[House Hearing, 106 Congress]
[From the U.S. Government Publishing Office]
OVERSIGHT OF THE 2000 CENSUS: ACCURACY AND COVERAGE EVALUATION [ACE]--
STILL MORE QUESTIONS THAN ANSWERS
=======================================================================
HEARING
before the
SUBCOMMITTEE ON THE CENSUS
of the
COMMITTEE ON
GOVERNMENT REFORM
HOUSE OF REPRESENTATIVES
ONE HUNDRED SIXTH CONGRESS
SECOND SESSION
__________
MAY 19, 2000
__________
Serial No. 106-207
__________
Printed for the use of the Committee on Government Reform
Available via the World Wide Web: http://www.gpo.gov/congress/house
http://www.house.gov/reform
U.S. GOVERNMENT PRINTING OFFICE
71-389 WASHINGTON : 2001
_______________________________________________________________________
For sale by the Superintendent of Documents, U.S. Government Printing
Office
Internet: bookstore.gpo.gov Phone: (202) 512-1800 Fax: (202) 512-2250
Mail: Stop SSOP, Washington, DC 20402-0001
COMMITTEE ON GOVERNMENT REFORM
DAN BURTON, Indiana, Chairman
BENJAMIN A. GILMAN, New York HENRY A. WAXMAN, California
CONSTANCE A. MORELLA, Maryland TOM LANTOS, California
CHRISTOPHER SHAYS, Connecticut ROBERT E. WISE, Jr., West Virginia
ILEANA ROS-LEHTINEN, Florida MAJOR R. OWENS, New York
JOHN M. McHUGH, New York EDOLPHUS TOWNS, New York
STEPHEN HORN, California PAUL E. KANJORSKI, Pennsylvania
JOHN L. MICA, Florida PATSY T. MINK, Hawaii
THOMAS M. DAVIS, Virginia CAROLYN B. MALONEY, New York
DAVID M. McINTOSH, Indiana ELEANOR HOLMES NORTON, Washington,
MARK E. SOUDER, Indiana DC
JOE SCARBOROUGH, Florida CHAKA FATTAH, Pennsylvania
STEVEN C. LaTOURETTE, Ohio ELIJAH E. CUMMINGS, Maryland
MARSHALL ``MARK'' SANFORD, South DENNIS J. KUCINICH, Ohio
Carolina ROD R. BLAGOJEVICH, Illinois
BOB BARR, Georgia DANNY K. DAVIS, Illinois
DAN MILLER, Florida JOHN F. TIERNEY, Massachusetts
ASA HUTCHINSON, Arkansas JIM TURNER, Texas
LEE TERRY, Nebraska THOMAS H. ALLEN, Maine
JUDY BIGGERT, Illinois HAROLD E. FORD, Jr., Tennessee
GREG WALDEN, Oregon JANICE D. SCHAKOWSKY, Illinois
DOUG OSE, California ------
PAUL RYAN, Wisconsin BERNARD SANDERS, Vermont
HELEN CHENOWETH-HAGE, Idaho (Independent)
DAVID VITTER, Louisiana
Kevin Binger, Staff Director
Daniel R. Moll, Deputy Staff Director
David A. Kass, Deputy Counsel and Parliamentarian
Lisa Smith Arafune, Chief Clerk
Phil Schiliro, Minority Staff Director
Subcommittee on the Census
DAN MILLER, Florida, Chairman
THOMAS M. DAVIS, Virginia CAROLYN B. MALONEY, New York
PAUL RYAN, Wisconsin DANNY K. DAVIS, Illinois
MARK E. SOUDER, Indiana HAROLD E. FORD, Jr., Tennessee
------ ------
Ex Officio
DAN BURTON, Indiana HENRY A. WAXMAN, California
Jane Cobb, Staff Director
Lara Chamberlain, Professional Staff Member
Michael Miguel, Professional Staff Member
Andrew Kavaliunas, Clerk
Michelle Ash, Minority Counsel
C O N T E N T S
----------
Page
Hearing held on May 19, 2000..................................... 1
Statement of:
Prewitt, Kenneth, Director, Bureau of the Census, accompanied
by John H. Thompson, Associate Director for Decennial
Census; and Howard Hogan, statistician, Chief, Decennial
Statistical Studies Division............................... 29
Letters, statements, etc., submitted for the record by:
Maloney, Hon. Carolyn B., a Representative in Congress from
the State of New York:
Letter dated May 18, 2000................................ 11
Prepared statement of.................................... 13
Miller, Hon. Dan, a Representative in Congress from the State
of Florida:
CRS report concerning sampling........................... 71
December 1992 Federal Register........................... 44
Prepared statement of.................................... 4
Various editorials....................................... 17
Prewitt, Kenneth, Director, Bureau of the Census, prepared
statement of............................................... 35
Ryan, Hon. Paul, a Representative in Congress from the State
of Wisconsin, letter dated May 17, 2000.................... 23
OVERSIGHT OF THE 2000 CENSUS: ACCURACY AND COVERAGE EVALUATION [ACE]--
STILL MORE QUESTIONS THAN ANSWERS
----------
FRIDAY, MAY 19, 2000
House of Representatives,
Subcommittee on the Census,
Committee on Government Reform,
Washington, DC.
The subcommittee met, pursuant to notice, at 9:30 a.m., in
room 2247 Rayburn House Office Building, Hon. Dan Miller
(chairman of the subcommittee) presiding.
Present: Representatives Miller, Ryan, Maloney, and Davis
of Illinois.
Staff present: Jane Cobb, staff director; Chip Walker,
deputy staff director; Lara Chamberlain, Michael Miguel, and
Amy Althoff, professional staff members; Andrew Kavaliunas,
clerk; Michelle Ash, minority counsel; David McMillen and Mark
Stephenson, minority professional staff members; and Earley
Green, minority assistant clerk.
Mr. Miller. Good morning. We will begin with opening
statements first, and then we'll hear from Director Prewitt and
then we'll proceed.
Director Prewitt, thank you for being with us here today.
I'm pleased to hear that the Census Bureau is proceeding on
schedule for nonresponse followup. This is the most difficult
part of the census with respect to ACE, or the estimation plan.
As we move into the politically charged arena of the
Bureau's estimation plan, one of the greatest concerns to
Congress and the scientific community is that we will not be
provided all of the information and data necessary to evaluate
the results in a timely manner.
I want to get the Director's assurances today that these
numbers will be fully scrutinized by the Bureau and the
scientific community at large prior to their release for public
use.
As many of you know, in 1990 there were numerous errors
found in the sampling plan known as the PES. After the census
in 1991 the Bureau discovered a computer error in the PES that
inflated the undercount by 1 million people. Then, during a
series of evaluations that took almost 2 years, the Bureau
discovered more errors in the PES leading to even more
erroneous enumerations.
In 1992, the experts at the Bureau who reviewed the 1990
census estimation plan issued what's known as the CAPE report.
These experts reported that, ``About 45 percent of the revised
estimated undercount is actually measured by us and not
measured undercount.'' In other words, in 1990 almost half of
the statistical adjustment was wrong.
Once States draw their district lines you can't come back a
year later and say, ``sorry, we made a mistake, we added or
subtracted too many people.'' If it took almost 2 years in 1990
to find the errors, how can you ensure that we don't have the
same errors this time in only a few months?
From a practical perspective, there is no guarantee that
this plan is even viable. Despite claims to the contrary, the
National Academy panel has not given the ACE a full blessing.
While certain groups have endorsed statistical estimation as a
concept, this is a far cry from an endorsement of the actual
plan.
To give you an analogy, we all know that it's possible to
build a spaceship to go to the moon or Mars. Yet, as with all
very complex scientific tasks--and the estimation plan is
immensely complex--your spaceship could blow up on the
launching pad or burn up in the atmosphere as has happened
twice recently.
What assurances do we have from the Director that their
scientific plan won't blow up? Just because something may be
theoretically possible doesn't mean it can be done.
Despite claims by the Democrats, Republican opposition to
the estimation plan is based on fundamental, unresolved
problems. Is the plan constitutional, is it legal, is it in the
best interest of our Nation as a whole, and simply, will it
work?
In January 1999, the Supreme Court ruled that sampling or
estimating portions of the population was illegal. Democrats
read the decision to outlaw the issue of sampling for purposes
of apportionment only, while Republicans read the decision to
prohibit the use of estimation for redistricting as well.
In the wake of that decision, the nonpartisan Congressional
Research Service issued an opinion, ``A closer examination of
other parts of the Court's opinion indicates that it did not
interpret those other purposes as necessarily including at
least intrastate redistricting.''
Unfortunately, this administration was not going to be
deterred by even the Supreme Court. In a political move clearly
against the best interests of the Nation, the Clinton-Gore
administration decided to conduct a two-number census. This
unwise decision will clearly throw the States into legal
turmoil over the census, the likes of which this Nation has
never seen.
And while the Democrats and their so-called experts have
claimed that it is perfectly legal to use estimation for
purposes of redistricting, I would simply offer a few words of
caution. These are the same so-called experts that said
estimation could be used for apportionment. Those on the
estimation side of this disagreement have yet to win a court
case.
The fundamental purpose of the census is to fairly and
accurately apportion and distribute political representation.
Our political system, for the most part, is the envy of many
other nations. One of the foundations of our system is its
relative transparency. Our elections are carefully scrutinized
and the appeals process clear. If warranted, for example, an
election can be challenged, voter registration records can be
checked and rechecked, ballots recounted. With estimation, it's
simply not possible to verify whether or not a person added
actually exists or if a person subtracted was done so
rightfully.
Additions and subtractions exist only in a virtual world, a
world based not in reality, but in the complex mathematical
formulas that could be right or wrong and understood only by a
few select statisticians and government bureaucrats. Census
estimation, no matter who is crunching the numbers, is not a
system that lends itself to trust and integrity, two
cornerstones of our electoral process.
And while we have spent billions of dollars to motivate
people to participate in the census, something that all sides
agree is a civic ceremony, what would motivate someone to
participate in the census when they can sit back and be
estimated? Why fill out your census forms at all if the
government will compensate for you anyway? And even worse, how
can it be acceptable that someone does their civic obligation,
fills out their form on time and sends it in, only to have the
government say they count as less than a whole person? Is this
not a violation of one man, one vote? Can the Director
guarantee that every person who filled out a form and only one
form will be counted as one person and not less? The answer,
disturbingly, is no.
The fact not widely talked about in the Bureau is, there
will be people who do everything right, fill out their census
forms, send them in on time and will be counted as less than a
whole person.
While I fully support expending the resources to reach the
undercounted, I wholeheartedly oppose the concept of counting
someone as less than a whole person.
The census has traditionally been constructed of millions
and millions of individuals. However, this estimation plan has
introduced a new level of demographic grouping that is very
dangerous in its assumptions. The assumption that people within
racial groups act alike and have the same tendencies is
something that this Nation has been trying to overcome for over
100 years, but now the Census Bureau has gone down that exact
path.
No longer are we individuals. Now we are White males, 25 to
35, who rent or own. We are Cubans in Miami, Mexicans in Texas,
Puerto Ricans in New York, that according to the Census Bureau
are all alike and thus are grouped together. We are Asians,
including Chinese, Japanese and Koreans from Seattle to
Washington, DC, grouped together like so many choices.
The by-products of this estimation plan are not healthy for
our Nation. From civic disengagement to simply throwing one
man-one vote out the window, we in the long run hurt our
Nation.
We must do everything possible to eliminate the undercount.
We must also remain faithful to our Constitution, the law and
the civic health of our Nation. Clearly this administration is
putting politics ahead of sound public policy. Unfortunately,
it will take the courts once again to protect the integrity of
our census.
[The prepared statement of Hon. Dan Miller follows:]
[GRAPHIC] [TIFF OMITTED] T1389.001
[GRAPHIC] [TIFF OMITTED] T1389.002
[GRAPHIC] [TIFF OMITTED] T1389.003
Mrs. Maloney.
Mrs. Maloney. Thank you, Mr. Chairman. I must say I find
the title of today's hearing very curious. It's called Status
of the 2000 Census, Accuracy and Coverage Evaluation, Still
More Questions Than Answers. Yet it seems to me, Mr. Chairman,
that there are almost no unanswered questions, only questions
which you don't like the answers to.
Despite the cautious stance taken by the Census Bureau, I
believe that the 2000 census may well be the best, fairest and
most accurate census ever, a fitting way to start the 21st
century. It will be that not just because of the operational
successes we have seen to date, but because it incorporates
modern scientific methods into its design.
We all know the problems of the 1990 census. It contained
millions of errors and was the first to be less accurate than
the census before it. The 1990 census had an error rate of over
10 percent, 8.4 million people were missed, 4.4 million people
were counted twice and 13 million people were counted in the
wrong place. And we know who the people were that were missed.
