of Dr. Ericksen follows:]
 Prepared Statement of Eugene P. Ericksen, Professor of Sociology and 
         Statistics, Department of Sociology, Temple University
                            1. introduction
    I am a Professor of Sociology and Statistics at Temple University, 
where I have taught since 1971. I teach courses on survey design and 
methods, general statistics, and demography. Last year, I taught a 
graduate seminar on the United States census, with an emphasis on the 
nature, causes, and consequences of census error.
    I completed my doctoral dissertation at the University of Michigan 
in 1971. In this dissertation, I developed a method for calculating 
local population estimates when census data are unavailable. The Census 
Bureau provided financial support for this work. Since completing the 
dissertation, I have done substantial research on methods of 
calculating population estimates, both in census and non-census years. 
Over the past two decades, I have extended my research into the area of 
estimating local undercounts.
    In October 1980 I advised plaintiffs in a lawsuit filed by the City 
and State of New York attempting to compel an adjustment of the 1980 
Census. Between October 1989 and July 1991, I served as Co-Chair of a 
Special Advisory Panel appointed by then-Secretary of Commerce Robert 
Mosbacher to advise him on the possibility of adjusting the 1990 
Census. After completing my duties on the Advisory Panel, I again 
advised plaintiffs in a lawsuit seeking to compel the adjustment of the 
1990 Census. For both the 1980 and the 1990 Census, I conducted 
substantial amounts of research on data relevant to the undercount.
    In June 2000, the Census Monitoring Board, Presidential members, 
hired me as a statistical consultant. My duties have included reviewing 
Census Bureau plans and reports regarding the 2000 Census, and 
analyzing certain data from the census that the bureau has provided to 
us. In particular, I have had access to the individual level data 
records of persons included in the Accuracy and Coverage Evaluation 
(A.C.E.) samples used to estimate the levels of undercounting and 
overcounting in the 2000 Census.
                        2. basic census concepts
    There have been three major controversies of the 2000 Census. The 
first, settled by a Supreme Court decision, was whether sampling could 
be used as part of Non-Response Follow-Up (NRFU). The second, decided 
by Secretary of Commerce Donald Evans, was whether the results of the 
census should be adjusted to correct for the differential undercount of 
minorities for redistricting purposes. The third, still undecided, is 
whether the adjusted results should be released for public use. I 
understand that the Census Bureau has already calculated block level 
adjusted counts for the entire nation.
    These controversies are intimately related. To discuss them, it 
helps to define two terms, ``net undercount'' and ``gross coverage 
error.'' The net undercount is the difference between the number of 
people counted in the census and the bureau's independently calculated 
estimate of the national population. The net undercount was 1.6 percent 
in 1990. The bureau estimates that it fell to 1.2 percent, or 3.3 
million people, in 2000.
    These 3.3 million people are not the total number of people missed 
from the census. The net undercount is the difference between two 
quantities, omissions and erroneous inclusions. Omissions are people 
who should have been counted, but were not. Erroneous inclusions are 
counts that should not have occurred. They are frequently duplications 
of the same person counted in the same place. They may also be double 
counts at separate addresses. For example, college students may be 
counted not only at the dormitory where they actually live, but also by 
their parents at home. A family with a second home, perhaps used for 
weekends and vacations, may get counted at each address. It is 
conceivable that there could be large, but equal numbers of omissions 
and erroneous inclusions. Should this occur, the net undercount would 
be zero, and useless for the evaluation of census error. If the 
omissions and erroneous inclusions occur at different locations, some 
local areas would have overcounts and others would have undercounts 
even though the national net undercount was zero. For example, in 1990, 
the Census Bureau estimated a 7.76 rate of undercount for Black renters 
living in New York City. In that same census, there was an overcount of 
0.23 percent among ``non-Hispanic White and Other'' homeowners also 
living in New York City\1\.
---------------------------------------------------------------------------
    \1\ Howard Hogan, ``The 1990 Post-Enumeration Survey: Operations 
and Results,'' Journal of the American Statistical Association, 88: 
1047-1060, Table A.1.
---------------------------------------------------------------------------
    There were several million erroneous inclusions in both the 1990 
and the 2000 Censuses. Since these must be subtracted from the number 
of omissions to derive the net undercount, the actual number of people 
missed from the census is much greater than the net undercount. The 
Census Bureau has asserted that 8.4 million persons were omitted from 
the 1990 Census. The comparable number for the 2000 Census is between 
6.4 and 7.6 million persons (see Table 1).
    The ``gross coverage error'' is the sum of omissions and erroneous 
inclusions. In 1990, the Bureau told us that there were 8.4 million 
omissions and 4.4 million erroneous inclusions. The net undercount was 
4.0 million and gross coverage error was 12.8 million. In 2000, the 
Bureau indicates that the net undercount was 3.3 million and the gross 
coverage error was between 9.5 and 11.9 million people.
    The national net undercount is not a good indicator of overall 
census quality. If omissions and erroneous inclusions occurred in the 
same places, many of them would offset each other. Because the 
geographic distributions of omissions and erroneous inclusions differ, 
they offset each other only partially. The gross coverage error is a 
better indicator of census quality.
    Omissions tend to be concentrated among poor, typically minority, 
populations. They occur at especially high rates among the urban poor. 
Erroneous inclusions, while present among the poor, frequently occur 
among the affluent. In 1990, there were net overcounts among Whites 
living in owner-occupied housing units in the Northeast. In 2000, there 
were net overcounts among White homeowners in several post-strata 
located in the Northeast and Midwest.
    Census taking is controversial due to the perception that some 
groups are disadvantaged because of counting errors. The best statistic 
for evaluating this possibility is the differential undercount. This is 
the difference in the net undercount between White and minority 
populations (see Table 2). In 1990, the net undercount for Hispanics 
was 5.0 percent, for non-Hispanic Whites it was 0.7 percent, so the 
differential undercount between Hispanics and non-Hispanic Whites was 
4.3 percentage points. The net undercount for non-Hispanic Blacks was 
4.6 percent and the differential between non-Hispanic Blacks and Whites 
was 3.9 percentage points.
    The results of the 2000 Census show considerable improvement in 
this statistic. The net undercount for Hispanics was 2.8 percent, for 
non-Hispanic Blacks it was 2.2 percent, and for non-Hispanic Whites it 
was 0.7 percent. The differential between Hispanics and non-Hispanic 
Whites was 2.1 percent and between non-Hispanic Blacks and Whites it 
was 1.5 percent. These differentials for the 2000 Census are less than 
half of the comparable numbers for the 1990 Census. This is real 
progress, and I congratulate the Census Bureau.
    However, from the data I have seen to date, it is clear to me that 
statistical correction of the census would improve the accuracy. There 
were a considerable number of errors in this census--at least 9.5 
million by the Bureau's own account. The Bureau reduced the numbers of 
omissions more than the numbers of erroneous enumerations, and 
therefore reduced the net undercount. Although it reduced the 
differential undercount as well, important racial differences 
remain\2\, and these could be removed by statistical correction.
---------------------------------------------------------------------------
    \2\ These differentials are found whether we rely upon the A.C.E. 
or demographic analysis.
---------------------------------------------------------------------------
                   3. the status of the controversies
    The Supreme Court has settled the question of whether sampling 
should be used for Non-response Follow-up. We do not need to return to 
it now. The Secretary of Commerce, following the recommendation of the 
Census Bureau, has declared the unadjusted count to be the official 
result of the 2000 Census. We similarly do not need to return to this 
question. The remaining controversy concerns the release of adjusted 
block counts.
    I believe that these should be released. There are three main 
reasons for this. One is that while the Census Bureau did reduce all of 
the net undercount, gross coverage error, and differential undercount 
rates--important differentials remain. The American Indian undercount, 
while less than it was 1990, is at 5 percent. Both the American public 
and the scientific community need to understand the effects of the 
undercounts on the census results they rely upon.
    The second reason is that there are important ways in which the 
published census results appear to be incorrect. Ironically, the very 
improvement in census coverage that the Bureau accomplished creates 
error. There is confusion between the amount of actual growth and the 
amount of coverage improvements. Many localities, especially those with 
large minority populations, may have inflated impressions of the 1990-
2000 growth rates. The best way to fix this problem is to compare 
adjusted 1990 to adjusted 2000 census estimates. The use of adjusted 
data will reduce the inconsistencies created by improved coverage in 
the 2000 Census.
    The third reason is that both the American public and the 
scientific community need to have access to the adjusted counts to make 
their own evaluations on the nature of census error. There will be some 
purposes for which statisticians and other data analysts may deem it 
best to use the unadjusted counts, and other purposes for which they 
will prefer the adjusted results. The Bureau needs to release the 
adjusted data, along with their evaluations of these data, to permit 
these considerations.
    We also need to have a better understanding of the geographic 
distributions of gross error. Study of the gross error will help us to 
learn where there were important problems of census taking, and where 
improvements might have taken place in the 2000 compared to the 1990 
Census. For example, did the addition of addresses through LUCA reduce 
omissions? Finally, the scientific community needs to make its own 
evaluation of the bureau's estimates of the gross error rates.
                     4. state and local undercounts
    To illustrate the kinds of calculations and evaluations scientists 
need to make, I have calculated undercount estimates for each of the 50 
states and the District of Columbia. I have also done this for five 
large cities, Atlanta, Chicago, New York City, Philadelphia, and 
Houston, and the remainders of the states in which they are located. I 
relied upon the A.C.E. data provided by the Census Bureau to the Census 
Monitoring Board. Because I do not have access to the P.L.94-171 census 
count data, I cannot match the estimates of these quantities that the 
Census Bureau would calculate. I should be close, however.
    The state estimates vary across a narrow range. All but 10 states 
have estimates within one-half percentage point of the national 
average, 1.2 percent, i.e., and they are in the range of 0.7 to 1.7 
percent. States with larger shares of minority population tend to have 
higher rates of undercount, and the opposite occurs in states with 
smaller minority shares. Holding the minority shares constant, rates of 
undercount were higher in the West and lower in the Midwest. The main 
result, though, is that between-state variations are moderate.
    Turning to the cities, I have calculated rates of undercount for 
five of them. Although each of the cities includes large minority 
populations, the rates of undercount are lower than in 1990. Each of 
these cities has a higher rate of undercount than the remainder of the 
state in which it is located.
    These calculations are possible with the limited amounts of data 
that the Census Bureau provided to the Census Monitoring Board. To 
calculate comparable estimates for smaller areas, and to calculate the 
state and large city estimates with greater certainty, we need to have 
the adjusted block level data.
                   5. estimates of population growth
    With the reduction in the net undercount, especially of minority 
populations, we have difficulty interpreting rates of local population 
growth. Use of the currently unadjusted data leads to comparison 
problems. Because the level of undercount, especially in minority 
areas, was so much greater in 1990 than it is in 2000, use of the 
unadjusted results of the two censuses leads us to overestimate 
population growth. The problem is well illustrated by the case of New 
York City. Since 1990, Census Bureau population estimates have 
indicated small increases, with a total growth of 1.4 percent occurring 
between 1990 and 1999. Extrapolating to 2000, the expected population 
count was 7,452,184, an increase of 1.8 percent since 1990.
    The actual 2000 Census count was 8,008,278, an increase of 9.4 
percent since 1990. This amount is nearly 7 percentage points greater 
than the expectation. The result is implausible. Local experts believe 
that New York City has grown, but probably not by this amount. We 
cannot tell how much of the change is due to improved census coverage, 
and how much is real growth. In addition, because New York City added 
several hundred thousand addresses during the LUCA, this operation 
alone could account for much of the apparent growth. LUCA was not part 
of the 1990 Census.
    The New York City story was repeated consistently across the 
nation. I have calculated 2000 population projections simply by 
extrapolating the 1998--99 population change forward to 2000. In 
Philadelphia, for example, the Bureau estimated a loss of 17,367 people 
between 1998 and 1999. I simply assumed a comparable loss between 1999 
and 2000. The resulting projection of 1,400,234 is 7.7 percentage 
points below the 2000 Census count of 1,517,550.
    I have made comparable calculations for all counties that had at 
least a 500,000 population in 1990 for which the Census Bureau has 
published 2000 Census tabulations of P.L.94--171 data (see Table 3). 
There are 66 of these counties and 60 of them have counts that are 
higher than the population projections than I calculated. Only 6 have 
counts that are below the projections. This skewed pattern of 
difference reflects the fact that the 2000 Census counts are more 
complete than those of the 1990 Census.
    I have divided the counties into four categories, depending on the 
percentage minority in the 2000 Census. Among those counties where this 
percentage was less than 20, the average difference between the 
projected estimates and actual counts was moderate, 1.72 percent. 
Fourteen of the 16 population estimates were too low.
    Turning to the second category, where the percentage minority was 
between 20 and 29.9 percent, there were 22 counties, and 19 of them had 
estimates that were too low. The average difference between the 
population projection and the actual count was 2.35 percent.
    The third category included 16 counties where the percentage 
minority was between 30 and 49.9 percent. Fifteen of them had 
projections that were too low. The average difference between the 
population projection and the actual count was 3.78 percent.
    The fourth and final category included 12 counties where the 
percentage minority was greater than 50 percent. All of these counties 
had projections that were too low, and the average difference between 
the projection and the actual count was 6.66 percent. These counties 
were spread across the country, with 5 of them located in the 
Northeast, 6 of them in the South, and 1 in the West.
    Overall, we see that the discrepancies between projections and 
counts are greater where the percentage minority is greater. To 
understand these confusing patterns of population growth and coverage 
improvement, we need to have both the adjusted and unadjusted block 
counts. With these in hand, we can create our own combinations of local 
jurisdictions to study the extent to which the apparent growth is real 
or the product of changed and improved methods of census taking.
                               6. summary
    The 2000 Census is now complete, and its surprising results require 
analysis. The Census Bureau must complete its evaluation of the 
demographic estimate, and assess its consistency with the A.C.E. 
results. Demographers and statisticians across the country recognize 
this as a crucial project, and they would like to have their own looks 
at the data. In addition to understanding the predictors of undercount 
in the 2000 Census, the changing patterns of undercount between the 
1990 and 2000 Censuses, we need to figure out why the demographic and 
A.C.E. results are so discrepant.
    We also need to understand the patterns of gross and net error. 
Although I have not been able to carry out intensive research on the 
subject, much of the improvement in census coverage appears to be due 
to improvements in the census address list. The Local Update to Census 
Addresses program was a big contributor to this. If this were true, we 
would expect a substantial reduction in the number of ``whole household 
omissions'' in the 2000 Census relative to 1990. If the address list is 
improved, then the number of entire households missed should go down.
    On the other hand, many omissions occur in households where others 
are counted. These ``within household omissions'' occur when people 
filling out census forms misunderstand the instructions, and such 
omissions are especially prevalent among persons distantly or unrelated 
to the heads of household. Because such errors are generally unrelated 
to the completeness of address lists, or even to the proficiency of 
advertising and other outreach programs, we would expect the numbers of 
such omissions to be similar to those observed in 1990.
    Finally, we need good estimates of the numbers of omissions and 
erroneous inclusions. The Census Bureau has estimated that there were 
over 20 million non-matches, but only 6.4 to 7.6 true omissions in the 
A.C.E. The Bureau also estimates that only 3.1 to 4.3 million of the 12 
million apparent erroneous inclusions are real. Many of the apparent 
errors are due to missing data and other similar problems, and the 
numbers of apparent errors overstate the problems. However, the Census 
Bureau has not yet shared with us the logic behind these estimates.
    There is a great deal of research that needs to be done to 
understand the 2000 Census. Some of it must necessarily be carried out 
at the Census Bureau, but the bureau must make public all data relevant 
to this understanding so that researchers and academicians can draw 
their own conclusions.