They were children. They were minorities in urban and rural
areas.
The Census Bureau, working with the National Academy of
Sciences at the direction of Congress in a bipartisan way, has
tried to fix these errors, but there are politicians who, for
partisan reasons, have tried to make sure that it doesn't
happen. They have tried to make sure that minorities in poor,
urban and rural areas and children are not undercounted.
The Census Bureau first discovered the problem of the
undercount during World War II, 60 years ago, when more young
men showed up for the draft than the Census Bureau thought
existed. For young Black men, nearly 13 percent more showed up
for the draft than they expected, showing that there was a
disproportionate undercount, particularly for minorities.
We now know that the people missed in the census are the
urban and rural poor and minorities. We also know that the
people counted twice are primarily people who are fortunate
enough to have two homes. They're suburbanites. Those errors
shift economic resources and political representation unfairly.
Those who oppose the use of modern scientific methods in the
census would ensure that millions of people missed in the
census are left out permanently and the millions of people
counted twice are forever kept in. That is fundamentally unfair
and it is unjust and it must stop, and that is why this census
has been called the civil rights issue of the decade--whether
we will correct, knowing people that are left out, whether we
will do what every scientist says needs to be done to make sure
that they are counted and represented.
The closer the Census Bureau has gotten to developing a way
to fix these errors, the harder the opponents of a modern
census have worked to stop them. In 1987, the professionals at
the Census Bureau proposed a 300,000-household survey to
measure and correct for the errors of the census. The
politicians at the Reagan Commerce Department stopped that
planning dead in its tracks. Correcting the 1990 census would
have been stopped for good had not the great city of New York--
and I am proud to be a Representative from that great city--the
city of New York sued.
Finally, in late 1989, the Commerce Department allowed
planning for the quality control survey to go forward. However,
instead of allowing it to be a 300,000-household survey, the
politicians at then-President Bush's Commerce Department cut it
in half, and the Secretary reserved the right to block the use
of the corrected, modern, scientific numbers. Not surprisingly,
he overruled his own Republican-appointed Census Director, Dr.
Barbara Bryant, and the professional nonpartisans at the Census
Bureau, and he did block their use.
In 1997, the opponents of a fair and accurate census, they
held up a disaster relief bill to the Midwest because they
attached language to this important bill that would have
prohibited the use of modern scientific methods. They believed
that the President of the United States, when the country was
in disaster, people were suffering, their homes were under
water, that he would not have the nerve to veto the disaster
relief bill over the census, over an accurate census, yet the
President vetoed the disaster relief bill over the census
because of their crass attempt to manipulate the numbers in the
census; and he received editorial support clear across this
country.
And I would like permission, Mr. Chairman, to put all of
those editorials in the record of this hearing.
Mr. Miller. OK.
Mrs. Maloney. In 1997 and again in 1998, opponents of a
fair and accurate census, those who did not want minorities and
the poor and children in urban and rural areas counted, tried
to use the appropriations process, the budget process, to
legislate how census 2000 would be conducted by threatening to
hold up two budgets and close down the government. Their
attempt to block the use of modern scientific methods failed
again.
Principally at the direction of Congress, the National
Academy of Sciences has conducted extensive research and review
of the planning and implementation of the 2000 census
throughout the decade of the 1990's, working with the Census
Bureau. Four separate panels of independent experts have
consistently supported the use of modern scientific statistical
methods, in general. More recently, a fifth panel, the
Academy's panel to review the 2000 census has endorsed the
Bureau's specific plans for the ACE program in census 2000; and
I would like permission to put in the record the five
statistical reports and scientific reports that have come out
in support of the Bureau's plans.
May they be put in the record?
Mr. Miller. No objection.
Mrs. Maloney. The overwhelming majority of the scientific
community has concluded that if we are to have a 2000 census
that is fair and accounts for all residents of this country,
regardless of race or economic status, it must be a census that
uses modern methods. And I would now like to put into the
record this list of organizations that supports the use of
modern scientific statistical methods, and it includes all
kinds of associations from across this country.
Mr. Miller. Without objection.
Mrs. Maloney. The General Accounting Office, the Commerce
Department's Inspector General and George Bush's Census
Director, Dr. Barbara Bryant, are all on record in their
support.
The Census Bureau has presented its plans for the use of
modern methods to the scientific community on a continuing
basis since 1996. This subcommittee and the Census Monitoring
Board have been kept apprised of those plans since their
inception, and the Secretary's 2000 Census Advisory Committee,
Race and Ethnic Advisory Committees and Census Advisory
Committees of Professional Associations have all been briefed
on these plans as well.
Again, Mr. Chairman, there are not any unanswered questions
about the ACE program, only answers that the Republican
Conference and the RNC doesn't like. I have heard the opponents
of modern methods say repeatedly that they are a, ``risky
experimental plan that is inaccurate, that all we need to do is
use old methods and try really hard to just count everyone.''
Well, the only thing that is risky is not using modern methods.
Over the course of the years, working on this issue, I have
repeatedly heard from people, how do you know that you missed
people, how do you know that there is an undercount? Well, the
Census Bureau is unique among government agencies in that they
tell you how well they have done. And the only way we know that
the 1990 census was less accurate than the one before it was
the 1990 post-enumeration survey, the use of modern statistical
methods.
And the only way we will be able to determine the most
accurate count for the 2000 census is from the results of the
ACE program, the use of modern statistical methods. In the end,
it is only through those methods that we will have the most
accurate census possible.
I would also like to comment briefly on the Supreme Court
case. Very simply, in the Supreme Court case, the Republicans
won one and the Democrats won two. The Court held that you
could not use modern scientific methods for the apportionment
of seats between the States, but if at all feasible, you could
use it for the distribution of Federal funds which is
tremendously important since the Federal Government distributes
over $185 billion a year based on funding formulas that are
tied to census numbers. That means that over $3 trillion in the
next decade will be distributed on these numbers, and we need
to make them accurate; and they also held that it could be used
for redistricting within the States, and that is why we have to
come forward with two numbers, one for reapportionment between
the States and one for redistricting and the distribution of
funds.
I'd also like, Mr. Chairman, to briefly comment on the e-
mail that you made public last week. Mr. Chairman, I continue
to believe that it was inappropriate to make this public
without first talking to its author, especially since you used
it to imply a vast conspiracy by the census to hide information
from Congress. And I remain disturbed by the fact that I had
little more than an hour's notice of this e-mail's existence
prior to our hearing. And I still am disturbed that GAO, a
supposedly nonpartisan independent body, contacted the staff of
the majority but did not contact the staff of the Democratic
minority.
The e-mail may be poorly worded, but after speaking to its
author, I called the gentleman. He is an honest, hardworking
American. He is a former Marine. He is working very hard now
for his country on this great civic ceremony, the census. I
believe him, that he made an honest mistake that is not in any
way evidence of a systematic attempt to deny information to
Congress.
Nevertheless, it seems to have raised questions in your
mind which should be put to rest. Therefore, I have written to
the Comptroller of GAO asking him to investigate this incident
as soon as possible and to have the GAO determine if there is
an intentional, systematic attempt to hide information from
Congress. I'd like to introduce that letter into the record.
[The information referred to follows:]
[GRAPHIC] [TIFF OMITTED] T1389.004
Mrs. Maloney. I'd like to thank the chairman, and I would
also like to put into the record, since at times this issue has
been called partisan, and regrettably sometimes it has been
partisan in our comments--so, therefore, to bring the debate
above a partisan level, I would like to introduce into the
record all of the editorials from across this country in
support of the Census Bureau's plans, in support of the use of
modern scientific methods to correct the undercount, and it
comes from the Miami Herald, the Houston Chronicle, newspapers
all across this country.
Mr. Miller. I think we've already accepted that.
Mrs. Maloney. Thank you.
[The prepared statement of Hon. Carolyn B. Maloney
follows:]
[GRAPHIC] [TIFF OMITTED] T1389.005
[GRAPHIC] [TIFF OMITTED] T1389.006
[GRAPHIC] [TIFF OMITTED] T1389.007
Mr. Miller. Mr. Ryan.
Mr. Ryan, before you begin, you mentioned the editorials.
There's some very large number of editorials opposed to the
concept of statistical adjustment and manipulation, and I ask
consent that we enter those in the record, and without
objection, those editorials will be included.
[The information referred to follows:]
[GRAPHIC] [TIFF OMITTED] T1389.008
[GRAPHIC] [TIFF OMITTED] T1389.009
[GRAPHIC] [TIFF OMITTED] T1389.010
[GRAPHIC] [TIFF OMITTED] T1389.011
Mr. Miller. I am glad you talk about getting away from the
partisanship, but it's something you make a statement and it
really bothers me to set claims, Mrs. Maloney, that we don't
want to count people. This Congress has provided every penny
and provided all the resources the Bureau has asked to get
everybody counted. That is our objective, that is our goal; and
we are not trying to not count people, and to say that is just
political rhetoric. So I just want to make sure we clear the
record.
Let me clear one other record and that is the question
about scientific endorsement. The National Academy of Sciences
panel to review the 2000 census has not endorsed the ACE. I had
a long meeting with Janet Norwood only 2 days ago, and she
emphatically stated several times that neither she nor any
member of the panel has made any determination as to the
quality or outcome of the ACE. She explained what is quite
obvious to most people, that you can't evaluate a statistical
process if, one, it's not complete and, two, you don't have the
data. So please stop misrepresenting the truth here. The ACE is
a statistical plan, but at the moment it is mostly just that, a
plan.
Mr. Ryan.
Mr. Ryan. Thank you, Mr. Chairman.
Mr. Chairman, let me followup on this. We are going into
that touchy part of the census where I think we have done very
remarkable accomplishments in the enumeration process, and I'm
excited about hearing more about how well the enumeration is
working, but now we are getting into that touchy area and now
we are hearing the kinds of discussions, the kind of political
rhetoric that is unfortunate.
First of all, it's not all Democrats against all
Republicans. Democrats in my home State of Wisconsin are
against sampling because they know it is not good for the State
of Wisconsin. So I object to the characterization that
opposition to the ACE plan means opposition to counting people.
The same people who oppose the ACE plan, the chairman and
myself, are the same people who have provided $7 billion to
improve the census, especially in the traditionally
undercounted areas and communities in the inner cities, $7
billion provided for advertising, $7 billion provided for
partnership, hiring that far exceeded any previous census. So
saying that the people who provided these resources don't want
the people to be counted is wrong and is actually racially
divisive.
Post-census local review, we passed that out of Congress.
We can't get it signed into law. Post-census local review, in
my opinion, is a very good idea. It simply says in those hard-
to-count areas, the inner city of Milwaukee, the inner city of
New York, please look at the data, local, elected officials,
tell us if we missed anybody. Did we miss a public housing
complex? Did we miss a neighborhood that's tough to reach? Did
we miss a Latino neighborhood that didn't want to comply with
the census? If so, we'll go back and get those people counted.
That's a very common-sense idea.
It's a common-sense idea that was supported in 1990 by
Democrats and Republicans alike. The mayor of Chicago, mayors
all across the country supported a common-sense idea like post-
census local review. We passed post-census local review to try
and improve the count, to try and make sure that those people
who are historically undercounted get counted, that local
officials, mayors, county board members, city council people,
pastors in inner-city churches get a chance to look at the data
before it's finalized to make sure that their citizens weren't
missed. Well, this administration blocked post-census local
review. We don't have post-census local review.
LUCA was a good idea. LUCA worked well, but it can be
improved upon. I still think we should do post-census local
review. So to suggest that those of us who have questions about
the statistical adjustment are somehow against getting the most
accurate census is a ridiculous claim, No. 1.
No. 2, the scientific community is clearly not unified on
this point, so also to suggest that the scientific community is
completely behind statistical sampling is not correct. I have a
letter here from the Statistics Department of the University of
Berkeley in California, Dr. David Freeman and Kenny Walter. I
ask unanimous consent that this letter be inserted into the
record, Mr. Chairman.
Mr. Miller. Without objection.
[The information referred to follows:]
[GRAPHIC] [TIFF OMITTED] T1389.012
[GRAPHIC] [TIFF OMITTED] T1389.013
[GRAPHIC] [TIFF OMITTED] T1389.014
[GRAPHIC] [TIFF OMITTED] T1389.015
Mr. Ryan. Also, the National Academy of Sciences has not
endorsed the ACE plan. So to suggest that the scientific
community believes that this is a unified point, that's just
not the case.
Sampling didn't work in 1990. We found that years later we
had dire problems. So one thing that I think in today's
hearing, hopefully we can get into, is the compressed time line
that this plan involves. I am concerned that this rushed time
line is going to give us the errors we will discover down the
road when it's too late.
So with that, Mr. Chairman, I'd like to yield back the
balance of my time. My friend from New York, the national
academy of scientists hasn't officially endorsed the ACE plan.