           Table 1.--Estimated Error Counts from 1990 and 2000
                              [In Millions]
------------------------------------------------------------------------
                                                          1990     2000
------------------------------------------------------------------------
Estimated total population............................    252.7    284.7
Census count..........................................    248.7    281.4
Measured net undercount...............................      4.0      3.3
Measured Gross Erroneous Enumerations.................      4.4      3.1
Implied gross omissions...............................      8.4      6.4
``Gross coverage error''..............................     12.8      9.5
Assumed Errors in Reinstated "Potential Duplicates"...  .......      1.2
Total assumed gross erroneous enumerations............  .......      4.3
Implied gross omissions...............................  .......      7.6
Implied ``Gross coverage error''......................  .......     11.9
------------------------------------------------------------------------
Source: Obtained from the Census Bureau on March 21, 2001.


                        Table 2.--Reduction in the Differential Undercount, 1990 to 2000
----------------------------------------------------------------------------------------------------------------
                                                                Net Undercount      Difference from
                                                                     Rate            non-Hispanic       Ratio,
                           Group\1\                          --------------------       Whites          2000 to
                                                                                 --------------------    1990
                                                               (1990)     (200)    (1990)    (2000)   Difference
----------------------------------------------------------------------------------------------------------------
Hispanics...................................................       5.0       2.9       4.3       2.1      48.8%
Non-Hispanic Blacks.........................................       4.6       2.2       3.9       1.5      38.5%
Non-Hispanic Whites.........................................       0.7       0.7  ........  ........  ..........
----------------------------------------------------------------------------------------------------------------
\1\ Asians, Pacific Islanders, American Indians and Alaska Natives were not included in this table as these
  racial categories were treated differently in 1990 and 2000.
 
Source: United States Commerce Release Wednesday, February 14, 2001.

[GRAPHIC] [TIFF OMITTED] 88017.002

[GRAPHIC] [TIFF OMITTED] 88017.003

    Senator Kerry. Thank you very much.
    Dr. Wachter.

          STATEMENT OF KENNETH W. WACHTER, PROFESSOR, 
                CHAIR, DEPARTMENT OF DEMOGRAPHY,

                        UNIVERSITY OF CALIFORNIA

    Dr. Wachter. I am joined in this testimony by my colleague, 
David Freedman. We think there is widespread agreement on two 
chief points: First, Census 2000 succeeded in reducing 
differential undercounts from the 1990 levels. Second, there 
were serious questions about the accuracy of proposed 
statistical adjustments. The bureau advised the Secretary to 
certify the unadjusted counts, and we concurred, as did the 
Secretary's other outside advisors.
    As you have heard, Demographic Analysis and A.C.E. point in 
opposite directions. Demographic Analysis is precious, because 
it is independent of the census and A.C.E., tested over time, 
and rests on fewer and simpler assumptions than A.C.E. While DA 
is hardly perfect, analysis already shows that it would be a 
stretch to blame Demographic Analysis for the whole or most of 
the discrepancy with A.C.E.
    Mistakes in statistical adjustments in the census are 
nothing new. Our studies, David Freedman's and my own, on 1980 
and 1990 data, have described three kinds of error, processing 
error, correlation bias, and heterogeneity. In the face of 
these errors, it is hard for adjustments to improve on the 
accuracy of census numbers for states, for counties, 
legislative districts, and smaller areas.
    Statistical adjustments can easily put in more error than 
they take out, because the census is already very accurate. The 
changes, as Gene Ericksen mentioned, to state counts from 
adjustment are very small, measured in dozens of parts per 
million when you look at those changes, so the chief question 
is, what went wrong with A.C.E. in 2000?
    Errors in responses to the survey, or in the statistical 
operations, may from some perspectives have been under better 
control than in 1990, but it appears that processing errors 
must have been worse in other respects. Research is underway to 
pinpoint the difficulties.
    The bureau is investigating a form of error called 
balancing error. We suspect that troubles also occurred in a 
new treatment of movers, and in detection of duplicates, which 
were especially numerous in 2000, and which occur for minority 
populations as well as for nonminorities.
    Correlation bias and heterogeneity are endemic problems 
that make it extremely difficult for adjustment to improve on 
the census. Correlation bias is the tendency for people missed 
in the census to be missed by A.C.E. as well. Correlation bias 
in 2000 seems to have amounted, as it did in 1990, to millions 
of persons. These people cannot be evenly distributed across 
the country, so statistical adjustments create a distorted 
picture of census undercounts.
    Heterogeneity means that undercount rates differ from place 
to place within population groups treated as homogenous by 
adjustment. Heterogeneity puts limits on the accuracy of 
adjustments for areas like states, counties, or legislative 
districts. Our studies, along with recent work at the bureau, 
show that heterogeneity remains a serious concern.
    We must keep in mind that the census is used to distribute 
representation and resources directly among places, not ethnic 
groups. For A.C.E. to find an undercount of Hispanics or 
African Americans or Asian Americans, does not help if A.C.E. 
puts the missing people in the wrong places, and that is what 
A.C.E. is likely to do.
    Census 2000 achieved a high level of accuracy. Given that, 
and given the problems with statistical adjustments, the 
Secretary s decision to certify the census counts was the right 
decision.
    Thank you.
    [The prepared statement of Dr. Wachter follows:]
Prepared Statement of Kenneth W. Wachter, Professor, Chair, Department 
                of Demography, University of California
    I am joined in this testimony by my colleague David Freedman. We 
think there is widespread--although by no means universal--agreement on 
two chief points. First, Census 2000 succeeded in reducing differential 
undercounts from their 1990 levels. Second, there are serious questions 
about the accuracy of proposed statistical adjustments. The bureau 
advised the Secretary to certify the unadjusted counts and we 
concurred, as did the Secretary's other advisors.
    Statistical adjustment faced a new problem in Census 2000. 
Independent population estimates are derived by Demographic Analysis 
from administrative records, including birth and death certificates and 
Medicare files.\1\ These estimates show the Census overcounted the 
population by perhaps 2 million people. Proposed statistical 
adjustments would have added another 3 million people, making the 
overcounts even worse. Demographic Analysis is independent of the 
Census and the A.C.E. survey which underlies proposed adjustments. 
Demographic Analysis and A.C.E. point in opposite directions. While 
Demographic Analysis is hardly perfect, it is a stretch to blame 
Demographic Analysis for the whole of the discrepancy with A.C.E.\2\
---------------------------------------------------------------------------
    \1\ The bureau's estimates from Demographic Analysis are presented 
in the B-4 Report; see especially Appendix Table 2. The estimated total 
national populations are as follows: Demographic Analysis: 279.598 
million; Census 2000: 281.422 million; A.C.E.: 284.684 million.
    \2\ The Demographic Analysis estimates for net undocumented 
immigrants and some other categories of non-citizen residents may be 
somewhat low. But, as the B-4 Report spells out on page 11, implausibly 
large revisions would be required to bring the totals into agreement 
with A.C.E.
---------------------------------------------------------------------------
    Mistakes in statistical adjustments to the Census are nothing new. 
Our studies of the 1980 and 1990 data have described three kinds of 
error: processing error, correlation bias, and heterogeneity.\3\ In the 
face of these errors, it is hard for adjustments to improve on the 
accuracy of Census numbers for states, counties, legislative districts, 
and smaller areas. Statistical adjustments could easily put in more 
error than they take out, because the Census is already very accurate.
---------------------------------------------------------------------------
    \3\ See, e.g., Freedman and Navidi (1992), Freedman and Wachter 
(1994), and Brown et al. (1999).
---------------------------------------------------------------------------
    What went wrong with A.C.E. in 2000? Errors in responses to the 
survey or in the statistical operations may from some perspectives have 
been under better control than they were in 1990. But, it appears, 
processing errors must have been worse in other respects. Research is 
underway to pinpoint the difficulty. The Bureau is investigating a form 
of error called balancing error.\4\ We suspect that troubles also 
occurred in a new treatment of movers and in the detection of 
duplicates, which were especially numerous in 2000.
---------------------------------------------------------------------------
    \4\ The Bureau's research on balancing error is summarized on pages 
24-25 of the ESCAP report.
---------------------------------------------------------------------------
    In July 1991, the Bureau recommended adjusting Census 1990 by 
adding 5.3 million people: processing errors were estimated at 1.7 
million (these figures are net, nationwide). The figure for processing 
error later increased to 3.0 million, although independent estimates 
range up to 4.2 million. Our estimate is 3.6 million (Wachter and 
Freedman, 2000a).
    In 2000, A.C.E. would add 3.3 million persons to the census count. 
The preliminary estimate of processing error is 2 million, as may be 
seen by doing some arithmetic on the percentages in Table 24 of the B-
19 Report. That table allows 1 million for correlation bias. However, 
the underlying model (B-12) repeats the error discussed in Wachter and 
Freedman (2000a). Compare page 16 of B-12, although page 46 of the B-1 
Report acknowledges the problem.
    Thus, the estimate for correlation bias needs to be increased, 
perhaps by another million or so. If the population estimates from 
Demographic Analysis are approximately correct, the estimates for 
processing error will need to increase by several million, as in 1990.
    Correlation bias and heterogeneity are endemic problems that make 
it extremely difficult for adjustment to improve on the Census. 
Correlation bias is the tendency for people missed in the Census to be 
missed by A.C.E. as well. Correlation bias in 2000 may have amounted, 
as it did in 1990, to millions of persons. These people cannot be 
evenly distributed across the country. If so, statistical adjustments 
create a distorted picture of census undercounts.\5\
---------------------------------------------------------------------------
    \5\ Brown et al. (1999),Wachter and Freedman (2000a).
---------------------------------------------------------------------------
    Heterogeneity means that undercount rates differ from place to 
place within population groups treated as homogeneous by adjustment. 
Heterogeneity puts limits on the accuracy of adjustments for areas like 
states, counties, or legislative districts. Our studies, along with 
recent work at the Bureau, show that heterogeneity remains a serious 
concern.\6\
---------------------------------------------------------------------------
    \6\ Adjustment assumes that coverage rates (i.e., rates of census 
undercount or overcount) are constant within population groups called 
``post strata,'' across wide stretches of geography. Failures in this 
assumption are called heterogeneity, or called ``synthetic error'' in 
the B-studies. The Bureau's research on heterogeneity is summarized on 
pages 22-24 of the ESCAP report. Our work is described in Freedman 
andWachter (1994) as well asWachter and Freedman (2000b).
---------------------------------------------------------------------------
    Census 2000 achieved a high level of accuracy. Given that, and 
given the problems with statistical adjustments, the Secretary's 
decision to certify the census counts was the right decision.
                               References
    Anderson, M. and S. E. Fienberg (1999). Who Counts? The Politics of 
Census-Taking in Contemporary America. Russell Sage Foundation, New 
York.
    Breiman, L. (1994). The 1991 census adjustment: undercount or bad 
data? Statistical Science 9: 458-537 (with discussion).
    Brown, L. D., M. L. Eaton, D. A. Freedman, S. P. Klein, R. A. 
Olshen, K.W.Wachter, M. T.Wells, and D. Ylvisaker (1999). Statistical 
controversies in Census 2000. Jurimetrics 39: 347-75.
    Cohen, M. L., A. A. White, and K. F. Rust, eds. (1999). Measuring a 
Changing Nation: Modern Methods for the 2000 Census. National Academy 
Press, Washington, D. C.
    Fay, R. E. and J. H. Thompson (1993). The 1990 Post Enumeration 
Survey: statistical lessons, in hindsight. Proceedings, Bureau of the 
Census Annual Research Conference. Bureau of the Census, Washington, D. 
C.
    Freedman, D. A. and W. C. Navidi (1992). Should we have adjusted 
the U. S. census of 1980? Survey Methodology 18: 3-74 (with 
discussion).
    Freedman, D. A., P. B. Stark, and K.W.Wachter (2000). Aprobability 
model for census adjustment. http://www.stat.berkeley.edu/census/
557.pdf To appear in Mathematical Population Studies.
    Freedman, D.A. and K.W.Wachter (1994). Heterogeneity and census 
adjustment for the intercensal base. Statistical Science 9: 476-485 
(with discussion).
    Freedman, D. A., K. W. Wachter, D. R. Cutler, and S. P. Klein 
(1994). Adjusting the census of 1990: loss functions. Evaluation Review 
18: 243-280.
    Prewitt, K. (2000). Accuracy and Coverage Evaluation: statement on 
the feasibility of using statistical methods to improve the accuracy of 
Census 2000. Federal Register 65: 38373-38398.
    Skerry, P. (2000). Counting on the Census. Brookings, Washington, 
D. C.
    U.S. Census Bureau (2001). Report of the Executive Steering 
Committee for Accuracy and Coverage Evaluation Policy. Cited as the 
ESCAP report. The supporting documentation consist of reports B-1 
through B-19. All these reports are available on the web. http://
www.census.gov
    Wachter, K.W. and D. A. Freedman (2000a). The fifth cell. 
Evaluation Review 24: 191-211.
    Wachter, K.W. and D.A. Freedman (2000b). Measuring local 
heterogeneity with 1990 U. S. Census data. Demographic Research, an on-
line journal of the Max Planck Institute. Volume 3, Article 10. http://
www.demographic-research.org/Volumes/Vol3/10/

    Senator Kerry. Thank you very much, doctor. Doctor, in your 
testimony I do not think you addressed the question of the 
immigration issue, in the DA. I mean, it is my understanding DA 
does not measure immigration, it measures births and deaths.
    Dr. Wachter. That is right. The immigration component is an 
assumption that is built in.
    Senator Kerry. But isn't it critical? I mean, we have read 
article after article in the papers about increased numbers of 
immigrants, particularly in urban areas, and more even in rural 
areas than we previously thought, so would you not acknowledge 
that there is an impact on the DA by immigration?
    Dr. Wachter. On the Demographic Analysis?
    Senator Kerry. Yes.
    Dr. Wachter. Yes. In my complete testimony in footnote 2 we 
take up this question. We believe that Demographic Analysis 
estimates for net undocumented immigrants and some other 
categories of noncitizen residents may be somewhat low. But 
there are constraints from data on the percent foreign-born 
from the Current Population Survey and elsewhere, which put 
bounds on the plausible adjustments that you can make.
    We think you might be able to bring the Demographic 
Analysis estimates up closer to the census counts to make the 
apparent overcount in the census go away. But to make those 
estimates come up anywhere near the A.C.E., to suggest that 
there is not some set of major errors in A.C.E., seems really a 
great stretch.
    There is not enough scope in plausible estimates for 
undocumented immigration. The apparent overcounts of some 
groups of African Americans, younger African Americans, 
according to some constructions of the multiple race issue also 
put constraints on how much leverage you can get. So I do not 
know of anyone who believes the Demographic Analysis can be 
plausibly brought up to agree with A.C.E. It may be brought up 
to agree with the census. To bring it up to agree with A.C.E. 
is too much.
    Senator Kerry. If the Demographic Analysis had reported a 
higher number greater than A.C.E., would you have supported 
using corrected data?
    Dr. Wachter. No, I would not. The major uses of the census 
are in distributional matters. The important issue is in 
accuracy of distribution, and there are these endemic errors, 
including correlation bias, which argue that A.C.E. or suggest 
that A.C.E. would be putting people systematically in the wrong 
places.
    Representative Maloney has departed, but New York is a good 
example. By our estimates the population share of New York 
State would be brought down by adjustment, despite the strong 
presence of urban central city minority groups. And this would 
have happened in 1990 also. We believe that the culprit is 
likely correlation bias, that this adjustment in New York State 
would be a mistake. It would carry the population shares in the 
wrong direction, because of distributional errors in the A.C.E. 
survey.
    Senator Kerry. Dr. Ericksen, how do you respond to both 
answers with respect to the DA?
    Dr. Ericksen. Well, first of all I think that there is a 
very serious problem in the level of research that has been 
done on Demographic Analysis. There simply has been enough 
done, but we may be facing a situation where a good demographic 
estimate of the population simply is not possible, because 
after all, undocumented immigration is just that, and we can 
try to make assumptions after assumptions after assumptions to 
change the number.
    Senator Kerry. Is it a mistake to make assumptions because 
it is undocumented?
    Dr. Ericksen. It depends on the amount of ancillary data 
you have. Now, back in 1980 and 1990, demographers were able to 
make good estimates of the amount of undocumented immigration 
by looking at things like births and deaths in the areas where 
undocumented people were concentrated.
    I am told by demographers looking at this problem that it 
is more difficult to do now, and so I think there has to be a 
very large question mark to the use of Demographic Analysis for 
anything having to do with an immigrant population. I just do 
not think we know.
    As to the other issue, I think that I go back to my written 
testimony. In the state of New York you have both extremes. You 
have minority areas in New York City where the undercounts are 
very high. You also have non-Hispanic white areas outside of 
New York City where you have just the opposite situation.
    Now, how those two factors balance out is an interesting 
question. I do not know the answer to that question at this 
moment, but it is not implausible that they should be canceling 
out to bring New York State very close to, or just a little bit 
above, just a little bit below the national average.
    Senator Kerry. Now, you have heard a number of people today 
cite the difficulty and inadvisability of releasing the A.C.E. 
data in order to permit you and others to make some 
calculations, would you comment on that question of 
advisability and necessity?
    Dr. Ericksen. Well, I think there are two different points 
to make. First, in terms of general policy, we actually have 
had some debate in here today about how do we know the census 
is better. Well, we know the census is better because of the 
A.C.E., but we do not want to use the A.C.E. because we think 
it has errors in it. We go around and around on that.
    I think the way to get out of that vicious circle is to do 
an analysis. If the census is better in 2000 than it was in 
1990, why did that happen? What were the crucial decisions that 
people made?
    I think that I have identified two of those decisions. One 
decision was the use of the local update of the census address 
program. We need to evaluate that. The only way that we can 
evaluate that is to have local data to compare places where 
that program was successful, and where it was not put into 
place.
    Secondly, there was a decision to close data collection 
earlier and have a broader use of imputation. That also 
occurred on a local basis. The only way we could evaluate the 
result of that is on a local basis, so we need to have the 
block level data so we can understand the effects of that, and 
on the undercount, but we have to be able to aggregate the 
blocks in different ways.
    The other part of my answer is that there are a tremendous 
number of localities across the United States who do not 
understand the census data, and they cannot understand their 
census data because they do not know how much of the change is 
due to real change and how much of it is due to the fact that 
the net undercount of this census is smaller than the net 
undercount of the last one.
    Senator Kerry. What do you say to the notion that it is not 
ready, that it is premature, that they need to do more work 
with it?
    Dr. Ericksen. Well, my understanding is the block level 
data have already been calculated, and so they could be given 
out tomorrow.
    Senator Kerry. So you are saying that that is just not a 
legitimate answer?
    Dr. Ericksen. Well, I think that there might be a 
misunderstanding here, because there are two issues. One is, 
could the block level data be released? The answer--I 
understand they have all been calculated. They could be 
released. Two, are the block level data more accurate? My 
understanding there is, the Census Bureau is contending that 
the research on the question needs to be done, and we will see 
what their answer is.
    Senator Kerry. Is there any harm done by letting other 
people review that data now?
    Dr. Ericksen. I think every one understands that the final 
decision on accuracy will be made sometime in the future, so as 
long as people are cognizant of that, I do not see that there 
could be any harm.
    Senator Kerry. Do you have any way to make a judgment 
whether additional research will show that the A.C.E. is more 
accurate as a way of measuring?
    Dr. Ericksen. Well, I am very cognizant of my colleague to 
my left, Ken Wachter. I am very cognizant of the opinions of 
other colleagues who feel differently. I think it is an open 
question.
    Senator Kerry. Dr. Wachter.
    Dr. Wachter. It is an open question, and we need to 
research it. I would presume that one issue in the minds of the 
Census Bureau in terms of the early release of the block level 
adjusted data is the possibility that there is something like a 
computer coding error, which we had in 1990, in the A.C.E. 
estimates, and if there is, they would like to track it down 
before they put all these numbers out for the many uses that 
they would have. I do not have a strong view myself on the 
release or nonrelease.
    Senator Kerry. What is a fair amount of time within which 
one ought to be able to do that? When would it be fair for 
people to have an expectation that this data could be released 
so we could avoid this controversy?
    Dr. Wachter. I should say I am not opposing the immediate 
release. I think there are scientists who----
    Senator Kerry. Dr. Murray, you are opposing immediate 
release?
    Dr. Murray. Senator, I want to be very careful here. For 
research purposes there are great uses and values to having 
A.C.E. data available. You have to ask the question, for which 
purpose, and it is important to recognize that there is some 
sense in which the A.C.E. data are really not finished, 
finished in the sense of, they are attached, each of them, each 
of the numbers that could be provided has error bars attached 
to it, has a standard error that is associated with it that is 
a fluctuation.
    We have been, I think, provided a couple of 
misapprehensions in this discussion about the standing of the 
A.C.E. data, and if I might, just for a moment, explain, it has 
been treated as if it was only a matter of time, and if only we 
had more time, things would have been resolved.
    The actual ESCAP report, the Executive Steering Committee 
on A.C.E. Policy, recommended to the bureau that they not use 
the adjusted A.C.E. data, and the bureau Director, Mr. Barron, 
also said very clearly we did not run out of time, we ran out 
of data. What we have is an irreconcilability, and we may in 
the future be able to resolve that.
    Senator Kerry. I am puzzled by that.
    Dr. Murray. Let me find a cite for you that says that.
    Senator Kerry. I thought they specifically said that they 
needed to more fully investigate. They just needed more time.
    Dr. Murray. That is part of the issues, Senator. I 
appreciate you reading it directly. Let me point to errors in 
the way we have had this entered into the testimony. We have 
heard so far today that the committee believes it likely that 
further research may establish that the adjustment based on the 
A.C.E. would result in improved accuracy. However, the very 
next sentence, which was not quoted, says, however, the 
uncertainty due to these concerns is too large at this time to 
allow for recommendation on adjustment.
    Now, let us take that line over into the last page of the 
A.C.E. statement where they say, the ESCAP is unable to 
conclude at this time that the adjusted data are superior 
because--and here is the critical concern--further research on 
these concerns could, in fact, reverse the finding of the 
adjusted data's superior accuracy. It is indeterminate.
    Senator Kerry. I understand that. Nobody questions that, 
but the question is, why can't other people be engaged in that 
similar research so that there is a sort of countercheck?
    Dr. Murray. What our concern, Senator, is that--I think the 
concern here is that we cannot presume that the adjusted data 
are somehow superior, or will be found to be superior.
    Senator Kerry. I am not making that presumption.
    Dr. Murray. What we might find is, the data are actually 
less reliable than we expected, and here, as Dr. Wachter has 
said very clearly, redistribution, the issue of distributive 
accuracy, of getting the data correct at the levels of the 
local governments where they are actually applied, is the 
census' primary responsibility, and the issue with the A.C.E., 
the concern we have is that the A.C.E. does not put back the 
undercounted where they were actually lost. The A.C.E. is 
indeterminate.
    Senator Kerry. We have a confidence question here about 
whether or not the undercount will be accurately reflected, put 
back by whatever methodology is arrived at, and therefore the 
question is, when is the appropriate time for that data to be 
judged by people outside who do not have the same interests, or 
sort of historical investment, if you will, in the process, who 
will look at it independently and help give confidence in the 
numbers.
    I would assume--it is a question I would ask of both you 
and Dr. Wachter, that if you have 6.4 million considered to be 
undercounted, do you consider that very accurate, leaving out 
6.4 million people?
    Dr. Murray. Mercifully, Senator, it is an improvement.
    Senator Kerry. That is not what I asked you.
    Dr. Murray. It is a gross error, and it is troubling to me, 
and let me explain. It was my privilege, Senator, to live 9 
years in Massachusetts, and you represented me.
    Senator Kerry. Well, being represented by me and living 
there means you can say whatever you want, and I will forgive 
you.
    [Laughter.]
    Dr. Murray. Thank you, sir. I learned something very 
directly about you, which was your own passion and commitment 
to the problems and needs of the undercounted communities, the 
underserved and underprivileged communities. Senator, let me 
assure you, I share that, and have tried in every way possible 
to address the problem that we face, the social problem of the 
undercounted people. The heart of the question is, what is the 
best mechanism, the best instrument that will help us reach and 
encourage and incorporate those people and give them a fair 
share and bring them into the system?
    Senator Kerry. I completely agree.
    Dr. Murray. One of the difficulties is that the A.C.E.--oh, 
would that it did work according to its promise, but it 
represents the triumph of hope over experience to think that 
A.C.E. actually solves the problem. It has the tendency to 
introduce more error, to actually add people back into already 
overcounted communities and worsen the problem, and to fail to 
put people back in the undercount. It does not seem to solve 
our issue.
    Senator Kerry. You are eloquently and forcefully making a 
case, but you are simultaneously, it seem to me, making a case 
for having this data analyzed by more people in order to come 
up with a fair methodology.
    Dr. Murray. There is no principled or inherent reason not 
to evaluate the data.
    Senator Kerry. My question is, when can we anticipate 
fairly that whatever adjustments need to be made, or judgments 
about the data, so that it could be released in order to permit 
that review before we all wind up with a decision that has been 
made and there is no recourse? Could I anticipate--it is now 
the end of March, or April. Could we have it by May?
    Dr. Murray. In my discussions with bureau professionals who 
are doing this task, they anticipate the need for more data 
from the long form that will help them understand the foreign-
born population, for readdressing this problem about whether or 
not Demographic Analysis is accurately capturing the 
undocumented, and that they have spoken about a target of 
potentially this fall having a better grasp of the problem one 
way or the other, and it seems to me to be reasonable.
    Senator Kerry. This fall is too late, which is why we have 
engaged the statistical sampling.
    Mr. Vargas, what would that mean to you?
    Mr. Vargas. Mr. Chairman, let me first thank you for your 
leadership on this issue.
    It would have a tremendous impact on the Latino population. 
Let me just say that as a representative of the 33 million 
Latinos out there who have been counted in the census, that we 
are speaking for the 1 million that were missed, and these are 
1 million who would not have fair representation in 
congressional districts and in state legislative districts. It 
would result in malapportioned districts, districts in the 
Central City of Los Angeles, for example, that would be 
inherently larger than other districts in other parts of the 
state.
    Let me also bring to your attention, Mr. Chairman, another 
issue. I did not mention in my oral testimony that I am a 
member of the Secretary of Commerce's Advisory Committee on the 
Decennial Census.
    With respect to this subsequent evaluation that the Census 
Bureau professionals will be doing with respect to the A.C.E., 
I asked that when they make that recommendation to the Acting 
Director, who will make the decision as to whether or not to 
further release adjusted data? The Acting Director of the 
Census Bureau? Mr. Lee Price, who is the Acting Under Secretary 
of Commerce, plainly said that they did not know whose 
authority it will be to decide whether or not in the fall to 
release adjusted data.
    They indicated that Secretary Evans' decision to rescind 
the rule that his predecessor put in place, Secretary Mineta, 
to allow the Department of Commerce--I'm sorry, the Census 
Bureau Director to make the decision on the release of the data 
only pertained to data for redistricting purposes, that if the 
professionals at the ESCAP committee recommend to release 
adjusted data, that they did not know whose authority it would 
be to approve that release. They did not know if it would be 
Acting Director Barron's decision or the Secretary's decision.
    I think this is an issue that certainly this Committee 
should investigate, as to who ultimately has the authority to 
decide on whether or not----
    Senator Kerry. If I had had more time to be here with the 
Secretary, that is one of the questions I will ask, but we will 
submit that question to him to try to determine that as a part 
of the record of this hearing.
    Before we wrap this up, Dr. Ericksen, Dr. Wachter, what is 
your sense of what the timing might be able to be that would be 
able to be effective here so we could try to not wind up--I 
mean, look, my goal here is not to--I do not want an inaccurate 
count, but I do not want to live with the estimate for 10 years 
that 6.4 million of our citizens were left out.
    President Bush has said he wants to leave no child behind. 
If you do not count them, you do not have a prayer of not 
leaving them behind.
    Now, it seems to me there is just a fundamental judgment to 
make here. If everybody has agreed there is an undercount, and 
I have heard nobody suggest otherwise, there has to be a way 
through smart application. I mean, you can improve the census 
any number of ways, but statistical sampling has been the most 
accepted scientific method of trying to do it, so you are not 
just playing Solomon and trying to cut it down the middle.
    Now, what could we anticipate, fairly speaking, so that the 
transparency of this is sufficient that people will not feel 
that a Republican administration and a Republican-run Commerce 
Department is trying to do what people saw a Republican 
administration do previously in the Commerce Department when 
they denied the use of statistical sampling? Everybody knows 
what the impact of that is.
    Now, I want to try to get away from the political 
gamesmanship here. I do not want this to be political. I am not 
trying to have a Democratic outcome, I am not trying to have a 
Republican outcome. I would like an American outcome, which is 
to count American citizens as fairly as we know how.
    Now, is there some fair way for both of you, as sort of 
representing--I mean, one is a Republican Member, one is a 
Democratic appointee, but is there a way to come at this and 
have some agreement about when the statistics could be subject 
to scrutiny at large, so that people could make some judgments 
about the numbers that the American people have paid for to be 
gathered by an American agency? What do you think?
    Dr. Wachter. I do not think either Gene nor I really comes 
from a partisan political viewpoint on this, and we have often 
worked together in the Special Advisory Panel of 1990.
    Senator Kerry. Do you know when that data might be able to 
be subject to scrutiny?
    Dr. Wachter. For the 1990 census, by the beginning of the 
summer of 1991 most of the relevant data that we worked with 
was there in good form. Isn't that right, Gene?
    There are apparently certain kinds of larger problems with 
A.C.E. than there were with the Post Enumeration Survey in 
1990, so I have heard people talk about the summer. But as I 
said, from the point of view of scientists like ourselves 
working with the data, I would have no opposition to seeing the 
immediate release.
    Senator Kerry. Dr. Ericksen.
    Dr. Ericksen. I think it is a basic principle of science 
that we make our data publicly available. Just to give you an 
example, the Journal of Public Opinion Quarterly has a policy 
that we not publish an article unless the author of the article 
makes a questionnaire available.
    In my view, the data should have been made publicly 
available the day the monitoring board got them back in March. 
There is only going to be improvement for having more people 
enter this debate. The more I analyze the data, the more I see 
questions to ask the Census Bureau. For example, in every state 
that I have looked at that has sizable numbers of blacks and 
Hispanics, the net undercount there of those groups appear to 
be greater than they are of non-Hispanic whites.
    That is a very important question to know, because we want 
to know about consistency, so the data in my view should have 
been released in March or February, whenever they became 
available. There is no harm whatsoever in engaging in the 
scrutiny. I also think that people in city planning offices 
across the country would be very interested in looking at the 
effects of a possible adjustment on their localities.
    That is a different issue than making the decision. I think 
the mistake is to think nobody can look at the data until the 
decision has been made, and we are arguing the opposite.
    Senator Kerry. Well, I am arguing the opposite very 
strongly. It seems completely inconsistent to me to be making 
the statements that this is the most accurate census we have 
ever had, and to base your judgment that it is the most 
accurate census you have ever had because it is supported by 
the A.C.E., but you are not wiling to let the information on 
which you make that judgment be scrutinized.
    So I am going to press very hard with colleagues and others 
to see if we can open this debate up. I mean, this is an 
important debate, and I have never known legitimate 
information, statistics to be available to people. I know they 
can distort them, and I know they can twist them, but there are 
a lot of folks out there who can analyze this and begin to make 
some judgment about accuracy and about where we go.
    That is the nature of the American process, is to allow 
more information, not less, particularly information that 
number 1, that the American people paid for and pulled 
together, and number 2, that the American people are being 
asked to live with for 10 years that has a profound impact on 
their communities and their planning, and lots of other public 
officials who would like to, I think, have independent analyses 
of what happened in their community. So I think it is very, 
very important to try to get that information available, and to 
continue to have this debate.
    I assume if we do not attempt t resolve this issue now, it 
will be with us for 10 years, with a lot of acrimony and 
divisiveness that is really unnecessary that should not 
accompany something like this effort to simply count the people 
who are here living in this country.
    So with that said, unless there is an urgent need to add 
something, at this point I will leave the record open for 7 
days for any colleagues who may have additional questions, or 
for any additional testimony we want to gather in writing, and 
I thank you all very, very much for taking time and for 
participating today.
    Thank you. We stand adjourned.
    [Whereupon, at 11:55 a.m., the hearing was adjourned.]
                            A P P E N D I X