The scientific community is divided on this. So I hope that we
can move forward on an even keel, on an objective basis; and I
hope that we won't get into this heightened political rhetoric
where we are impugning the motives of each of the two parties
involved.
All of us want an accurate count. All of us want everybody
to be counted in the neighborhoods where they live. Democrats
and Republicans in Wisconsin want that. All of us want that. So
let's keep the discussion at that level if we may.
With that, I yield back the balance of my time.
Mrs. Maloney. I request permission to respond. I think that
I should be able to respond.
Mr. Miller. Mr. Davis.
Mr. Davis of Illinois. I yield time, Mr. Chairman.
Mrs. Maloney. Very briefly I feel that we should let the
facts speak for themselves, so I would like to put into the
record the legislation, bipartisan, that went--that passed
calling on the National Academy of Sciences to come forward
with a plan to correct the undercount. The plan that they came
back with, which was the use of modern scientific methods, I
would like to put into the record the language that the
Republican majority attached to the disaster relief bill that
would have prohibited the use of modern scientific methods. I
would like to put into the record that the Republican majority
tried to attach to two budget bills, holding up two budgets,
that would have limited and prohibited the use of modern
scientific methods.
The facts speak for themselves, and I will put that in the
record, and it is clear--and it is clear what the intention of
such actions would do and how it would affect and continue an
undercount. Knowingly, the majority tried to stack the deck so
that millions of Americans would be intentionally left out of
representation and funding dollars in this country. It is
unfair. It is unjust. It is the civil rights issue of this
decade.
Mr. Ryan. Will the gentlelady yield for an honest point of
clarification--not a tit for tat, just an honest point of
clarification?
Mr. Davis of Illinois. Reclaiming my time, I yield.
Mr. Ryan. Thank you, Danny. Appreciate it. That was 2 years
ago in the last Congress. That was then; this is now. Let's
move forward with not a lot of political rhetoric. Let's move
forward and debate this objectively, and let's not impugn the
motives of each other. We all want an accurate count. With
that, I yield.
Mr. Davis of Illinois. Reclaiming my time, Mr. Chairman,
let me thank you for convening this hearing regarding oversight
of the 2000 census and the impact of the accuracy and coverage
evaluation, that is, the ACE process. As enumerators begin the
process of going door-to-door to those households that did not
send in their census forms, it is important that we examine the
ACE operations. The ACE process was added to the 2000 census to
replace the post-enumeration survey of 1990 in an effort to
improve the accuracy of the census.
We all know that accuracy must be the goal. We can ill
afford to go back to the days of 1990 when too many people lost
from an inaccurate census. The constituents of my district, the
Seventh Congressional District in Illinois, deserve and need an
accurate count of the entire population. They realize that too
much is at stake to get a less-than-accurate count.
In 1990 Chicago lost millions of dollars in Federal funds
because of a census undercount. According to the Bureau, at
least 10 people, including at least 113,831 in the State of
Illinois, 81,000 in Cook County and 68,000 in the city of
Chicago were not counted in the 1990 census. Many of those
missed were obviously children and women who live in minority
communities. Because the 1990 census missed counting millions
of people in Chicago, every one of our residents were
shortchanged on money to repair roads and streets. They were
shortchanged on money for mass transit and senior citizen
homes. They were shortchanged on money for schools, parks and
job training. Perhaps the most egregious shortchange was that
of political representation, and in a democracy representation
is essential to having a voice in local, State and Federal
Government.
I represent many hard to count people. According to the
Census Bureau 165,000 of them live at or below the poverty
level in my district. I'm pleased that we're holding these
hearings in an effort to make certain that the Census Bureau
and others are doing everything possible to get an accurate
count. Yes, many people are indeed difficult to count.
Therefore, we must use every effort to try and make sure that
the past evils and transgressions of our Nation do not continue
to negatively impact upon the reality of our being, and if
there is to be fairness, we must indeed make use of every
method available to us.
Just 2 days ago the mayor of my city, Mayor Daley announced
a $400,000 radio and television advertising campaign to be
funded by the city to encourage people to cooperate and
participate in the 2000 census. This advertising campaign
coupled with what the Census Bureau has already committed to
will go a long way toward a more accurate census. However, I
tell you that you cannot undo with radio and television ads
what 400 years of slavery, deprivation, discrimination, denial
of equal opportunity, lack of opportunity to be educated, to
understand, to be a part of the mainstream, you cannot undo
that with radio, television and newspaper advertising. You
cannot even undo it by sending people door to door, looking for
people that you cannot find, people who in many instances are
unreachable, untouchable and unfindable, and I don't care how
much we put in, unless we make absolutely certain that every
available technique, every scientific advancement, every
opportunity exists to count every single person, account for
every single person in this country, then we once again will
come up lacking. Once again, individuals will be left out. Once
again, individuals will be shortchanged and once again, this
Nation will have shortchanged itself.
So, Mr. Chairman, I am pleased that we're having this
hearing and look forward to the information that Dr. Prewitt
will share with us. So I thank you and yield back the balance
of my time.
Mr. Miller. Thank you, Mr. Davis. You do outline the real
challenges of this massive undertaking that we're in the
process of. Dr. Prewitt, if you and Mr. Hogan and Mr. Thompson
would stand and raise your right hands, we'll get you sworn in
and proceed, in case Mr. Thompson and Mr. Hogan is needed to
assist you.
[Witnesses sworn.]
Mr. Miller. Thank you and please be seated. The record will
reflect that Mr. Thompson, Dr. Hogan, Dr. Prewitt answered in
the affirmative. Welcome. Thank you. You may proceed with the
opening statement.
STATEMENT OF KENNETH PREWITT, DIRECTOR, BUREAU OF THE CENSUS,
ACCOMPANIED BY JOHN H. THOMPSON, ASSOCIATE DIRECTOR FOR
DECENNIAL CENSUS; AND HOWARD HOGAN, STATISTICIAN, CHIEF,
DECENNIAL STATISTICAL STUDIES DIVISION
Mr. Prewitt. Thank you, Mr. Chairman, and I did solicit the
chairman's permission this morning to spend just a few minutes
returning to the e-mail incident that was addressed last week
at the subcommittee hearing with the GAO, and I have also
informed the minority and also the leader and also Mr. Davis
and Mr. Ryan that I would like to address that quickly.
I'd like to start by saying that it is understandable that
in the absence of facts the offending sentence instructing the
LCO managers not to share a given report with the GAO could
have led to the strong reaction of the chairman, of Congressman
Ryan and of Mr. Mihm of the GAO. But the facts do in fact
mitigate this reaction and I would like to quickly put them in
the record, and here I paraphrase from a subsequent e-mail by
Mr. Rodriguez, who was the author of that initial e-mail.
The report in question, he explains is, a regional level
report and information from it should be shared only by the
regional level. It in turn generates local office reports, and
it is this information that can be shared by local managers. As
he writes in a subsequent e-mail to us, that, as per our
instructions on May 8th at about 3 p.m., the report, the
offending report--not the offending report but the one that
initiated the incident--was to be shared at the area manager's
discretion with their local census office but any one local
census office was not to share the production information of a
different LCO; that is, each LCO was only to share its own
information and not other patterns of information. If anyone
outside the Census Bureau wanted to obtain a regional level
report, they can get that from the regional census center, and
then he goes on to explain why he sent the report he did: ``My
intentions were to provide the report to my offices as a tool
to encourage friendly competition and thereby productivity,
nothing more. I regret that my intentions have been
misinterpreted and rightly so because of the way my e-mail was
worded, and I apologize for any inconvenience I may have
caused.''
I would like to say, Mr. Chairman, that obviously I join in
that apology but nevertheless do want to make certain that we
understand that the facts themselves give no warrant for the
accusation that the Census Bureau is preventing the GAO from
discharging its responsibilities.
Mr. Chairman, we have taken this accusation so seriously
that yesterday with Deputy Director William Barron I met with
David Walker, the Comptroller General, with Nancy Kingsbury and
with Chris Mihm to reiterate the Census Bureau policy in regard
to access, and I do believe that we all fully understand that
there's nothing in our policies that are designed to prevent
any access by the GAO.
Mr. Chairman, more than a week ago at the subcommittee
hearing to which I've referred, you said that there were Census
Bureau employees, ``in very influential positions who are
dangerous.'' This I take as a very serious charge. If
substantiated, I would take corrective action. Obviously if I
am, myself, the person who is dangerous, then I would expect
you to bring that to the attention of the Secretary and he
would take corrective action.
In that same hearing, you asked the GAO to investigate the
incident that led you to make this extremely serious charge.
Yesterday I asked David Walker, Comptroller General of the GAO,
if his organization had any evidence that would corroborate
your charge that the Census Bureau has people in very
influential positions who are dangerous. He replied in the
negative, and in this he was seconded by Christopher Mihm, who
also was present at the meeting.
Mr. Chairman, it is now more than a week after your charge.
I know that you and your staff had conducted your own
independent investigation, and I wait for the evidence on which
this accusation is based, for I am unable to take corrective
action until I know who these people are and what it is that
makes them dangerous. So may I respectfully request that you
please provide me, as soon as possible, the names of the
dangerous people, the nature of the danger they pose and of
course the evidence that would substantiate this charge, and I
promise to you that I will take corrective action. Thank you,
sir.
Now, if I may, I'll turn to my opening comments on the
topic of this hearing.
Census 2000 operations continue on track and on budget. I
earlier reported that the mail-back response rate at 66 percent
was very encouraging, and in my written testimony I indicated
that we had completed 39 percent of the nonresponse followup
workload. I would like to update that number through yesterday.
The workload is now 50 percent complete. Putting these two
numbers together, the census enumeration is now approximately
85 percent complete. That is, 85 percent of the housing units
have now responded or we have identified them as vacants.
Still, the Census Bureau does not anticipate at this point
that census 2000 will have better coverage than the 1990 census
because many of the factors that led to the undercount in 1990
are still present in American society, and indeed, as a
proportion of the population have grown--more gated
communities, more recent immigrants, more linguistically
isolated households, more persons living in irregular housing
and perhaps more anger toward the government.
The Census Bureau has both measured and documented the
existence of a substantial undercount since 1940, and this has
already been referenced in the opening comments. The Census
Bureau has been running harder but believes this will only
allow us to stay even. That is, we expect that neither the
overall coverage levels nor the differential undercount rates
in census 2000 will show improvement over 1990. The Census
Bureau strongly hopes to be proven wrong in this assessment,
and the ACE will give us the information to determine whether
this is so.
The ACE provides a final quality check on how well we have
done in the initial census. The alternative is not to do the
ACE and never know how we have done below the national level
where demographic analysis does provide a benchmark. The ACE
also provides the means to generate a more accurate count.
The 1990 version of the ACE, or the accuracy and coverage
evaluation, was called the post-enumeration survey. It provided
information that was used during the 1990's to improve
statistical programs. The population estimates the Bureau of
Labor Statistics asked us to incorporate into the current
population survey program following the 1990 census were
corrected for the undercount identified through the 1990 PES.
The Bureau of Labor Statistics also requested adjusted
population controls for the consumer expenditure survey. All
other major national demographic surveys conducted by the
Census Bureau or other agencies in the Federal statistical
system also were converted to this adjusted population base.
And Katherine Abraham, the current Commissioner, of course
testifies that in the absence of this correction their
published demographic distribution of unemployment and other
measures would have been inaccurate.
We believe we have an obligation to the Bureau of Labor
Statistics and the many, many other users of our data products
to make our data as accurate as possible. I have said
previously that the Census Bureau currently expects that the
corrected numbers using the accuracy and coverage evaluation
will be the more accurate numbers. If the Census Bureau does
not have confidence in the results, we will not use them. The
decision whether to release the statistically corrected data
should take into consideration operational data to validate the
successful conduct of the ACE, whether the ACE measurements of
undercount are consistent with historical patterns of
undercount and a review of selected measures of quality.
In the fall of this year the Census Bureau will discuss the
review process and criteria with the statistical community and
other interested parties. We will set forth how we will assess
whether our operational functions for the ACE were met. All
major operations have been designed and documented and the
details have been available for review and comment. Every
document requested by the subcommittee has been forwarded.
Here, however, Mr. Chairman, is a complete set. It is possible
there are documents here that you have not yet requested, but
we can provide you the entire complete set of our decision
documents that go into the design of the ACE.
Now, let me very quickly try to describe the operations as
requested in your invitation letter. Several major operations
have now been completed. One is ongoing and others will follow
the completion of nonresponse followup. All operations are
currently on schedule.
The basic concept behind the ACE is the comparison of the
data from two systems, an independent survey and the initial
census. Because of its small size relative to the initial
census, we believe we can do a better job enumerating people in
the housing units in a sample. We can be more selective about
the interviewers, train them longer, pay them more and provide
more quality assurance.