      Response to Written Questions Submitted by Hon. John McCain 
                           to Donald L. Evans
    Question 1. Mr. Secretary, your testimony highlights a large 
reduction in the undercount of minorities by the Year 2000 Census. 
However, some minority groups still complain that the Census does not 
adequately count them. Could you please describe what steps the Census 
Bureau took to reduce the minority undercount, and how the Census 
Bureau intends to improve upon its 2000 results?
    Answer. The Department of Commerce is gratified by recognition of 
the strides made by the Census Bureau in reducing the undercount of 
minorities and others in Census 2000. This was achieved by implementing 
many new and expanded procedures to improve participation in the 
census, including: expanded partnerships; paid promotion, allowing both 
a visible national campaign and promotion targeted at specific 
population groups; increased availability of census forms and ways to 
respond; involvement of local/tribal governments in updating the 
address listing to improve housing unit coverage; tailoring special 
enumeration procedures to population groups and geographic locations; 
enhanced language program; expanded recruitment of census workers.
    Further reductions in the undercount need to be addressed on 
several fronts. While the Census Bureau is proud of the results of 
Census 2000, it intends to improve upon them. This will be increasingly 
challenging because the population is becoming progressively more 
diverse.
    The Census Bureau plans to reengineer the entire census process, 
with improved coverage as a major objective. The reengineering will be 
based on three strategies: A significantly enhanced and improved Master 
Address File and the geographic data base supporting it; Long form data 
collected and tabulated every year by the American Community Survey, 
and therefore not included as part of the 2010 Census; Early planning 
and design of the 2010 Census to fully take advantage of opportunities 
made possible by the other two strategies.
    To achieve these goals we will need the support and endorsement of 
Congress, in addition to the necessary funding.
    Question 2. The next major decision that you will have to take 
regarding the Census is whether or not to release adjusted data for 
federal fund allocation. I understand that you intend to make this 
decision in the middle of October. What processes have you set up and 
criteria you will consider as you make this decision?
    Answer. The Census Bureau has prepared a plan for continuing the 
analysis of the Census 2000 data over the next several months, a 
process which will lead to a recommendation next fall concerning the 
use of adjusted data for use the inter-censal population estimates 
program and for sample controls in demographic surveys. These data may 
also be used to allocate federal funds under various grant programs.
    The plan contemplates that the Census Bureau will continue with its 
investigation into the quality of the Accuracy and Coverage Evaluation 
(A.C.E.), the demographic analysis, and Census 2000. That investigation 
will rely on new data collected from the long form, as well as more in-
depth reviews of the data and methodologies that supported the 
recommendation of the ESCAP Committee last March to use the unadjusted 
data for redistricting purposes.
                                 ______
                                 
     Response to Written Questions Submitted by Hon. Sam Brownback 
                           to Donald L. Evans
    Question 1. For the first time since 1880, the 2000 Census short 
form did not gather data on marital status. As a result, important 
information on family structure, which is a vital indicator of child 
well-being, was lost. Last year, the Senate voted unanimously in favor 
of a resolution expressing support for restoring the marital status 
question to the short form. Does your office plan to resume the 
gathering of marital status data on the short form of 2010?
    Answer. For Census 2000, in compliance with directions from 
Congress to reduce the number of questions asked on the decennial 
questionnaires, the Census Bureau undertook a comprehensive review of 
each question contained on both the short and long form questionnaires, 
eliminating those which were not specifically required by law. Through 
this process, the Census Bureau moved five questions from the short 
form to the long form because these questions produced data that were 
not needed at the block level. One of those questions was the item on 
marital status. The Census Bureau submitted its proposed questionnaire 
content for Census 2000 to Congress on March 31, 1998. Following this 
submission, the Census Bureau fully considered all expressed concerns 
about question topics before finalizing the questionnaire content. The 
Census Bureau received no comments regarding the content of the marital 
category, and thus left that question on the long form only. When the 
Senate's resolution on the marital status question was passed, it was 
too late to consider changes to the content of census questionnaires.
    The formal planning work for the 2010 census again will undertake a 
comparable large-scale effort to elicit information from Congress and 
the public on what data should be collected in the next decennial 
census. In addition, the American Community Survey (discussed in more 
detail in a later question), if funded, will provide important new and 
more relevant data on family structure in America.
    Question 2. It is my understanding that collection of some vital 
statistics data from the states can be spotty, and that the marital 
status question on the short form of the Census was used to flesh out 
those numbers. Some demographers have claimed that removing the 
question from the short form may impair the ability to make accurate 
state and local estimates of the married population. Is this a concern?
    Answer. The Census Bureau expects no data quality deficiencies 
resulting from having the marital status question on the long form. The 
Census Bureau received over 17 million long forms and believes that 
marital status information from the long form, which is provided down 
to the census tract level, will be sufficiently accurate to provide 
demographers concerned with state and local estimates with the data 
necessary for their statistical needs.
    Question 3. What specific steps are being taken to ensure that the 
American Community Survey (ACS) improves and enhances data collection 
on family structure in America?
    Answer. Several questions in the ACS measure family structure, 
including questions regarding marital status, relationship, and births 
in the past year. The ACS has many advantages over a single day 
snapshot taken every ten years because the ACS provides more detailed 
measures of change. In addition, the quality of data collection, 
coding, and processing will improve through the use of a permanent, 
well-trained staff and continuous data collection.
    Through the richness of the ACS content and the ability to create 
estimates for small geographic areas, America will he able to study a 
variety of family issues, such as matters related to children living in 
homes with a disabled veteran parent, children relying on the financial 
support of grandparents, characteristics of children living with 
immigrant parents who are not proficient in English, and other matters. 
The Census Bureau expects the ACS to produce quality data to answer 
these questions at all important levels, including the national, state, 
city and town levels.
                                 ______
                                 
      Response to Written Questions Submitted by Hon. John McCain 
              to Kenneth W. Wachter and David A. Freedman
    Question 1. You have just given testimony concerning possible 
causes of problems in the A.C.E. methodology. How would you design a 
method for adjusting the Census data to account for correlation bias, 
heterogeneity, and processing errors?
    Answer. Given the small level of error in the 2000 Census, as 
presently estimated, we do not believe that adjustments can be designed 
which would improve on the Census counts.
    Question 2. The ESCAP cited the discrepancy between the A.C.E. 
data, actual Census data, and Demographic Analysis as a major reason 
for releasing unadjusted data to the states. What processes would you 
recommend that the Census Bureau use to examine this discrepancy?
    Answer. The most urgent need is for the Bureau to tabulate the 
results from A.C.E., the Census, and the Demographic Analysis with a 
consistent set of categories of age, race, and sex. These tabulations 
will be helpful for deciding on the next stages of analysis.
    We recommend that the Bureau publish additional data for 
independent analysis by the scientific community, including (a) the 
targets used in their loss function analyses, and (b) the Census 
populations of Congressional districts by A.C.E. poststrata and the 
residual population.
    We thank the Chair and the Committee for their interest and 
attention.
                                 ______
                                 