The first step in the ACE process is to design and select a
sample which consists of approximately 314,000 housing units or
about one-fourth of 1 percent of the total housing units. The
basic units of this sample are what we call block clusters, and
there will be about 11,800 block clusters in our sample. The
sample was designed and selected to provide sufficient
precision to estimate the true population for various groupings
of the population that we call post-strata which I will
describe below.
The next step in the process is to create an independent
listing of housing units. By independent we mean that we do not
start with or refer to the master address file from census 2000
but instead have census staff systematically canvass the block
clusters to list the addresses. This operation was completed in
the fall of 1999, checked and keyed and 100 percent quality
controlled in our national processing center.
The Census Bureau then matches this list of housing units
to the master address file, first by computer and then
clerically if necessary, using this additional information, we
continue to improve our address list. The purpose of this
housing unit match is to create an accurate linked list of
housing units in the block clusters. This work was also
completed on schedule.
To provide sufficient data to compare the ACE to the
initial census the Census Bureau of course must conduct
interviews to collect data from each of the housing units that
were independently listed. We initiated the ACE interviewing
with a telephone phase using laptop computers in a technique we
call computer assisted personal interviewing. This is the
laptop computer that we're using that in that procedure, and we
would be delighted, of course, to provide a staff briefing of
how it is used.
We began in late April telephoning households in the ACE
sample at unique addresses for which a census 2000
questionnaire had been mailed back, processed through data
capture and for which a telephone number was provided. As of
today we have completed over 60,000 interviews by telephone,
more than 20 percent of our workload for the ACE. In addition
to getting an early start on interviewing, the benefits include
providing experience for our supervisors and a final testing of
our automated system. The Bureau, of course, has had extensive
experience with telephone interviewing. We designed this phase
of the ACE based upon our testing of the methodology in the
dress rehearsal.
As you know, we do not begin personal visit interviewing
until nearly all nonresponse followup work is completed in all
the ACE block clusters in an LCO. This is one of the ways we
preserve independence between the ACE and the census. If
nonresponse followup and ACE, field interviews are working the
same areas simultaneously, they could affect each other's work,
and that's why we wait to complete nonresponse followup before
starting the personal visits.
Interviewers, whether on the telephone or personal visit,
focus on reconstructing the Census Day household; that is,
determining who lived at the address on Census Day at the time
of the ACE interview and collecting as much information as
possible for those who lived at the address on Census Day but
have moved out, so we also have special procedures, of course,
for movers.
All of these interviewers will use the CAPI technique. This
is a technique that improves the accuracy of the operation
because it permits a more structured interview and more probing
questions. We have extensive processes for conducting quality
assurance to identify data quality or falsification problems,
though for data quality purposes we do not widely publicize
these processes. Most personal visit interviewing will be
conducted in late July or August, but some may begin in mid-to-
late June. Personal visit interviews are conducted only with a
household member during the first 3 weeks if the case is
available. If an interview is not obtained after 3 weeks,
interviewers will attempt to interview another knowledgeable
person, and during this latter part we use our very best
interviewers of course who are trained to convert reluctant
respondents.
We then do person matching. This will occur in October and
November. Census Bureau staff conduct the various stages of the
matching of persons listed in the ACE interviewing to those
persons counted in the same block clusters as part of the
initial census, and we have designed, of course, these matching
processes to minimize errors. Incorrect matching determinations
generally result from incomplete, inaccurate or conflicting
data or from poor judgment, and so we have several stages at
which we conduct this matching process each with its own
quality assurance process.
We then turn to dual system estimation. We use data from
the ACE and the census to estimate the true population using a
statistical technique called dual system estimation. The DSE
will be conducted for each of over 400 groupings of people or
post-strata. The dual system estimator of true population is
then used to calculate a coverage correction factor for each
post-stratum, which is the ratio of the DSE to the initial
census count. The variables that define the post-strata
grouping include race, ethnicity, age, sex, owner and nonowner,
return rates, whether in or out of a metropolitan area and, if
in, the size of the area, the type of census enumeration
method.
These are characteristics that our research indicates are
correlated with a likelihood of inclusion in the census. An
example of one post-stratum is non-Hispanic Black males age 18
to 29 in nonowner units in mail-out mail-back areas of
metropolitan areas with 500,000 or more people in a tract with
a low return rate in the census.
Coverage correction factors are then applied the census
files. For example, if the coverage correction factor for a
non-Hispanic Black male in the specific post-stratum described
above is 1.02, this means the Census Bureau measured an
undercount of 2 percent for this post-stratum and for every 100
people counted in the census in these areas two records will be
added. This process is sometimes called synthetic estimation.
After this, the corrected census file can be used to produce
the corrected tabulation for all uses of census data.
Mr. Chairman, I have tried to give a rather simple and
quick basic description of the ACE and the documents listed in
the appendix, and, of course, this fuller set of documents can
be investigated for any more detailed questions that you might
have. Thank you sir.
[The prepared statement of Mr. Prewitt follows:]
[GRAPHIC] [TIFF OMITTED] T1389.016
[GRAPHIC] [TIFF OMITTED] T1389.017
[GRAPHIC] [TIFF OMITTED] T1389.018
[GRAPHIC] [TIFF OMITTED] T1389.019
[GRAPHIC] [TIFF OMITTED] T1389.020
[GRAPHIC] [TIFF OMITTED] T1389.021
[GRAPHIC] [TIFF OMITTED] T1389.022
Mr. Miller. Thank you, Director Prewitt. I'm not going to
enter all that in the record. It will make it too long and
lengthy. So I will appreciate having the access to that. Let me
briefly make a comment about the issue of transparency and
openness, and I hope like you that we get beyond this very
quickly.
However, let me just explain the foundation for my
concerns. Issues of transparency and access go right to the
heart of one of the reasons of why I called this hearing today.
The census is like a business placing an order. Last year the
Congress placed a $7 billion order with the Census Bureau. This
Congress had done everything it can to make sure we had the
money to pay. Now the inventory is coming in, and I am equally
responsible to those Senators and Representatives and the
people they represent to make sure we get what we paid for.
It's my job to check the inventory.
It's also the job of the GAO, the monitoring board and the
Inspector General and the National Academy of Sciences to
review the 2000 census. Unfortunately, when I tried to check
the inventory, the Bureau tells me I can't open certain boxes
or I have got to wait 3 weeks while they check with
headquarters or there are some boxes that are off limits. Let
me give you a few examples.
Last year the Bureau refused to provide this subcommittee
with data from the 1990 census. They claimed it was protected
by title 13 and it wasn't. And we are all sworn anyway, but the
result was to delay the request for months.
The Bureau also delayed providing information requested for
the dress rehearsal for an entire year, effectively preventing
analysis. Two months ago, I entered into the record a list of
information requested by the monitoring board that were delayed
by more than 60 days or refused. The Bureau also produced a set
of guidelines that limited access to local offices by the GAO,
the board and the subcommittee. And just last week we received
a copy of an e-mail that gave me, my staff and representatives
of the GAO reason to believe a Bureau employee was instructed
to withhold information and was instructing subordinates to do
the same.
This week I received a letter from the Director requesting
that I not call any Bureau employees without a Democratic
staffer or a member of the Bureau present. Is this what is
meant by transparent census free from political manipulation?
If this is the routine during the relatively simple census,
delayed information, limited access and obstructed
investigations, how can we have confidence in the extremely
complex statistical adjustment?
How can we honestly say this process is free from political
manipulation if we are not allowed to review the process, or if
we are only allowed to look at certain parts of it under
certain conditions with proper supervision of Democrats and
Bureau employees?
These developments are increasingly troubling and do not
add to the credibility of this or future decennial censuses.
My comments certainly do not reflect on you, Director
Prewitt, and I think the people behind you. The nature of the
concern is there is a contempt for Congress and the
responsibility that we have, as the elected officials have for
overseeing, not only want the $7 billion of the census, but the
critical role the census is for our entire electoral process.
It is rare that we have a copy of an e-mail like that and
it is legitimate for us to be looking at that.
Let me make another comment about the use of the PES and
the BLS adjusted numbers. My understanding is that the BLS
accepted adjusted numbers for very large areas, national
populations, and for large States. But you are proposing
releasing adjusted numbers for every State, county, city, and
block in the country. That is a completely different use of the
numbers than what BLS is using. At small levels of geography,
adjusted numbers are not reliable, and in fact the Census
Bureau doesn't use the adjusted numbers. Which raises a
question of I see in your statement instead of using
``adjusted'' numbers you started using ``corrected'' numbers,
which implies that you have already decided that the
adjustments are the correct ones, which to me almost
politicizes the use of that word. So I am concerned that you
start saying, well, this is the corrected one. You have
obviously made a decision--not obviously but apparently instead
of calling it adjusted numbers you are using corrected numbers
and that is a political way to refer to those numbers.
Dr. Bryant, the Director of the Census Bureau in 1990,
originally supported adjusting the national numbers. But she
decided not to adjust the intercensual estimates after
extensive evaluation by the CAPE Committee showed the 1990
adjustment was 45 percent error. Let me read from her decision,
this is from the CAPE report: ``Work suggests, the CAPE
committee's work suggests that no survey, either the high
quality, well-controlled and interviewed 1990 PES of 170,000
households or a larger one can be used to make a post census
fine-tuning of an average undercount as small as 1.6 percent in
all types of places, counties, and States. Given that, from
little or no evidence that adjustment would improve the quality
of sub-state estimates other than for a limited number of large
places, the decision is not to adjust.''
This is from the December 1992 Federal Register, which I
would like to enter into the record, and without objection it
will be included in the record.
[The information referred to follows:]
[GRAPHIC] [TIFF OMITTED] T1389.023
[GRAPHIC] [TIFF OMITTED] T1389.024
[GRAPHIC] [TIFF OMITTED] T1389.025
[GRAPHIC] [TIFF OMITTED] T1389.026
[GRAPHIC] [TIFF OMITTED] T1389.027
[GRAPHIC] [TIFF OMITTED] T1389.028
[GRAPHIC] [TIFF OMITTED] T1389.029
[GRAPHIC] [TIFF OMITTED] T1389.030
[GRAPHIC] [TIFF OMITTED] T1389.031
[GRAPHIC] [TIFF OMITTED] T1389.032
[GRAPHIC] [TIFF OMITTED] T1389.033
[GRAPHIC] [TIFF OMITTED] T1389.034
[GRAPHIC] [TIFF OMITTED] T1389.035
[GRAPHIC] [TIFF OMITTED] T1389.036
[GRAPHIC] [TIFF OMITTED] T1389.037
[GRAPHIC] [TIFF OMITTED] T1389.038
[GRAPHIC] [TIFF OMITTED] T1389.039
[GRAPHIC] [TIFF OMITTED] T1389.040
[GRAPHIC] [TIFF OMITTED] T1389.041
[GRAPHIC] [TIFF OMITTED] T1389.042
Mr. Miller. Director Prewitt, if you plan to release
adjusted numbers at block level, please be prepared to defend
the accuracy at that level. Don't tell me that because some
agencies have elected to use adjusted numbers at the national
level we should adjust the population for all 6 million blocks
across the country.
A part of your statistical design hinges on using race, age
and other characteristics to characterize people on what you
call post-strata. For example, one category would be Asian
women living in small metropolitan areas, age 30 to 49, renting
their living space. And while the exact number has changed
several times, I believe the current design is made up of 448
such categories; is that correct?
Mr. Prewitt. Yes, sir.
Mr. Miller. In these post-strata or profiles, when you say
Asian, do you make distinctions between Japanese, Chinese,
Laotians, Koreans or other Asian cultures?
Mr. Prewitt. No, sir.
Mr. Miller. Do you make a distinction between Japanese,
Chinese, Laotian, Koreans or others, or is it all just Asian?
Mr. Prewitt. It's all Asian.
Mr. Miller. On Hispanics, do you lump confidential Cuban
Americans, Puerto Rican Americans and Mexican-Americans?
Mr. Prewitt. Yes, sir.
Mr. Miller. And they are all Hispanics?
Mr. Prewitt. They're all Hispanic.
Mr. Miller. And so the assumption is they all respond in a
similar manner?
Mr. Prewitt. Correct. That is the assumption, yes, they
have similar capture probabilities, is how we would put it.
Mr. Miller. Well--so, the Cubans in Miami, the Puerto
Ricans in New York, the Hispanics in Houston or Los Angeles all
have the same characteristic response rates? Guatemalans,
Hondurans in Miami, they all have the same as the Cubans? I
find that hard to understand and grasp.
I think if you talked to the Cubans in Miami or the
Mexicans in Los Angeles, they may not totally agree with that,
that they are all homogeneous as you assume.