      Response to Written Questions Submitted by Hon. John McCain 
                           to David W. Murray
    Question 1. Some observers argue that it is important to release 
data adjusted with the A.C.E. results to more accurately account for 
population growth and enhance our understanding of Census methodology. 
Given your concerns about errors in the A.C.E. methodology, what would 
be the effect of releasing A.C.E.-adjusted data?
    Answer. The A.C.E. process does indeed have the capacity to 
``enhance our understanding of Census methodology,'' and hence, 
studying the results of the A.C.E. in comparison to Census enumerated 
data can make an important contribution to our overall understanding of 
the country's population growth during the last decade. Similarly, 
studying the data captured in the Demographic Analysis could likewise 
contribute to our understanding of both the A.C.E. process and the 
Census enumeration. The comparison of the three measures of the 
population is important particularly because the respective portraits 
of the population that they offer can not be reconciled with each 
other, at this time and given currently available data. That is, there 
is an important research role for the use of A.C.E. data (which role 
does not necessarily require their ``release'' to the respective 
states).
    Moreover, not only are the three data sets internally 
irreconcilable, they place us in an historically unprecedented 
situation, in that the Census enumeration shows a larger population 
than does the Demographic Analysis, a reversal of the expected 
relationship. That is, from the perspective of the Demographic 
Analysis, the Census enumeration data show a net overcount of the 
population, and not an undercount. The A.C.E. data, at least at this 
preliminary juncture in the analysis, take us even further away from an 
ability to reconcile the Demographic Analysis with the Census 
enumeration, in that the A.C.E. data suggest a population count even 
higher than either of the other two measures. Determining which 
measurement is giving us the most accurate data is not self-evident, at 
this stage.
    These remarks characterizing the respective population portraits 
refer, however, only to the overall summary totals of the population, 
and do not address the even larger challenge of determining 
distributive accuracy (the proportionally correct allocation of the 
population accurately positioned in correct geographical areas), which 
is the essential function of the Census numbers. Given inherent 
uncertainties of the A.C.E. adjustment process, it seems very likely 
that the Census enumeration data, the unadjusted data gathered by a 
direct census, will prove to be the most correct and accurate data for 
distributive accuracy, a consideration that appeared to prevail in 
forming the recommendation by the ESCAP committee not to release 
adjusted data as the official redistricting data file.
    In addition, there appear to be misconceptions on the part of the 
public and some public officials about the quantitative meaning of 
A.C.E. adjusted numbers. While some people take the A.C.E. data as 
concrete, specific population counts, in reality all numbers derived 
from a sampling process are subject to probability fluctuations. That 
is, rather than deriving a single, stable adjusted count number (for a 
particular post-stratum), what a sampling adjusted process really gives 
you is the value drawn from within a spread of values that varies in 
its range for various post-strata. Importantly, each of those estimated 
numbers must be accompanied by a Standard Error that expresses the 
confidence intervals for the value in question (confidence intervals 
providing the likely upper and lower bounds within which the mean 
figure is believed to fall a high probability of the time).
    For instance, for the Major Group ``Native Hawaiian/Pacific 
Islander,'' we might hear that the A.C.E. data for Census 2000 show a 
net undercount of 4.6 percent, while the A.C.E. undercount figure for 
Asian Americans (non-Hispanic) is pegged at .96 percent undercount 
(source: Census Bureau B-1 document, table 2a, ``Percent Net Undercount 
for Major Groups: 2000 A.C.E.''). Some people improperly regard those 
numbers as specific and concrete ``findings'' from the A.C.E. But in 
reality, the A.C.E. process actually reports those undercount estimates 
along with a Standard Error (SE) estimate, a plus/minus spread within 
which the true mean is felt to lie. Hence, for Native Hawaiians/Pacific 
Islanders, the actual A.C.E. data show an undercount of 4.6 percent 
with an SE of 2.77 percent. That means that the actual value for that 
post-stratum's undercount could range from 1.83 percent to 7.37 
percent. Likewise, for Asian Americans, the .96 percent undercount has 
an SE of .64, producing a range of values going from .32 percent to 1.6 
percent. Which number represents the ``true'' undercount adjustment 
which must be applied to any local population?
    Notice further that in these two cases, the SE is larger than half 
the value of the estimate mean figure. In such circumstances, the 
probability fluctuation possible is proportionately very large indeed, 
rendering our portrait of the population of those post-strata 
relatively indeterminate within the overall range. With a strikingly 
accurate Census enumeration that materially reduced the undercount, 
including dramatic reductions in minority post-strata, we rapidly come 
to the point where adjustment begins to give us diminishing returns; 
that is, we begin to replace one relatively accurate enumerated figure 
(which albeit has an estimated undercount percentage) with a second 
figure, the A.C.E. adjusted figure, that is not better or more 
accurate, just different. Moreover, its margins of error are in many 
instances as large or larger than the undercount percentage that it 
seeks to remedy.
    Yet many observers continue to think that the adjusted A.C.E. 
number is somehow a straightforward percentage correction that can 
simply be added in to the respective post-strata populations. This is 
not so. Releasing adjusted data sets to the states could well abet this 
misapprehension, and even lead to ill-founded legal actions. Finally, 
the margins of error problem only increases as one approaches smaller 
and smaller geographical units, which it is the primary responsibility 
of the Census data to specify with accuracy for apportionment purposes.
    Question 2. Some observers have suggested that a ``sampling'' 
method, such as the A.C.E., would be helpful in accurately measuring 
previously undercounted areas, such as parts of Arizona. Could you 
please discuss why the A.C.E. might lead to an incorrect adjustment to 
an area that is largely rural and has a large minority population?
    Answer. Once again, we must keep foremost the importance of local 
level accuracy for the Census to accomplish its purposes of accurately 
apportioning political power and funding to actual people where they 
actually live. A serious difficulty encountered by the Census in rural 
areas, particularly those with a low density of a minority population 
such as found in Arizona, is the problem of inaccurate or incomplete 
address files for residences. Getting an accurate enumeration is a 
challenge in such circumstances, and, as the lessons of the Census 2000 
amply demonstrate, requires an active mobilization at the local level 
of officials, agencies, and field personnel who are familiar with the 
geographical area, who have indigenous command of the respective 
languages and cultures of the area, who are trusted and accepted by 
local people, and who have a stake in achieving an accurate count.
    It appears to be an unfortunate by-product of the sampling 
adjustment process that just such effective mobilization is not 
activated, and may actually be undermined by the process of statistical 
adjustment based on estimation assumptions for an area that is 
sometimes derisively referred to as ``fly-over country.'' Rather 
obviously, ``flying over'' an area and adjusting it with estimated data 
derived from nationally-distributed samples cannot provide the kind of 
local knowledge, and community engagement, that a direct enumeration 
can accomplish.
    The problem of an accurate adjustment is particularly acute when 
address files are incomplete or inaccurate, since the A.C.E. design 
depends upon a matching process, comparing the count from the 
enumeration with the count for the A.C.E. sample for each address 
covered by the sampling method. Errors in addresses lead to errors in 
matching, and these together lead to erroneous adjustments.
    As our September, 1999 Report to Congress (``Unkept Promise'') 
argued, the adjustment process has a tendency, particularly at local 
levels of geography, to mis-allocate the adjustment, adding people 
erroneously where they were not really missed and, sadly, failing to 
provide a commensurate correction for areas where striking undercounts 
actually occurred. Many undercounted communities remain undercounted 
after adjustment, the missing persons erroneously having been allocated 
by the adjustment process to other areas, or spread across wide 
geographical dispersions where they were not undercounted in the first 
instance. (Realizations such as these may well have motivated the 
language of the ESCAP committee's recommendation not to use adjusted 
data where they noted that, under any scenario yet devised, adjusted 
data were less accurate than enumerated data for counties of less than 
100,000 people.)
    Somewhat perversely, however, officials acquire the illusion that 
they have addressed their community's undercount by adjustment, when in 
fact they have not proportionally done so. Perversely, as well, it 
appears that the ``unkept promise'' of adjustment substitutes for the 
very community based actions that have a far better chance of actually 
locating the undercounted minorities, and even engaging them in civic 
participation by the encounters and community mobilization required for 
a direct Census enumeration.
    The difficulties of the address mis-matching are further amplified 
by the fact that the post-strata adjustment factor applied to many 
rural minority communities suffer from heterogeneity problems that the 
A.C.E. design could never sufficiently resolve. This means that 
sociological assumptions about a minority post-stratum being homogenous 
throughout the country (in their probability of being counted or missed 
in the census) were not born out in practice in actual communities.
    As I noted in my earlier written testimony submission, one should 
not apply the same adjustment factor to a recently-arrived Puerto Rican 
immigrant to New York City as one applies to an Hispanic ranch owner in 
Arizona whose family has been in the community for two centuries simply 
because both parties identify themselves as Hispanic; yet this is 
exactly what the A.C.E. designs tries to do, using a nationwide post-
stratum of Hispanics wherein Arizona state data are adjusted using data 
actually derived from adjacent or even distant states. As a further 
example, it would probably surprise many observers to learn that a 
post-stratum category that experienced one of the highest overcount 
rates was Native American male homeowners not on reservations. Simply 
put, the Native American community, ranging from the densely-clustered 
Hopi villages to the remotely distributed Navajo ``outfits'' to the 
remarkably wealthy and modernized Connecticut community of the casino-
owing Mashantucket Pequot, has enormous heterogeneity in its actual 
social and economic make up, a factor that introduces indeterminacy 
into the A.C.E. assumptions.
    It was considerations such as these that were reflected in the 
ESCAP committee's recommendation not to use adjusted data when they 
discussed remaining A.C.E. design difficulties termed ``synthetic 
error.'' A second set of concerns in the ESCAP report, referred to as 
``balancing error'' and addressing the problems of differential search 
areas in the Census and the A.C.E. for determining matches, represent 
an additional problem for rural areas of low density and household 
dispersion, as do the tendencies of some minority groups (such as 
Native Americans) to live in household groupings that do not coincide 
with traditional census assumptions about residence patterns, and the 
noted transience of workers in many rural minority communities.
    For these reasons (among many), the A.C.E. process cannot be shown 
to be more accurate for local communities than the Census enumeration 
data that they were to replace, and may even encourage a false hope 
that the undercount problem has been ``solved'' while simultaneously 
reducing the incentive to adopt Census enumeration strategies and 
practices that could prove more effective in ensuring that undercounted 
communities are identified, accurately enumerated, and accordingly, are 
provided with the political power and public funding that are their 
just expectation.
                                 ______
                                 
      Response to Written Questions Submitted by Hon. John McCain 
                            to Arturo Vargas
    Question 1. You have emphasized that over one million Latinos were 
missed by the Census Bureau. What suggestions would you have to improve 
the Census Bureau's enumeration process for counting Latinos?
    Answer. The first place to begin in determining how to improve the 
enumeration of Latinos for the next census would be to build upon those 
elements of Census 2000 which were particularly effective in this 
regard. From our extensive work in conducting community outreach and 
public education for the census, we believe the following efforts 
contributed to the relative success of Census 2000:
     The early waiving of the U.S. citizenship requirement for 
the hiring of enumerators helped ensure that adequate numbers of 
bilingual individuals were available to be hired by the Bureau in areas 
with large non-English speaking populations. However, there continued 
to be a shortage of bilingual enumerators in areas with small and 
emerging non-English speaking populations.
     The Census in the Schools helped to convey the importance 
of the census to families by using schoolchildren as key messengers.
     The paid advertising campaign in English and non-English 
languages helped raise awareness in the census considerably.
     The partnership program was particularly successful in 
involving as many community leaders and organizations as possible in 
being messengers of the importance of the census.
    One of the major obstacles to ensuring a full census count of 
Latinos, however, is a structural element of the enumeration process. 
As long as the Census Bureau continues to rely primarily on the mail-
out/mail-back method, Latinos and other populations with higher rates 
of poverty than the general population, non-English speaking 
populations, and groups of individuals who fear contact with the 
government, will be particularly vulnerable to being missed in the 
census. It is our understanding that the census enumeration method may 
be vastly different in 2010. Any changes to the enumeration process 
should be specifically designed to remove these inherent barriers to 
the counting of low-income populations, children, immigrants, and 
others at historical risk of being undercounted. We understand that a 
fundamental aspect of ``recreating'' the census is the introduction of 
the American Community Survey (ACS), which will be able to provide 
annual data on the U.S. population and eliminate the need for the long 
form in the decennial census. We believe that the ACS has particular 
promise in collecting data that will be critical in designing a more 
effective decennial census and targeting resources for outreach and 
enumeration in 2010.
    Based on our experience with Census 2000 and information we have 
received from other Latino organizations and leaders involved in census 
promotion and outreach, we make the following recommendations:
     Promoting the census must become an ongoing initiative, 
and not an activity that is left to being a decennial task.
     Congress should promote annual appropriations for the 
Census Bureau to carry out sustained partnerships between the Bureau 
and the community, including Latino organizations. The Census Bureau 
and the Department of Commerce have expressed a commitment to 
maintaining these partnerships throughout the decade, but it will 
depend on the willingness of Congress to support these partnerships 
through appropriations.
     The Congress should support and monitor the American 
Community Survey, an annual sample survey of the American population 
that is designed to collect the data now asked in the long form. It is 
the Census Bureau's plan to eliminate the need for the long form in the 
decennial census through implementation of the ACS. The ACS can also 
serve as an ongoing tool to reinforce the confidentiality of the census 
throughout the decade.
     The American Community Survey must be fully funded by 
Congress every year.
     Members of Congress should take a leadership role in 
promoting the census through the public schools by encouraging the 
school districts in their states and congressional districts to 
incorporate the census as a permanent element of the school curriculum. 
Census in the Schools was a particularly successful element of the 2000 
Census outreach plan and should happen every year.
     The Congress should support and promote supplemental 
outreach funding by cities and states in the decennial census. In 2000, 
jurisdictions such as California and Houston allocated their own 
resources to census outreach and promotion with considerable success in 
increasing their mail-back response rates.
     The Congress should advocate for funding for the Census 
Information Centers. Several Latino organizations have been designated 
as Census Information Centers, but are unable to raise the funds 
necessary to analyze, publish and distribute the data they receive. 
Congress should appropriate resources to make these Centers viable.
     The Congress should carefully evaluate all aspects of the 
Census 2000 and work with the Bureau and the community to improve those 
areas that fell short. Some key concerns in Census 2000 include the 
following issues:

           As in 1970, 1980 and 1990, it again was difficult 
        for the public to receive the Spanish-language form in a timely 
        and easy manner, which contributed to confusion and frustration 
        during the enumeration. We strongly recommend that the Census 
        Bureau determine how to make the non-English language census 
        forms more easily accessible to the public. For Census 2010, 
        data from the ACS on levels of English-language proficiency and 
        non-English language use would be especially valuable in 
        identifying areas where non-English language forms could be 
        made more readily available, including by mailing the non-
        English language forms in targeted communities with particular 
        non-English languages predominate.
           While it appears that adequate numbers of bilingual 
        enumerators were employed in areas with large Latino 
        populations, such as Los Angeles and south Texas, there was a 
        severe shortage of bilingual census workers in areas with 
        emerging Latino communities, such as western Pennsylvania and 
        western New York. For Census 2010, data from the ACS on levels 
        of English-language proficiency, non-English language use, and 
        nativity would be especially valuable in identifying all the 
        areas where bilingual enumerators will be required.
           While the paid advertising campaign appeared to be 
        successful overall, there were limitations on the media outlets 
        that were selected to broadcast the spots. Future advertising 
        should be distributed to those media outlets most watched/
        listened/read by Latinos.