Mr. Prewitt. No, I said they all have the same capture
probability, which is to say the chances of enumerating them in
the basic census are roughly similar. Now, let us make certain
that we understand that we're not just talking about Cubans or
Hispanics, we are also talking about 17 or 16 other sets of
characteristics. Do they own? Do they rent? What is their age?
What is their gender? So it is not simply the ethnic or racial
characteristics. It is a cluster of characteristics that create
a post-stratum.
Mr. Miller. As I understand, the post-strata design
adjustments for various categories will be applied the same way
across the entire Nation. For example, Hispanic men age 18 to
29 in large cities in rental housing in areas with high mail
return rates. As I read that, it sounds like as long as they
fit that description, Cubans living in Tampa will get the same
adjustment as Mexicans living in Houston and Puerto Ricans
living in New York and so on across the country. That is
correct?
Mr. Prewitt. Yeah, it turns out they have very similar
capture probabilities. That is why it is correct.
Mr. Miller. Let me ask you about Puerto Rico. You have 84
post-strata classifications for Puerto Rico; is that right? For
Puerto Rico, you have 84?
Mr. Prewitt. Checking.
Mr. Miller. OK. OK. My understanding is there was 84. Does
Puerto Rico have the separate strata but Texas, Chicago,
California and New York are treated separately? How is that?
Why don't we have separate classifications of post-strata for
Texas, California, Chicago or New York City, but we have all of
these strata classifications for Puerto Rico? Why is Puerto
Rico singled out? I mean, they are all Hispanic or most all
Hispanic. But then----
Mr. Prewitt. Before we turn to Puerto Rico, let me address
the first part of your question. Can I just go back a bit? You
have asked a large number of questions to get to this very
particular one. And I don't want to readdress all the access
questions, but I do think that nearly every one of those things
that you mentioned in your response on the access question have
been answered before and continue to be answered. We talked
about a terrabyte of information, 52 million yellow pages worth
of information. We have met repeatedly with the GAO and the
monitoring board. There are no access questions, there are no
transparency questions that I know to be on the table right
now, sir.
And you chide us about having guidelines. On the other
hand, you were the one who asked me in a hearing to please
create some guidelines so that we could all sort of understand
and move forward in this census without disrupting it. So I
find it a little odd that now we try to have some guidelines
and we are chided for that. So it is kind of either way.
Mr. Miller. It is interesting that you have given me
guidelines on my behavior. I mean, my letter that you received
the other day is that you are telling me who I can call and who
can be present in a phone call. I'm the elected representative
of the people.
Mr. Prewitt. Of course, you can do whatever you want and
will do whatever you want. We then get pressure from the
minority wanting to make exactly the same calls. I don't know,
I don't know whether you want to call 100 people and put them
on tape, or 10 people, or 200 people.
And we're just trying to sort of manage the process right
now. We're trying to finish the census. It seems kind of
reasonable to us to say, if possible, let's try to coordinate
these phone calls. But no, you can call anyone you want to in
the country. We will give you the phone numbers of 500,000
employees.
Mr. Miller. By the way, it was very easy to locate the
gentleman in question and he was a very pleasant call.
Mr. Prewitt. I am also very pleased that you made reference
to the fact this is a rare e-mail. Of course it is rare. That
is why you had to pay such attention to it because there is not
a pattern of such e-mails.
But back to the ACE. You said in your opening comments that
the Bureau of Labor Statistics used the data at the national
level and State level and substate levels, but primarily you
are absolutely correct. I would just like to make certain that
you understand that the Census Bureau believes that local area
data are unstable irrespective of whether they are corrected or
not.
We think that the apportionment number, which is based upon
the basic enumeration, is unstable at the local level. We don't
have a whole lot of confidence in block level data, period,
however they are collected, because it is the nature of very
small area data that that is where errors can get magnified.
So we are just as worried about block level data
preadjusted or precorrected as we are post adjusted and post
corrected. It is just a fact of the nature of statistical
operations.
So when you say that we have to be absolutely correct at
the local level, we would--if that were the standard, we would
not be able to give this country the redistricting data based
upon the initial census, because we couldn't stand behind that
data at every local level. We just couldn't do it. So it is not
an issue of whether it is adjusted or unadjusted----
Mr. Miller. Which is more correct--which is more accurate
at the block level, the actual count or the adjusted, or you
want to call it the corrected number at the block level? Which
is more accurate?
Mr. Prewitt. Oh, undoubtedly the adjusted number is more
accurate across----
Mr. Miller. At the block level?
Mr. Prewitt. Oh, yes, absolutely. At the block level we are
missing--we know in certain blocks in Chicago, heavily
comprised of African-Americans who rent their housing who are
young males, we're missing a large number of them.
Mr. Miller. You have already decided then that the
adjustment is going forward and that is going to be more
accurate, the adjusted numbers at the block level. Why did the
Census Bureau not use the adjusted numbers for the intercensual
estimates? Why did the BLS use it only for the national numbers
or very large State numbers and not for--and we have all used
the unadjusted numbers? Now you are saying it is more accurate
or corrected, the political term you are using.
Mr. Prewitt. I am sorry that you are concerned about that.
We use the word ``adjusted'' and ``corrected'' interchangeably.
Mr. Miller. ``Corrected'' is a new use of the word, I
think. You have been using ``adjusted,'' and I think that is
more appropriate.
Mr. Prewitt. Well, I have been using the word ``corrected''
since I got here. I am happy to use the set of terms you want.
They are interchangeable as far as we are concerned.
The decision has not been made, sir. The decision that has
been made is to proceed with the ACE procedure. And indeed, as
I said, we are now over 20 percent finished with the collection
of the data.
We have theoretical reasons for believing that this will
produce more accurate numbers. We also have to test the
operations. Your concern about whether this is a spaceship to
Mars that may blow up is an understandable concern. That is
also true of the census. It is also true of the enumeration.
Any one of our operations could have blown up. We are very
pleased that we are now about 85 percent finished with the
census and it hasn't blown up. And we've had many, many
hearings about why that is so. We are really very pleased with
the operational robustness of this census.
But it is not in the nature of one operation versus another
operation that it can turn into difficulties. Any operation
will run into difficulties, including the new operation--not a
new, but an operation that we have not talked about in this
subcommittee called coverage improvement followup, which is
going to have 7\1/2\ million households in it. We have not done
that one yet. It may not work well. If it doesn't work well,
that will have an impact upon the quality of the apportionment
numbers.
So it is not something special with the ACE which makes it
vulnerable. Any big complex field operation is vulnerable. The
good news is we're 85 percent finished with the census without
having had one. We still have 15 percent to go and it is a
very, very hard 15 percent because we are now down to the
difficult cases.
So we have made the decision that based upon statistical
theory, capture-recapture technologies are able to improve a
basic count. That's the decision that we have made and we have
designed an operation to do that.
Mr. Miller. Well, I think we are all pleased that the full
enumeration is proceeding as it is. I think from the mail
response rates to the nonresponse followup is proceeding
apparently ahead of schedule. And that is the positive thing.
But ACE, there is legitimate differences within the
statistical community, as you are well aware of, and to say
that this is already going to be the corrected number, and the
other one is an incorrect number by inference--when you say one
is correct that means the other one is incorrect--is
politicizing the process, and that is unfortunate.
Mrs. Maloney.
Mrs. Maloney. Thank you very much. I feel that the chairman
often will invoke a person's name who is not here and use it as
evidence and they're not here to speak for themselves. He did
it last week with Mr. Rodriguez, a Marine, the civil servant,
the young man who was working in the census office. He wasn't
here to speak for himself. Now we have his written letters. We
know what he said. But earlier he mentioned the names of Dr.
Barbara Bryant, the Census Director under former President
Bush, and let's have her come here and speak for herself on how
she feels about modern scientific methods.
And he mentioned the name of Dr. Janet Norwood and his
conversation with Dr. Janet Norwood from the National Academy
of Sciences, implying that she did not support the plan of the
agency. Well, may I suggest, respectfully, that we invite Dr.
Norwood to come and speak for herself. I have a May 3rd, 1999,
letter from Dr. Norwood that I would like to put in the record
that appears to indicate that she supports the plan of the
agency, ``In general, the panel concludes that the ACE design
work to date is well-considered. It represents good, current
practice in both sample design and post design as well as the
interrelationship of the two.'' And I'd like to put that in the
record.
So very respectfully, I suggest that we have these people
come and enter their own testimony as opposed to an
interpretation by the chairman.
And I would like to ask you, Dr. Prewitt, have you had any
conversations with Dr. Norwood? And what is your interpretation
to date on her support of the agency's plan?
Mr. Prewitt. Well, Dr. Norwood is of course the chairman of
the standing committee now of the National Academy of Sciences
that is looking over all of our plans for the accuracy and
coverage evaluation as well as the basic census, so obviously I
go to all of their meetings that are publicly open. I am
frequently asked to testify or to present materials before that
committee, so yes, I have attended every one of the meetings of
this committee and have had conversations with Ms. Norwood in
that context.
I believe that to try to get the facts exactly on the
table, the letter to which you refer, which is of course over a
year old, written May 3, 1999, was based upon the degree of
work that had been done to that date, and there has been a lot
more work done on the ACE design since that date. And that's
what is represented by this stack. That is--the size of the
stack about a year ago would have been, you know, a quarter to
a fifth of this size because we hadn't done a lot of the
technical work then. So based upon the technical work that had
been done, which was the early sampling design this is the
judgment that the committee wrote in that letter of May 3rd.
I think it is correct to say that the National Academy
committee has not, ``signed off on'' the full design because
they haven't met since the full design has been completed.
There are now 106 major decisions that have to be made with
respect to the ACE design; 104 of those have been made. The two
which have not been made have to do with weight trimming
methodology, and varience estimation, the specific criteria for
weight trimming methodology. And under variance estimation we
haven't fully finished talking through technically the specific
criteria for incorporating controlled rounding into generalized
variance estimation. Those are the only 2 out of 106 major
decisions that have not yet been completed.
You know, we're talking about those all the time right now.
In the next couple of weeks we will have resolved those. We
will put them in a piece of paper and they will join this stack
and they will be sent to this subcommittee if they want them
and also to the National Academy.
So I think it is fair to say that they haven't, ``signed
off on'' the final design because no one could have. The final
design did not exist on May 3rd nor indeed, as of yesterday
whenever the chairman talked to Ms. Norwood it did not exist.
So I think the National Academy committee has been extremely
useful to us, looking at our work, judging it, passing back
suggestions and recommendations to us, holding public events.
They had a major public event on the design earlier this year
and another one is scheduled for September.
So we don't expect them to have signed off on a design that
has not yet been completed. I think the chairman is quite
correct. They need to see it and they will see it as soon as it
is completely finished, and we are very close to having it
completely finished.
Mrs. Maloney. Dr. Prewitt, a number of people have
suggested that the use of statistical methods to correct errors
in a census opens the process to political manipulation. Would
you please explain to us whether or not you believe that to be
true, and what can be done to assure the public that no such
manipulation occurs?
Mr. Prewitt. Well, needless to say, there are few charges
that bother the Census Bureau as much as that one does. To my
knowledge, the decision memo by Secretary Mosbacher in 1991 was
the first time a senior official of the U.S. Government ever
put on record the possibility that the Census Bureau could
design a procedure in anticipation of it having a given
partisan outcome. And what he said is that the political
outcome of a choice, that is of a statistical procedure, can be
known in advance. He says: ``I'm confident that political
considerations played no role in the Census Bureau's choice of
an adjustment model for the 1990 census. I am deeply concerned,
however, that adjustment could open the door to political
tampering with the census in the future.''
This put on record the idea that the Census Bureau could
design something having a known partisan outcome. Let me just
say that this strikes me as ludicrous on the face of it. The
Census Bureau does not have the competence to predetermine
partisan outcomes. It has no statistical expertise in
reapportionment or redistricting, no expertise on trends in
voting behavior. To predetermine partisan outcomes the Census
Bureau would need to bring to bear such expertise when it
selected data collection methodologies several years in advance
of when the census counts are actually to be used for
reapportionment or redistricting.
It is simply way beyond the Census Bureau's competence or
capacity. Even if the Census Bureau intended to do it, it would
not know how to do it. It doesn't have the competence, it does
not have the interest, and it certainly does not have the
professional position that that is what this job is.
I would like to point out that there are a large number of
oversight agencies--this subcommittee, the Congressional
Monitoring Board, the Inspector General--and there are some
several dozen reporters who follow the census very closely.
There are public watchdog organizations. There are National
Academy committees. There are a large number of people who, if
they could find partisan manipulation of this census, would be
the first to report it. And I can only say we are now getting
toward the end of the census and no such incidence has ever
been revealed. Where is the evidence that the Census Bureau is
designing things to have a partisan outcome? What kind of
capacity would we have to have? We don't have it. We wouldn't
know how to do it. We don't care about it. We don't pay any
attention to redistricting data or voter trend data or what
Governor controls what State. We don't pay any attention to
that stuff, because we're actually trying to do a census and
that is the kind of capacity we have.