     The Congress should encourage and prod the Census Bureau 
to hire more Latinos in key policy making position at the national 
office where Latinos presently are significantly under-represented. 
Latinos also are under-represented in senior management positions in 
the regional offices as well. Having Latinos in key positions 
throughout the Census Bureau is key to sustaining a successful decade-
long outreach effort in the Latino community.
    Question 2. You also highlight the importance of an accurate Census 
count to the Latino community. The Census Bureau says that there is a 
large discrepancy between their Demographic Analysis, official Census 
data, and A.C.E. results. What factors should Secretary Evans consider 
as he reviews criteria for releasing adjusted data?
    Answer. We are particularly concerned about the discrepancy 
suggested by the Demographic Analysis; we believe the Census Bureau 
should fully research possible errors in this analysis. First, analyses 
of the 2000 Census data suggest that thousands of persons may have been 
counted in 2000 who were in the country on April 1, 1990 and missed by 
the 1990 Census. Second, it also has been suggested by the analyses 
from Census 2000 and the A.C.E. that the assumptions about the levels 
of immigration during the 1990s used the Demographic Analysis were 
significantly miscalculated. These possible errors would have a 
particular impact on the Latino population and we strongly recommend 
that the Secretary consider these possibility of greater error in the 
1990 Census than what has been assumed, as well as mis-assumptions 
about immigration levels during the 1990s, which may have flawed the 
Demographic Analysis.
    In addition, I would like to bring to the attention of the Senate 
Committee the recommendations on this point forwarded to Secretary 
Evans by the Secretary of Commerce's Advisory Committee on the 
Decennial Census. These recommendations were forwarded by the Acting 
Chair of the Committee to Secretary Evans in a letter dated April 6, 
2001. The recommendations read as follows: (1) The Committee recommends 
that the Department of Commerce and the Bureau of the Census complete 
the work of evaluation of the Census 2000 and Accuracy and Coverage 
Evaluation (A.C.E.) data, report the data as completely as possible 
including by race, ethnicity, and geography, and complete the report 
expeditiously. (2) If the Bureau of the Census recommends adjusting the 
data, the Committee recommends that the Secretary publish the adjusted 
data at all levels of geography and incorporate the adjusted data into 
all data products. (3) If the Bureau of the Census does not find the 
adjusted data to be more accurate than Census 2000 data, then the 
Committee recommends that the adjusted data should be available for 
research purposes. (4) If the Bureau of the Census makes the 
recommendation to adjust the data, and the Department of Commerce 
decides to not release the adjusted data for official purposes, then 
the decision and the adjusted data should be made public.
                                 ______
                                 
      Response to Written Questions Submitted by Hon. John McCain 
                         to Eugene P. Ericksen
    Question 1. The Census Bureau's ESCAP Panel recommended against 
releasing adjusted Census numbers, because of the large discrepancy 
between the Demographic Analysis, the Accuracy and Coverage Evaluation 
(A.C.E.) sample, and the actual Census data. What may have caused this 
discrepancy?
    Answer. The national estimates of population provided by these 
three methods are: Demographic analysis: 279.6 million; Census count: 
281.4 million; A.C.E. estimate: 284.7 million.
    The demographic analysis estimate is probably too low. There are 
indications that the volume of international immigration, especially of 
the undocumented type, was greater than the Census Bureau had 
estimated. I believe that the valid construction of such estimates has 
become more difficult than it was in 1980 or 1990, and we must consider 
the possibility that valid demographic estimates may not be possible 
for 2000.
    The census count is probably too low, but perhaps not by very much. 
As indicated in the ESCAP report (see pp. 4-6), both demographic 
analysis and the A.C.E. indicate a Black undercount rate between 2 and 
3 percent. Estimates for the non-Black population provided by the two 
methods differ, primarily because they differ greatly for Hispanics. 
Indications from the A.C.E. suggest that net undercounts for Hispanics 
and similar to those for Blacks; undercount rates for non-Hispanic 
Whites and Asians are lower, but for Native Americans they are higher.
    The A.C.E. estimate may be too high. My reading of the ESCAP 
report, and various backup reports, suggests that the possibility that 
the Census Bureau underestimated the rate of erroneous enumeration. The 
bureau features two factors leading to this possibility. One is 
``balancing error,'' resulting from an inconsistency in the geographic 
search areas used in the P- and E-samples in the A.C.E. The other is 
due to high rates of duplication in the ``late census adds.''
    When deciding whether individual people were counted correctly in 
the A.C.E., the Census Bureau searches the block(s) surrounding each 
sample member's address. For example, if a person were missed at her 
own address, but counted next door, the bureau would not count such a 
person as an omission. Problems in locating surrounding addresses in 
the correct blocks could have created an imbalance in the searches. The 
Census Bureau is now investigating this possibility.
    The Census Bureau used a sophisticated computer program, and other 
methods to estimate that 6.5 million people had been counted twice at 
the same address. The bureau removed them from the count. Upon later 
reflection, the bureau decided that about 2.2 million of them may not 
have been counted twice, and they returned them to the count as ``late 
census adds.'' It is likely that a substantial number of these people 
were in fact duplicates although evidence for this is inconclusive.
    I have been told that the bureau may have underestimated erroneous 
enumerations by as many as 2 million people (see ESCAP report, p. 25). 
Should this be the case, the net undercount would be reduced to 1.3 
million people. It is likely that racial differential undercounts would 
remain, and that some groups would have net overcounts.
    Question 2. In your testimony, you argue that comparison of 
adjusted 1990 and 2000 Census data is required to accurately measure 
population growth. Why cannot sets of 1990 and 2000 unadjusted Census 
data be used to measure population growth.
    Answer. My comments referred to estimates of population growth, for 
local areas, where the rates of change are more variable than they are 
for the nation as a whole. As I testified, the reductions in the net 
undercount between 1990 and 2000 are much greater for minority than for 
the non-Hispanic White population.
    Let us consider two cities of 100,000 true population. One, City A, 
is 90 percent White and 10 percent Black and the other, City B, is 30 
percent White and 70 percent Black. Let us further assume that the 
local undercounts in each place mirror the national undercounts for 
each of the 1990 and 2000 Censuses. This is to say that the White 
undercount is 0.7 percent in both years, but that the Black undercount 
was 4.6 percent in 1990 and 2.2 percent in 2000.
    The undercount in City A was 1.09 percent in 1990 and 0.85 percent 
in 2000, a small improvement of 0.24 percent. For City B the comparable 
numbers are 3.43 percent in 1990 and 1.75 percent in 2000, a larger 
improvement of 1.68 percent. The City A population count would appear 
to have grown from 98,100 in 1990 to 99,050 in 2000. The City B count 
would appear to have grown even more from 96,570 to 98,250.
    For cities where the shares of rental housing are greater, the 
year-to-year discrepancies would be greater than this. As a 
consequence, a city's or county's population could appear to grow even 
though its population had not changed, due to the fact that the Census 
Bureau counted better in 2000.
    Many of the population counts obtained in 2000 have surprised local 
officials, and they are inconsistent with the Census Bureau's local 
population estimates for 1998 and 1999.
    I have taken these estimates and projected a 2000 estimate by 
projecting the 1998-99 change. For example if the Bureau's estimate was 
660,000 in 1998 and 650,000 in 1999, the expected 2000 estimate was 
640,000. I then compared the expectation to the actual count. I did 
this for all counties with at least 500,000 people in the 1990 Census.
    As shown in the attached table, most of the expected values were 
lower than the counts. This was especially true in those counties with 
large minority populations. I obtained the following results:

    Average Percentage:

    Percent Black Or Hispanic, 2000: 40 percent or more/Difference, 
Estimate And Count: 3.84 percent.
    Percent Black Or Hispanic, 2000: 20 to 39.9 percent/Difference, 
Estimate And Count: 3.09 percent.
    Percent Black Or Hispanic, 2000: 0 to 19.9 percent/Difference, 
Estimate And Count: 1.96 percent.

    The difference is defined as (count-estimate)/count. A positive 
number means that the count was greater than the estimate. Of the 97 
counties in my studies, 86 had positive discrepancies. The above 
results show that the discrepancies were on average large in counties 
with greater shares of Black or Hispanic populations.