So I just think the charge that the Census Bureau itself
has a partisan agenda should be dismissed. I invite the
Congressional Monitoring Board to try to determine this and
find it and reveal it. I begged the cochairmen to have a
hearing just on this issue, and they have not yet done it.
So I just find that concerning this charge that was put in
the books 10 years ago, almost 10 years ago, there has never
been any evidence put on the table. I just wish someone would
put the evidence on the table so we could answer it.
Mrs. Maloney. I'd like to ask an operational question on a
non-ACE topic. You stated today that the Bureau had completed
50 percent of the nonresponse followup, and that sounds
remarkably good at this point. Are you ahead of schedule,
behind schedule? Could you comment on this number and exactly
where are you?
Mr. Prewitt. Well, we are always cautious, of course.
Choosing my words carefully, we are not displeased to be at 50
percent. On the other hand, the hard cases are yet before us.
We are now running into gated communities, a much higher
density of gated communities, where we are having a very
difficult time getting past the doorman. And yet we know we
have a low response rate from those areas. We are certainly now
running into the difficult cases in the inner city and the
African-American population that Congressman Davis just
referred to. We are in the difficult cases in the immigrant
populations.
So to say that we are 50 percent where we need to be is,
indeed, good news. And as I say, when you put that together
with the mail-out, we are about 85 percent finished with the
census. But we are now down to the hard cases and as I have
said in my written testimony and I have said many, many times,
at the end of the day, we will not get everyone. We would love
to be proven wrong, but we have had too many people already say
I will not answer this or I don't care what you say, you can
put me in jail--I have given you some of that evidence before--
so we will not be able to get everyone in the census.
Mr. Miller. Thank you. Mr. Ryan, if you will let me make
one comment first, and that is the question on the political
manipulation of the census. I would like to insert in the
record the Supreme Court decision concurring that Justice
Scalia said, and he used the phrase that an estimation was more
likely to be politically manipulated than the full count.
The Supreme Court is even saying that there is a difference
of potential political manipulation. And I would also like to
include in the record the CRS report on adjustment, because the
CRS said, as I said in my statement, that the Supreme Court
ruling precluded the use of sampled data for redistricting. Use
of money and other things is a different issue. But for
purposes of redistricting, it did not rule that it was illegal.
So with that, and with the inclusion of those, without
objection I am including both of those statements, Mr. Ryan.
[The information referred to follows:]
[GRAPHIC] [TIFF OMITTED] T1389.043
[GRAPHIC] [TIFF OMITTED] T1389.044
[GRAPHIC] [TIFF OMITTED] T1389.045
[GRAPHIC] [TIFF OMITTED] T1389.046
[GRAPHIC] [TIFF OMITTED] T1389.047
[GRAPHIC] [TIFF OMITTED] T1389.048
[GRAPHIC] [TIFF OMITTED] T1389.049
[GRAPHIC] [TIFF OMITTED] T1389.050
[GRAPHIC] [TIFF OMITTED] T1389.051
[GRAPHIC] [TIFF OMITTED] T1389.052
[GRAPHIC] [TIFF OMITTED] T1389.053
[GRAPHIC] [TIFF OMITTED] T1389.054
Mr. Ryan. Dr. Prewitt, let me start with a couple of
questions and then I would like to--I guess we are getting down
to this issue, the politicizing this thing. Have you made the
final decision to adjust the census numbers according to the
result of the ACE? You have made that final decision; correct?
Mr. Prewitt. No, sir, we will not make that final decision
until February, March 2001.
Mr. Ryan. In your written testimony, you indicate that you
would not release the adjusted numbers if they did not meet the
Bureau's standards of accuracy, and you said a review of the
ACE, quote, should take into consideration a review of selected
measures of quality.
Specifically what are our measures of quality and when were
they established?
Mr. Prewitt. Yes, sir, that is indeed the topic of our next
meeting with the National Academy of Sciences. We're working
these through right now. We are presenting these publically
well before we make the decision in September. Can I give you
one example?
Mr. Ryan. Sure.
Mr. Prewitt. Let us say we get the results back from the
accuracy and coverage evaluation and we have a higher than
expected undercount in inner cities of the White population
which owns its own home. That is something we don't expect.
Well, it may well be since 1990 to 2000 there has been a lot of
gentrification and a lot of this gentrification is now in gated
communities and these people are not returning their
questionnaires and we get an unexpected undercount in a
population group where we did not expect it.
When we get that pattern, what we will do is say can we
explain it? If it is a pattern we can't explain, it will make
us nervous and we will have to figure it out. If we can explain
it because there are now more gated communities in inner cities
that happen to be owned and inhabited by Whites who normally
give us answers to this, we will say now we have an explanation
for something that otherwise looks to be anomalous.
So that is what I would mean by looking at our own results
before we make the final decision.
Mr. Ryan. OK. And at your request, the National Academy of
Sciences has a special panel that has been convened, headed by
Janet Norwood, which we have been discussing today, to review
this. How important do you believe is the task of this panel?
And do you know what their time line is? What is going to be
the time line of the panel for evaluating the statistical
methods?
Mr. Prewitt. Well, we think this panel is very important,
and our work with the National Academy of Sciences has been
very important over the entire decade. However, the decision
itself about what the Census Bureau is obligated to do to
fulfill its constitutional and other statutory obligations is
clearly a decision of the Census Bureau and not an independent
committee of the National Academy.
Mr. Ryan. Let me ask you this: Will you wait for the
evaluation of the panel before you release your adjusted
numbers?
Mr. Prewitt. Well, of course not. We have to release our
adjusted numbers according to our statutory deadlines.
Mr. Ryan. Let's get to the political part of this.
Mr. Prewitt. I don't know when they are going to do their
evaluation. They are independent of us.
Mr. Ryan. Let me get to the political part of this, and I
understand your comments where you say this is ridiculous that
the census could be politicized. Well, I don't see you as a
political person. I don't see those who work with you at the
Census Bureau as political people. I see you as doing a job and
you have done a good job of enumeration and I would like to
give you credit for that. You are working at a statistical
adjustment. You are doing what you have been trained to do.
But your boss is the President of the United States. Your
boss below him is the Secretary of Commerce. Very political
people, the head of another political party. So you can
understand why you would see these kinds of allegations. I
don't think people are saying Ken Prewitt is a politician who
is seeking political ends with the statistical adjustment. But
you can see it is very rational to take a look at the situation
and who you work for and then make those conclusions.
The concern that I think many people have is the compressed
timetable. In 1991, the Bureau discovered a computer error in
the PES system that threw the undercount off by a million
people. Then during a series of evaluations that took about 2
years the Bureau discovered more errors in the system that
added millions of people erroneously. Now so far in the census
research has shown that the Bureau has had two computer errors.
One printed 120 million wrong addresses, the other failed to
print millions of surnames. These things happen. But given the
1990 experience, and given that small computer errors produce
millions of problems very easily in the adjustment, there is
cause for concern.
So can you understand that people in Congress and in the
scientific community are alarmed at the prospect of making
adjusted numbers official after less than 4 months of
evaluation? That is the cause for concern. And the other
question I have in that is are you trying to have the official
numbers done by January 20th? Is that a deadline that you are
trying to shoot for?
Mr. Prewitt. You mean the redistricting numbers? The
apportionment numbers?
Mr. Ryan. Yes.
Mr. Prewitt. Absolutely not, sir. There is no way we
could----
Mr. Ryan. The official adjusted numbers, not just the
redistricting numbers. And--well, I will let you answer.
Mr. Prewitt. I think I know where you are going. There is
of course the apportionment number, which is December 31, and
that will be finished on schedule. There is then the
redistricting number, which is April 1st. The current plan for
redistricting numbers is that they will be adjusted numbers or
corrected numbers. Under no circumstances would our schedule
allow us to produce that data tape prior to January 20th.
Mr. Ryan. I understand. I understand that the official
adjustment leads to the redistricting numbers. But don't you
think it is a reasonable concern that given the problems that
can occur with an adjustment, that a 4-month timetable is
relatively rapid?
Mr. Prewitt. Oh, yes, sir. In fact, I would say,
Congressman Ryan, that trying to get the basic census done in 9
months puts a lot of pressure on us. And, indeed, a coding
error can occur in the enumeration process as well as in the
correction process. It can just occur, and the ones that you
have cited occurred in the basic census. And indeed it is quite
possible that we will find out 2 years from now that we made
some error. We don't expect to find that in the enumeration,
but if so, we would have already reapportioned and we would
have to say ``too bad,'' we made an error and there it is.
It is not something unique to the ACE process, it is
something that is a characteristic of the entire process.
We obviously learned a lot based on 1990 and 1980, where we
did these exercises, and we have put in place--and this is what
this documentation is all about--we have put in place with
respect to our software development work enormous layers of
redundancy. We are double-coding every piece of software in the
ACE estimation process. And then we have compared the results
of two completely separate writings of the software code. And
we have built in quality assurance processes.
So we know it is a tight time schedule, but so is the
census a tight time schedule. Everything in this process is a
tight time schedule. We are pleased that the errors that have
been discovered so far did not have operational consequences
and they were, out of 2,500 different pieces of software, one
or two.
Mr. Ryan. The last thing I would want to do is force you to
do sloppy work by making you compress into an artificially
chosen timetable. Let me go back to the fact that the task of
the National Academy of Sciences, at your request, is to
evaluate the quality and accuracy of the ACE. Why will you not
wait for their review of your data before releasing your
official adjusted numbers?
Mr. Prewitt. Because we have a statutory deadline that says
we must release numbers by April 1st. The National Academy of
Science will take a couple of years. We always ask the National
Academy of Science to evaluate our work.
Mr. Ryan. What good is their analysis if you are not going
to wait for it?
Mr. Prewitt. Just the way the 1990 analysis helped us plan
2000, their 2000 analysis will help us for 2010. That is just
the nature of the system. We cannot delegate--we cannot
delegate, as the Census Bureau, the decision about what numbers
to give this country to an independent agency.
Mr. Ryan. I'm not saying you're delegating the decision to
an independent agency as to what numbers you give. But if
you're asking the scientific community to review your data, to
review the accuracy of your data before making them official,
you ought to wait for them to review your data before making
them official. That is where I think the point can be
adequately made under reasonable terms that there could be a
politicization of this process. That is the concern. If you are
not going to wait for the scientific community to look at the
data, to look at the accuracy, to make sure things were done
correctly, and rush to get these data--these adjusted numbers
out there in an official capacity, then why bother? Those
questions I think are very serious questions.
One more question and I see you are going to answer it. If
you were to release the numbers early how much kind of warning
would Congress have when you release the adjusted numbers?
Mr. Prewitt. If we release them early? You mean prior to
April 1st?
Mr. Ryan. Yes.
Mr. Prewitt. We historically have released adjusted numbers
on a flow basis. That is as soon as we finish a State we
release it. And we have certain States that have faster
deadlines than other States with respect to redistricting. And
that is what we have informed the States, we will get to them
as soon as possible. We do not expect to have any State
completed before early March. But I can imagine that we will
have States out as early as March 5th, some States out March
11th. We are going to be driving toward an April 1st deadline.
I'm not sure what you mean by informing Congress of this.
We normally simply release redistricting data tapes on a flow
basis starting as soon as we can. I am happy to tell the
Congress when we have that schedule. There is nothing secret
about that schedule.
Mr. Ryan. That would be appreciated.
Mr. Miller. Mr. Davis.
Mr. Davis of Illinois. Thank you very much, Mr. Chairman.
Mr. Chairman, let me yield for a moment to the ranking member.
Mrs. Maloney. I thank Mr. Davis for his yielding and for
his outstanding leadership on this issue.
I just wanted to respond to the series of questions that my
dear friend and colleague, Congressman Ryan, was putting forth
and ask you, Dr. Prewitt--Director Prewitt, can we expect from
you the same independence and independent judgment and action
that we saw in Dr. Barbara Bryant when she opposed former
President Bush and Secretary Mosbacher and came out for modern
scientific methods because she believed in them? We have a long
history of independence in the Census Department and in Census
Directors in speaking out for what they think is right for an
accurate count for America. Can we expect the same type of
independent action on your part?
Mr. Prewitt. Congresswoman Maloney, if the Census Bureau
looks at the adjusted data, the corrected data in February-
March and if I am then the Director and we decide that these
data have some serious flaw in them, we will simply not release
them. And irrespective of what the President of the United
States wants, whoever that may be at that time, irrespective of
what the Secretary of Commerce wants--now if they make us do
it, as Mr. Mosbacher overruled Dr. Bryant, I don't know what we
would do. But certainly the Census Bureau would not wish to
release any data product in which it did not have confidence.
I might say, continuing on this line if I could for a
moment, that I believe that in 1980 that the decision about
whether to adjust or not was left to Vince Barbara, then the
Director of the Census Bureau. I think that is the proper level
for this decision. And in 1990, the decision was not left to
the professionals at the Census Bureau; instead, it was made at
the Secretary of Commerce. I would strongly urge, strongly urge
that the decision in 2000 be made by the level of the Census
Bureau, regardless of who may be the Secretary of Commerce at
that time. But I believe that this is not a decision that
should be made at the level of the Commerce Secretary, but
should be made at the level of the Census Bureau itself and its
Director.
Indeed we have in place a standing committee that meets
every 2 weeks that goes through all of this technical stuff,
and it is designed to follow the ACE process very closely, both
in terms of its statistical theory, in terms of its operations,
and then make a recommendation to the Director as to whether to
use it or not.
Just if I could continue for a second, Congressman Ryan, I
did not fully--I do understand some of the concerns. I'm not
trying to dismiss the concerns. I'm only trying to say that
there is no evidence for those concerns. And even if a member
of the Supreme Court says that it could happen doesn't mean it
could happen. I don't know technically. If you think about it,
you are sitting there trying to generate these data and you are
now saying in what State is there a redistricting battle in
which there is a Governor of this party and a legislature of
this party and what are the processes and what would we have to
do to get the data--I mean, if you actually think about it for
a moment practically, how in the world would we do it?
Are we sitting there sort of looking at voter turnout in
different States? Are we sitting there looking at the balance
of power between the legislative and executive branches in
different States? Do we even understand how different States do
redistricting? If you actually look at it practically, it is
inconceivable that the Census Bureau in that environment of
trying to produce good data is now going to take on this extra
task of finding out what are the likely political implications
in a given State.
It's just not in the cards, and I don't see how people can
think it is in the cards. I don't care if they are on the
Supreme Court, sir. I don't know what evidence he has to make
that accusation. I simply don't know what the basis of that
accusation is.
Mr. Davis of Illinois. Reclaiming my time, you know as I
was listening to my friend and colleague from Wisconsin, I was
saying to myself, as he described the hierarchy relationship of
the executive branch that there is no way that he would think
that people with the name Daley and Clinton would be seeing
this in a political way.
Mr. Ryan. Never, ever.
Mr. Davis of Illinois. They wouldn't by no stretch of the
imagination.
Dr. Prewitt, let me try and make sure that I understand
some of the technical language. It is my understanding that
capture probability does not necessarily mean that everybody in
a category are the same, but there are enough similarities that
in terms of the probability of them being counted or enumerated
becomes essentially the same.
Mr. Prewitt. That is exactly correct, sir, yes, sir.
Mr. Davis of Illinois. So they don't have to have all of
the same characteristics but there are enough factors----
Mr. Prewitt. They are certainly not clones of each other as
was suggested. We have done a lot of research for 40 or 50
years on this, and what we do say is what are the probabilities
that we will include in the census people with this set of
characteristics. That's all it says. It doesn't say they are
alike in all other ways; just how similar are they with respect
to the probability of catching them in the census.
Mr. Davis of Illinois. In statistical language, is there a
difference between correctedness and accurateness?
Mr. Prewitt. Yes, sir. Accuracy really has to do with the
truth. And all statistical operations are estimations of the
truth. That is true of the basic census. There is a true number
of people who lived in the United States on April 1st. Our
census is an estimation of that. We use the ACE to get closer
to estimating that truth.
``Accuracy'' would be if we actually found and counted
every one of them. We will never be able to do that for you. We
believe we will get you closer to the truth by using this
process, or we wouldn't be doing it. Why else would we do it?
We have lots of things to do. We only do statistical procedures
because we believe they get us closer to the truth. So accuracy
has to do with how close to the truth can we get.
Mr. Davis of Illinois. And so the closest that you could
possibly get would be through the use of corrected data? Is
that accurate?
Mr. Prewitt. We believe so.
Mr. Davis of Illinois. And so it becomes almost--I mean, we
are trying to get as close as we can----
Mr. Prewitt. Yes, sir.
Mr. Davis of Illinois [continuing]. To making sure that
every person in the country is, indeed, accounted for.
Mr. Prewitt. Yes, sir.
Mr. Davis of Illinois. And so without using the corrected
data, we would obviously then just say to ourselves that we are
going to leave those individuals out.
Mr. Prewitt. Mr. Davis, I could answer that as follows: let
us say that in 2000 we were not doing an accuracy and coverage
evaluation. We were simply doing the basic count and then
stopping and then we came up with a number, 275,311,000 or
whatever. It would be my judgment that a more accurate number
to give to the country would be that number plus 1.6 percent.
Which is to say, I would still rather use the 1990 estimate of
the undercount even for the 2000 data if we were not doing an
accuracy and coverage evaluation. I would be convinced that
that number that we counted plus 1.6 percent would be a more
accurate number than simply stopping with the basic count. We
will do better because the accuracy and coverage evaluation
that we have in place for 2000 is a much better tool to use
than one from 1990, but even one from 1990 would give us a more
accurate count, one that was closer to the truth than simply
stopping with the basic enumeration.
Mr. Davis of Illinois. I know that we have talked a great
deal about enumeration and there is a cutoff period. There is a
time when we expect to have this done. Should we continue to
experience difficulty in some areas, will that cutoff date be
adhered to or is there any way to continue up to a point of
satisfaction?
Mr. Prewitt. We will certainly continue, Mr. Davis. We
expect across 520 offices to have completed most of our work in
most of them by our cutoff date, which is July 7th. But that is
not a cutoff date; that is a date in our master activity
schedule. But certainly, as was true in 1990 and all censuses,
there are always some local offices where we have not fully
exhausted all of our procedures and we will continue in those
areas until we have exhausted all of our procedures, until we
cannot think that going back yet again is likely to give a
response at that household.
Mr. Davis of Illinois. And so one can expect that every
effort or maximum effort will be made to make sure that we even
reach those individuals that we are having difficulty with.
Mr. Prewitt. Yes, sir, but at a certain point, we know that
we are simply wasting taxpayer dollars. And so at a certain
point we are better off--I mean, how many times do you want to
go back and knock on a door where nobody ever answers and the
person who answers says I don't care what you say to me, I'm
not going to give you that information. We could send that
person back 4 times, 6 times, 27 times----
Mr. Davis of Illinois. In some instances it would remind
me, if you just keep doing it of, you know, a young woman met a
soldier and wanted to get married and she said: Soldier,
soldier, would you marry me with your fife and drum, and he
said no, pretty miss, I can't marry you, I don't have any
shoes. So she ran and got him some shoes. Came back, same
thing, would you marry me with your fife and drum? No, pretty
miss, I can't marry you, I don't have a tuxedo to put on. So
she ran and got the tuxedo and came back. And said, soldier,
soldier, will you marry me with your fife and drum? Finally, he
says no, pretty miss, I can't marry you because I've got a
pretty little wife at home.
And so it seems to me that you are saying at some point,
people are going to say: Get away from my door, just don't come
back----
Mr. Prewitt. Yes, sir.
Mr. Davis of Illinois [continuing]. Anymore. I mean, those
individuals who are inclined to do so.
Mr. Prewitt. Yes, sir.
Mr. Davis of Illinois. Let me ask you, of course there has
been a lot of conversation about my city, the city of the big
shoulders, the city of Chicago, in terms of difficulty that we
are having. Could you elaborate on what's going on there and
what we are doing?
Mr. Prewitt. Yes, sir, and Congressman Ryan will also be
interested because it is not just the city of Chicago but the
Chicago region. I do want to say that the Chicago region as a
region is actually in fairly good shape. It is not our
strongest region but it is certainly not our weakest region.
Indeed, when you take into account both the mail-back response
rate and the completion of the nonrespondent followup workload,
the Chicago region is roughly in the middle right now and since
the whole scale is high right now that means we are in good
shape. Even in the very worst region we're actually in good
shape.
Now with respect to the city itself, I believe there are
now four local census offices where we believe that we have had
to improve the strength of our local management and we have
done so. In some instances we have actually changed the local
manager. In another instance we brought in additional
management help. Sometimes what happens, Congressman Davis, is
that there is more work going on than the system records
because stuff just stacks up and somebody doesn't have to
process that stuff every 6 hours to get it and so forth and so
on. We are finding that out. That may not be an explanation but
that may be part of the explanation in Chicago.
I certainly think that we are running in Chicago into deep
resistance to cooperating with the census. And that is actually
happening at both ends of the economic scale. We are running
into very difficult times in the near north in gated
communities. These are people who are very busy. They are, you
know, worried about their stock market returns and so forth.
They did not send the return in and now we're having a hard
time getting past the doormen who guard these buildings and it
is extremely difficult.
On the other hand, what we do is we do special things. We
go to the building manager. If that doesn't work we go to the
owner of the building. If not, we sometimes go to somebody
influential in the city and try to get them to make that call.
And at the other end of the economic scale, as you well know,
the poverty people that you mentioned in your own district,
those are very resistant people. They are disconnected from the
society. They are indifferent to their obligations. They do not
feel that the U.S. Government or the local government cares
about them and why in the world should they cooperate with
this?
That's why we do an accuracy and coverage evaluation. We
are doing one quarter of 1 percent of the households. When
you're doing nearly one quarter of 1 percent of the gated
communities with your very best people, you have a higher
probability of getting in than when you are trying to do the
entire universe of gated communities. It is the same thing with
the young African-American male in the Robert Taylor home. It
is very difficult to get them all.
On the other hand, when you're doing one quarter of 1
percent of them with your very best enumerators the
probabilities have just gone way up that you will get them. And
that is what we do in order to calculate the undercount.
Mr. Davis of Illinois. After this is all over, does the
Bureau have sociologists and researchers and people who will
try and study the situation and make some determinations
relative to this deep resistance that you spoke of?
Mr. Prewitt. Yes, sir. Yes, we do. We did--after 1990 we
had anthropologists and sociologists trying to help us
understand these population groups.
Mr. Davis of Illinois. Finally, I think it would certainly
be good, and I understand that you are trying to make a trip
out to Chicago to give whatever additional assurances to the
elected officials and the citizens there that every effort is,
in fact, being made to overcome the deep resistance that we
might be experiencing, and I certainly look forward to that
happening.
Mr. Prewitt. Thank you, sir.
Mr. Davis of Illinois. Thank you, Mr. Chairman. I yield
back.
Mr. Miller. Mr. Director, let me followup one more point on
what I was asking earlier about the broad classifications we
are using. And I would like to enter into the record a letter I
received from Dr. Friedman, who is head of the Statistics
Department at University of California at Berkley. He said,
``It is assumed that all Non-Hispanic Asians age 0 to 17 living
in rental units are equally likely to be undercounted from the
suburbs of Honolulu to Chinatown in New York. This assumption
is plainly false.''
They have done studies to show that there are huge
variations within post-strata across States and so there is a
real concern about that. You are well aware of that concern.
Let me now switch to the issue again of transparency with
respect to the ACE. You indicated your intention to make the
census fully transparent and free from charges of political
manipulation. Will you commit to releasing the E sample and the
P sample files from the ACE for analysis by the academic and
scientific community as soon as they are available to the
Bureau? They are not confidential files and for 1990 they were
not made available until 1998.
Mr. Prewitt. Yes, I don't know the 1990 to 1998 process,
but certainly they will be made available, yes, sir.
Mr. Miller. There is a real concern about the decision
process of which is going to be the more accurate set of data.
And if the National Academy of Sciences is not going to be able
to make a decision prior to March 2001, who is going to make
that decision?
Now, my understanding in 1991, when this decision was made,
that there was a panel within the Bureau of experts that was
basically equally divided to help, we may need more
clarification, but there was some panel of experts within the
Bureau. But you are not going to rely on the National Academy
of Sciences because they are going to take too long, I gather.
Mr. Prewitt. Yes, sir, we could not wait. The National
Academy of Sciences will do an evaluation--there is already a
2010----
Mr. Miller. So how is the decision going to be made? Is it
going to be made--of course you won't be there, unless whoever
is appointed President. We know there is going to be a new
President. But Mr. Thompson, Mr. Hogan will certainly still be
there. What experts? Are they strictly Bureau employees or who
is going to come up with the recommendations? I think there was
some outside people making recommendations, acknowledged
experts.
Mr. Prewitt. What will occur--let's just talk about 2000.
What will occur in 2000 and is occurring in 2000, we do have an
executive committee that follows the ACE process. As I say it
meets every 2 or 3 weeks, many members of which are here behind
me. I think there are maybe 9--no, it is larger than that,
maybe 13 members of that who represent all Census Bureau
employees who are math statisticians, demographers, field
operations experts and so forth. And they look at every one of
these processes, every one of these processes, and make a
judgment and deliberation about what will make the most
successful census.