                                        Errors in Population Estimates for 2000 by Counties and Percent Minority
--------------------------------------------------------------------------------------------------------------------------------------------------------
                                                                 Percent
                State                          County          Hispanic or      1990         2000         1998         1999         2000        Error
                                                                 NH Black    Population   Population    Estimate     Estimate     Estimate
--------------------------------------------------------------------------------------------------------------------------------------------------------
Texas...............................  El Paso................       80.98%      591,610      679,622      694,603      701,908      709,213       -4.35%
New York............................  Bronx..................        79.62    1,203,789    1,332,650    1,191,319    1,194,099    1,196,879        10.19
Florida.............................  Dade...................        76.28    1,937,094    2,253,362    2,150,877    2,175,634    2,200,391         2.35
Maryland............................  Prince Georges.........        69.29      729,268      801,515      776,907      781,781      786,655         1.85
District of Columbia................  67.31..................      606,900      572,059      521,426      519,000      516,574         9.70
Maryland............................  Baltimore City.........        65.74      736,014      651,154      645,664      632,681      619,698         4.83
Georgia.............................  De Kalb................        61.71      545,837      665,865      592,870      596,853      600,836         9.77
Texas...............................  Bexar..................        61.24    1,185,394    1,392,931    1,354,837    1,372,867    1,390,897         0.15
New Jersey..........................  Essex..................        55.72      778,206      793,633      748,322      747,355      746,388         5.95
New York............................  Kings..................        54.21    2,300,664    2,465,326    2,266,242    2,268,297    2,270,352         7.91
California..........................  Los Angeles............        54.03    8,863,164    9,519,338    9,223,807    9,329,989    9,436,171         0.87
New Jersey..........................  Hudson.................        51.92      553,099      608,975      553,030      552,819      552,608         9.26
Texas...............................  Harris.................        51.15    2,818,199    3,400,578    3,202,021    3,250,404    3,298,787         2.99
Pennsylvania........................  Philadelphia...........        51.07    1,585,577    1,517,550    1,434,968    1,417,601    1,400,234         7.73
Tennessee...........................  Shelby.................        50.98      669,018      897,472      867,804      873,000      878,196         2.15
Georgia.............................  Fulton.................        50.13      648,951      816,006      737,222      744,827      752,432         7.79
Texas...............................  Dallas.................        49.95    1,852,810    2,218,899    2,045,309    2,062,100    2,078,891         6.31
California..........................  Fresno.................        49.03      667,490      799,407      755,051      763,069      771,087         3.54
California..........................  San Bernardino.........        47.94    1,418,380    1,709,434    1,635,967    1,669,934    1,703,901         0.32
Illinois............................  Cook...................        45.79    5,105,067    5,376,741    5,192,396    5,192,326    5,192,256         3.43
Michigan............................  Wayne..................        45.69    2,111,687    2,061,162    2,116,540    2,106,495    2,096,450        -1.71
California..........................  Kern...................        44.11      543,477      661,645      631,615      642,495      653,375         1.25
New York............................  Queens.................        43.93    1,951,598    2,229,379    1,993,172    2,000,642    2,008,112         9.93
New York............................  New York...............        42.45    1,487,536    1,537,195    1,546,508    1,551,844    1,557,180        -1.30
California..........................  Riverside..............        42.19    1,170,413    1,545,387    1,480,708    1,530,653    1,580,598        -2.28
Alabama.............................  Jefferson..............        40.77      651,525      662,047      660,039      657,422      654,805         1.09
  ..................................  .......................  ...........  ...........  ...........  ...........  ...........      Average        3.84%
Texas...............................  Travis.................        37.21      576,407      812,280      709,182      727,022      744,862         8.30
Florida.............................  Broward................        36.78    1,255,488    1,623,018    1,507,770    1,535,468    1,563,166         3.69
Massachusetts.......................  Suffolk................        36.36      511,433      689,807      641,333      641,695      642,057         6.92
Florida.............................  Orange.................        36.31      677,491      896,344      804,489      817,206      829,923         7.41
California..........................  Ventura................        35.21      669,018      753,197      732,143      745,063      757,983        -0.64
North Carolina......................  Mecklenburg............        34.12      649,623      695,454      630,813      648,400      665,987         4.24
California..........................  Alameda................        33.60    1,279,182    1,443,741    1,397,050    1,415,582    1,434,114         0.67
Wisconsin...........................  Milwaukee..............        33.07      959,275      940,164      911,536      906,248      900,960         4.17
Florida.............................  Hillsborough...........        32.43      834,054      998,948      925,413      940,484      955,555         4.34
Texas...............................  Tarrant................        32.36    1,170,103    1,446,219    1,354,040    1,382,442    1,410,844         2.45
California..........................  Orange.................        32.26    2,410,558    2,846,289    2,723,782    2,760,948    2,798,114         1.69
Arizona.............................  Pima...................        32.19      666,880      843,746      790,333      803,618      816,903         3.18
California..........................  San Diego..............        32.18    2,498,016    2,813,833    2,766,123    2,820,844    2,875,565        -2.19
Florida.............................  Duval..................        31.64      672,971      778,879      734,664      738,483      742,302         4.70
Nevada..............................  Clark..................        30.79      741,459    1,375,765    1,161,259    1,217,155    1,273,051         7.47
Ohio................................  Cuyahoga...............        30.59    1,412,140    1,393,978    1,380,428    1,371,717    1,363,006         2.22
Tennessee...........................  Davidson...............        30.36    1,497,577      569,891      533,258      530,050      526,842         7.55
New York............................  Westchester............        29.17      874,866      923,459      900,861      905,572      910,283         1.43
Missouri............................  Jackson................        28.48      633,232      654,880      655,055      654,484      653,913         0.15
Arizona.............................  Maricopa...............        28.38    2,122,101    3,072,149    2,783,779    2,861,395    2,939,011         4.33
Indiana.............................  Marion.................        27.89      797,159      860,454      812,662      810,946      809,230         5.95
New Jersey..........................  Camden.................        26.96      502,824      508,932      504,268      503,093      501,918         1.38
California..........................  Contra Costa...........        26.84      803,732      948,816      917,970      933,141      948,312         0.05
California..........................  Santa Clara............        26.62    1,497,577    1,682,585    1,641,848    1,647,419    1,652,990         1.76
Maryland............................  Montgomery.............        26.33      757,027      873,341      839,158      852,174      865,190         0.93
Florida.............................  Palm Beach.............        25.91      863,518    1,131,184    1,032,872    1,049,420    1,065,968         5.77
California..........................  Sacramento.............        25.66    1,041,219    1,223,499    1,166,699    1,184,586    1,202,473         1.72
California..........................  San Mateo..............        25.24      649,623      707,161      701,080      702,102      703,124         0.57
Ohio................................  Hamilton...............        24.45      866,228      845,303      847,202      840,443      833,684         1.37
Oklahoma............................  Oklahoma...............        23.55      599,611      660,448      632,865      636,539      640,213         3.06
Connecticut.........................  Hartford...............        22.65      851,783      857,183      827,706      829,671      831,636         2.98
New Jersey..........................  Middlesex..............        22.18      671,780      750,162      712,638      717,949      723,260         3.59
Maryland............................  Baltimore County.......        21.77      692,134      754,292      721,556      723,914      726,272         3.71
California..........................  San Francisco..........        21.67      723,959      776,733      745,756      746,777      747,798         3.73
Connecticut.........................  Fairfield..............        21.48      827,645      882,567      837,476      841,334      845,192         4.23
Illinois............................  Lake...................        21.15      516,418      644,356      608,348      617,975      627,602         2.60
Ohio................................  Montgomery.............        21.03      573,809      559,062      570,141      565,866      561,591        -0.45
Connecticut.........................  New Haven..............        20.92      804,219      824,008      792,879      793,208      793,537         3.70
Kentucky............................  Jefferson..............        20.53      664,937      693,604      671,595      672,900      674,205         2.80
Missouri............................  Saint Louis............        20.38      993,529    1,016,315      997,347      996,181      995,015         2.10
Ohio................................  Franklin...............        20.01      961,437    1,068,978    1,021,578    1,027,821    1,034,064         3.27
  ..................................  .......................  ...........  ...........  ...........  ...........  ...........      Average        3.09%
New York............................  Nassau.................        19.72    1,287,348    1,334,544    1,300,995    1,305,057    1,309,119         1.91
Virginia............................  Fairfax................        19.41      818,584      969,749      927,895      945,717      963,539         0.64
Rhode Island........................  Providence.............        19.19      596,270      621,602      573,701      574,108      574,515         7.58
New York............................  Monroe.................        18.66      713,968      735,343      714,936      712,419      709,902         3.46
New York............................  Suffolk................        17.10    1,321,864    1,419,369    1,370,549    1,383,847    1,397,145         1.57
Oklahoma............................  Tulsa..................        16.81      503,341      563,299      543,417      548,296      553,175         1.80
New York............................  Erie...................        16.03      968,532      950,265      933,702      925,957      918,212         3.37
Pennsylvania........................  Delaware...............        15.87      547,651      550,864      542,592      541,502      540,412         1.90
Michigan............................  Kent...................        15.70      500,631      574,335      544,781      550,388      555,995         3.19
New Jersey..........................  Bergen.................        15.33      825,380      884,118      854,428      857,052      859,676         2.76
Ohio................................  Summit.................        14.00      514,990      542,899      537,160      537,856      538,552         0.80
New Jersey..........................  Monmouth...............        13.96      553,124      615,301      603,214      611,444      619,674        -0.71
Florida.............................  Pinellas...............        13.45      851,659      921,482      877,273      878,499      879,725         4.53
Pennsylvania........................  Allegheny..............        13.20    1,336,449    1,281,666    1,267,963    1,256,806    1,245,649         2.81
Oregon..............................  Multnomah..............        13.05      583,887      660,486      630,573      633,224      635,875         3.73
Massachusetts.......................  Essex..................        13.03      670,080      723,419      700,370      704,407      708,444         2.07
Minnesota...........................  Hennepin...............        12.91    1,032,431    1,116,200    1,058,943    1,064,419    1,069,895         4.15
Utah................................  Salt Lake..............        12.83      725,956      898,387      845,913      850,243      854,573         4.88
Michigan............................  Oakland................        12.45    1,083,592    1,194,156    1,175,057    1,179,978    1,184,899         0.78
Washington..........................  Pierce.................        12.27      586,203      700,820      675,962      688,807      701,652        -0.12
Illinois............................  DuPage.................        11.98      781,666      904,161      880,996      892,547      904,098         0.01
Washington..........................  King...................        10.77    1,507,319    1,737,034    1,654,329    1,664,846    1,675,363         3.55
Pennsylvania........................  Montgomery.............         9.41      678,111      750,097      719,569      724,087      728,605         2.87
Massachusetts.......................  Worcester..............         9.20      709,705      750,963      730,769      738,629      746,489         0.60
Hawaii..............................  Honolulu...............         8.94      871,768      876,156      871,768      864,571      857,374         2.14
Massachusetts.......................  Middlesex..............         7.73    1,398,468    1,465,396    1,422,465    1,426,606    1,430,747         2.36
Pennsylvania........................  Bucks..................         5.52      541,174      597,635      587,863      594,047      600,231        -0.43
Massachusetts.......................  Bristol................         5.43      506,325      534,678      516,975      520,258      523,541         2.08
Massachusetts.......................  Norfolk................         4.94      616,087      650,308      642,089      643,580      645,071         0.81
Michigan............................  Macomb.................         4.26      717,400      788,149      786,866      792,082      797,298        -1.16
                                      .......................  ...........  ...........  ...........  ...........  ...........      Average        1.96%
--------------------------------------------------------------------------------------------------------------------------------------------------------
Source: www.census.gov/People estimates/County Population Estimates/ County Population Estimates for July 1, 1999 and the Population Change from July 1,
  1998 to July 1, 1999/American FactFinder/Race Hispanic or Latino (2000)

     Response to Written Questions Submitted by Hon. Barbara Boxer 
                           to Donald L. Evans
    Question 1. Mr. Secretary, I am requesting two things. First, for 
the Census Bureau to immediately release all the data collected and the 
adjusted numbers it believes are inaccurate so that independent experts 
can examine the numbers and assumptions.
    Answer. In keeping with our goal of openness and transparency in 
decisions and processes, the Census Bureau signed a Memorandum of 
Understanding (MOU) with the U.S. House of Representatives Committee on 
Government Reform Subcommittee on the Census, the Census Monitoring 
Board, and the National Academy of Sciences' Committee on National 
Statistics, granting access to data to facilitate independent review of 
Census 2000. Presently, a second MOU is under consideration by these 
oversight groups. This MOU provides the opportunity for outside 
researchers to study the A.C.E. Census 2000 and demographic analysis 
data while appropriately controlling access to sensitive data and 
protecting the ongoing deliberative process of Census Bureau staff and 
other officials of the Department of Commerce. The data now being made 
available under this MOU have been used by the Census Bureau for 
assessing the accuracy of the Accuracy and Coverage Evaluation (A.C.E) 
survey results. We believe that providing these materials will further 
the oversight responsibilities of the Congress and the other above-
mentioned entities as well as satisfy the requests by many Members of 
Congress and other interested parties that the A.C.E. data be made 
available. Although the Senate had not previously participated in this 
review process, we have contacted the committees of jurisdiction to 
extend the same opportunity.
    Question 2. Second, for the Census Bureau to move with immediate 
haste to resolve the issues it believes have distorted the adjusted 
numbers and provide Congress with adjusted numbers well before the Fall 
deadline for determining distribution of federal funds. Can you assure 
myself and the Committee that these requests will be met?
    Answer. The Census Bureau is currently engaged in a comprehensive 
research and evaluation effort to address all outstanding issues and to 
resolve questions raised by the Executive Steering Committee for 
Accuracy and Coverage Evaluation Policy (ESCAP) report issued on March 
1, 2001. Our commitment to provide the Congress and the public with the 
most accurate and statistically sound data remains our primary 
objective. As soon as possible, the Census Bureau plans to release its 
findings and make a recommendation to the Secretary of Commerce 
concerning the accuracy of the adjusted data. We will provide these 
findings to the Congress as swiftly as possible.
                                 ______
                                 

                    [From USA TODAY, March 28, 2001]

          Report: Cities, Minorities Lose in Census Undercount

                 (By Haya El Nasser and Paul Overberg)

    California could claim half a million more people, Texas another 
350,000 and New York 300,000 if the 2000 Census were adjusted to make 
up for the people who were missed, according to a Democratic-backed, 
state-by-state undercount estimate out today.
    The undercount numbers for 50 states, the District of Columbia and 
five big cities are the first to indicate where an estimated 3.3 
million missed by the Census were living.
    The numbers, released by Democrats on the Census Monitoring Board, 
are likely to put more pressure on the Census Bureau to release 
adjusted population counts for all areas.
    The Census Bureau is still reviewing the numbers.
    The estimates were done by Eugene Ericksen, a statistics professor 
at Temple University and a consultant to Democrats on the monitoring 
board. He used the methods outlined by the Census Bureau in its own 
study of whether Census numbers should be adjusted.
    His analysis shows the largest percentages of people missed were in 
states with a high number of minorities and big cities. The net 
undercounts were smallest in some Midwestern and Plains states.
    Undercounts have broad implications for states and cities because 
Census numbers are used to redraw political districts and distribute 
$185 billion in Federal funds annually.
    Most Republicans are opposed to the estimates because they say it 
creates ``virtual'' people. Democrats say it's the only way to correct 
the disproportionate undercount of minorities and the poor.
    ``At this point, we don't have the confidence in (our adjusted) 
numbers to release them,'' says John Thompson, the Census Bureau's 
associate director for the decennial Census.
    Acting Census Director William Barron said he could not comment on 
Ericksen's estimates because he had not seen his methodology. Commerce 
Secretary Donald Evans, who oversees the Census Bureau, is to testify 
about the numbers today before a Senate panel.
    The estimates claim about 188,500 people were missed in New York 
City (2.3 percent undercount); more than 62,000 in Chicago (2.1 
percent); 44,000 in Houston (2.2 percent); 21,500 in Philadelphia (1.4 
percent); and 9,400 in Atlanta (2.2 percent).
    Census 2000 was the most accurate count ever. Undercount estimates 
released today show that 35 states--including the two biggest, 
California and Texas--had a smaller undercount in 2000 than in 1990.
    In 2000, officials said the Census missed at least 6.4 million and 
counted at least 3.1 million twice, a net undercount of 3.3 million. In 
1990, it missed 8.4 million and counted 4.4 million twice, a net 
undercount of 4 million.
    The Census Bureau recommended against using adjusted numbers 
because it couldn't explain huge gaps between the Census, adjusted 
numbers and estimates based on birth, death and immigration records.
[GRAPHIC] [TIFF OMITTED] 88017.004

  

                                  