They will continue to meet right through the entire
process. The way they are designed, that committee is chaired
by John Thompson and it is advisory to the Director. It will
make a recommendation to the Director, is the process.
Mr. Miller. One of the concerns I've had going back a
couple of years or so is you can have a bias within a
committee. If I select a committee or Mrs. Maloney selects a
committee, if we have sole responsibility for selecting it, it
will be a bias by who we select. I mean you ask Mr. Davis and
Mrs. Maloney, they are going to have one set of opinions. Mr.
Ryan and I will have another. My understanding was more of a
nonpartisan--if you select all people that are already biased
in favor of adjustment, you are going to have that conclusion.
And I'm sure Dr. Hogan, who is a respected statistician is bias
to some extent. Because he has had his heart and soul in this
for a decade, he's been working on this program.
But Mr. Friedman--Dr. Friedman at the University of
California--Berkeley, who is not going to have any input in it
is a respected statistician too. So, I mean are the only people
that are going to provide input just going to be people who are
``yes'' people?
Mr. Prewitt. No, sir, I don't know what you mean by ``yes''
people.
Mr. Miller. I don't consider Dr. Hogan a yes person.
Mr. Prewitt. You wouldn't if you met with him. These are
professionals.
Mr. Miller. I want to make sure there is a diversion of
opinions in the decision process.
Mr. Prewitt. If you sat and listened to some of the
arguments that go into this you would appreciate there is a
divergence of opinion. And it certainly includes people who in
1990 thought we should not have adjusted who are employees and
very senior, important employees at the Census Bureau. This is
not a committee that was sort of put together that way. It is a
committee all of whom have defined positions. These are the
senior positions, and so they are there by virtue of the
position they hold, not the kind of assumption they have. We
did not test anybody's viewpoint.
Mr. Miller. There are no outsiders participating in this?
Mr. Prewitt. No.
Mr. Miller. But there was in 1990 is my understanding.
Mr. Prewitt. No, that was a different process. I can
describe that. That is a different process.
Mr. Miller. I would be interested to have an explanation of
how the process or the decision will be made. In your written
testimony you say you will use something called dual system
estimation to estimate the degree to which each of the 448
categories of post-strata is overcounted or undercounted. Then
you would assign a certain weight to that category which would
tell you how many people to add or subtract from that
particular segment of the population; is that correct?
Mr. Prewitt. Is that correct? I don't think that is exactly
my wording. I don't think I talked about--I would have to look
it up, but I think it talks about statistical records, not
people.
Mr. Miller. Let me proceed. If it is possible to have
strata with adjustment factors of more than one, you have an
adjustment factor of--it may be 1.1, 1.2, it is also possible
to have post-strata with adjustment factors of less than 1.
That is people fitting a certain description could be
multiplied by a factor of 0.8 or 0.9.
Mr. Prewitt. Correct.
Mr. Miller. Correct? OK. Let's take an example, let's say
you are talking about the following: A non-Hispanic White woman
age 30 to 49 living in the suburbs who own their own homes in
the Midwest. Let's say the Bureau estimates a 5 percent
overcount of these women. The Bureau would give this group an
adjustment factor of 0.95. So if the unadjusted census counted
100 people--100 women in a block, the adjusted number would
show only 95 in that block.
The truth is if the actual census counts 100 people in a
block but in the ACE all of these people fall in a category
that the Bureau estimates were overcounted, the adjusted
population of the block will be less than 100. So we are in
effect deleting people from the census.
Mr. Prewitt. No, we're not.
Mr. Miller. I mean, if we have 100 people and their
adjustment factor is 0.95, we are only going to have 95 people
counted. No? I know we're not going to destroy forms. We're not
talking about the forms being there, but the fact if----
Mr. Prewitt. It is very important to make the American
public understand that 72 years from now, when you and I go
together to the National Archives, everyone who submitted a
form will find their record there. And there will be no form
there from anyone who did not submit a form. That is, the
actual census file itself will include everyone who cooperated
in this census.
Now, we are now talking about a statistical record. And
that is a different process. So it is not anything about people
being subtracted or virtual people or anything else. We're
talking about a statistical process. The answer, sir, is yes.
Where we have evidence that a certain population group was
double-counted, to leave records for those people in the
statistical record, means that we have now inflated some
number. We are now giving to the country something which we
know to be incorrect, and we don't think we should do that.
Mr. Miller. If two people completed the form, one in
Florida and one in New York and it is the same person, we don't
want that. I understand that. But the problem is my
understanding is that if you have 100 people living in an
apartment high-rise or something. If that is a statistical
classification that is considered overcounted--you have 100
people you count, we have 100 forms that are returned. All
right? And you have 100 people listed by name. But then because
that fit in a classification that is considered overcounted,
you are going to subtract people from that so the actual count
instead of 100 would be 98, or whatever adjustment number;
right?
Mr. Prewitt. Otherwise we would be giving the country
incorrect data.
Mr. Miller. Then you are deleting people from the census.
Mr. Prewitt. No, we were not deleting people.
Mr. Miller. Wait a minute. We're keeping the forms. I
understand the forms are going to be there. I don't know if
they are going to be physically kept, that is a different
issue. People are going to get counted less than a 1.0. You are
counted as a 0.98, a 0.95. You are going to have fewer people.
If you have 100 people that fill out that form, you have 100
names in that area and it comes out with a 0.98 adjustment
factor because of the statistical analysis, then you are only
going to have the number that is going to show up on the
adjusted, or you like to call it corrected, the adjusted number
is 98 people. Two fewer people.
Mr. Prewitt. Happy to call it adjusted. Again, if I could
just take a moment, the people that you are describing, that is
the category of persons that you are describing, we have
independent evidence that those kinds of persons were double-
counted at the rate of .02 percent, to use your example. And,
therefore, to leave statistical records of that category at the
level which you are recommending that we do means that
basically we're deliberately leaving in the census counts
people who have been double-counted, because they counted their
college student and we found their college student at the
dormitory. That happens.
And what we know from 1990 is there were as many as 4
million cases. So, yes, we have a statistical procedure that
for the purposes of giving this country accurate data for
reapportionment, for redistricting, for Federal funding, we
have a process that does not give the country incorrect data
when we know it is incorrect. Nothing more complicated than
that.
Mr. Miller. If you have 100 names in this area, in this
block, and your statistical analyses says that is an
overcounted population, so even though you have 100 names of
100 separate individuals, you are going to statistically remove
two, three, four, whatever the number of people of that overage
is. And this is one of the problems about all of these post-
strata. It is like the issue Dr. Friedman talks about, your
claim that you are getting these numbers from Asians in Hawaii
and Asians in Chinatown are the same. They have the same
response rate. Some studies show that they don't behave the
same. And, you know, I guess you have got proof that shows that
the Cubans in Miami respond at the same rate of response as the
Mexicans in Los Angeles or in El Paso or somewhere.
I mean you are saying they are exactly the same behavior.
Based on that, you can delete people or add people, which is
hard to say that--I have not been to Los Angeles----
Mr. Prewitt. That is your characterization, not ours, sir.
Mr. Miller. But aren't you using--well, you have already
said you're using--all Hispanics were one classification.
Whether you are a Hispanic in El Paso or Houston or New York
City or Chicago or Wisconsin, you get counted the same and you
get adjusted the same if you are Hispanic.
Mr. Prewitt. I didn't quite say that. I said----
Mr. Miller. Well, but aren't all Hispanic one category,
period?
Mr. Prewitt. No.
Mr. Miller. No?
Mr. Prewitt. All Hispanics who also are in census tracts
with low response rates who also rent their houses, who also
are between the ages of 18 and 29, who also are women, who also
are unrelated to anyone else in that household. All of those
people who have that set of characteristics constitute a
universe of those people.
And then we take a sample of those persons and, on the
basis of that sample, estimate for that universe of people who
have all of those characteristics--not just Hispanic, but all
of those characteristics--what are the probabilities that they
were caught in the census. And that is the process that we
used. It is not ``all Hispanics'' because we are--it's like
saying all renters or all people between ages of 18 and 29 or
all anything else.
Mr. Miller. But the post-strata for Hispanics is all
Hispanics, whether again they are in El Paso or Chicago or
Miami, they are all the same. Whether it is Guatemalan,
Honduran----
Mr. Prewitt. I just have to say this again. They are not
all the same.
Mr. Miller. Statistically, you're putting them in one
classification.
Let me ask another question. When you take someone--
subtract someone from the record, you subtract them, but it's
not because of a duplicate. It's just that some statistical
model says subtract one person. When you have 100 people on a
block and the statistical model says subtract somebody, it's
not because you have a duplicate. It's because you have 100
specific names there, but it's because of the statistical
models, not because of a duplicate.
Mr. Prewitt. That's a separate process. Subtracting
duplicates is a separate process.
Mr. Miller. Go ahead and finish what you were going to say.
Mr. Prewitt. We don't treat all Hispanics like all other
Hispanics. We treat Hispanics who rent, who are of a certain
age, who are of a certain gender, of a certain relationship to
the household. That is the post-stratum. Not all Hispanics.
They all have to live in a metropolitan area. So it's simply
incorrect to say that all Hispanics belong to the same post-
stratum.
Mr. Miller. But the Hispanics that meet those
classifications can be living in Los Angeles, El Paso, Houston,
Miami, or New York or Chicago as long as they meet those
general classifications. Then they are all adjusted. Asians in
Honolulu are being pooled with the ones in New York, and you're
saying they respond the same.
Let me go on to Mrs. Maloney.
Mrs. Maloney. Dr. Prewitt, would you please answer Chairman
Miller's line of questioning without interruption? I would like
to give you an opportunity to explain the process without
interruption.
Mr. Prewitt. I think the particular process we're talking
about is the structuring of the post-stratum which, as he said,
there are 448. These constitute identifications of population
groups, and one of the identifying characteristics is their
ethnicity or their race. It's only one of their identifying
characteristics. Another is whether it's a metropolitan area or
not, the size of the metropolitan area. Another is, as I say,
age, renter status and so forth and so on. That constitutes a
post-stratum, and it is our judgment that everyone who inhabits
that post-stratum has a more similar probability of being
captured in the census than someone in a different post-
stratum.
Everyone in the country is put into post-stratum. You are.
You're put in as a White female between 18 and 29----
Mrs. Maloney. Why is everyone laughing?
Mr. Prewitt [continuing]. Etc.
And we have--based on our experience, we have an assessment
of the probability of having caught you in the census, and
that's true for all of these groups. Nothing more complicated
than that. That's why they're put together.
We don't yet know until we actually conduct the census how
many of them we actually did catch in the census, but we think
they constitute a reasonable, plausible, universe of people who
have roughly similar probabilities of being captured.
I have not had the chance to read Mr. Friedman's and Mr.
Walker's letter. If they are saying that all Asians from every
place are put in one post-stratum, they are misreading our
post-stratum design. They are very sophisticated statisticians,
and I doubt they are misreading it. I doubt that, the way you
have characterized their letters, the way they have written it.
But I haven't read it, but, my guess, they understand our post-
strata structure, and it's not putting all Asians in one post-
stratum. It's not.
Mr. Miller. It's all Asians that meet the large
metropolitan areas and age brackets and such, too. But it's
correct that a Japanese American in Honolulu that meets that,
you know, other demographic characteristics and a China person,
a person from China from New York who meets that
classification, a large metropolitan area, age brackets----
Mr. Prewitt. Mr. Miller, it's just as--there are lots of
ways to rent a home. You can rent a condo, you can rent a co-
op, you can rent a mobile home. You can rent different kinds of
homes, so all renters are also put into a post-stratum because
that's one of our stratification variables. It's not just that
all renters constitute a post-stratum. It's that all renters
that also have these other characteristics create one.
So there's nothing magic about this process to say there
are a lot of different ways in which people rent, but
nevertheless we have decided that renters on balance behave
differently from owners, and we have a lot of evidence to that
effect.
Mrs. Maloney. Dr. Prewitt, approximately how many post-
strata of the 448 include the Hispanic characteristics,
approximately?
Mr. Prewitt. Fifty-six.
Mrs. Maloney. I think this whole issue of the undercount
and the deep resistance that my colleague Danny Davis
illustrated with the poem--the time that it was clarified to me
in the most stark way were the statements of a Republican-
appointed member of the Supreme Court, Justice Stevens, when he
asked the question of a Republican lawyer--and this was
before--the case that we've referred to before the Supreme
Court. And he asked her, how would you count a home, an
address, where six people lived, yet every time you went and
knocked on that door, whether it was in the morning or at night
or whatever time, no one answered the door? And she said, zero;
we would count it zero. Then he asked, what if you knew and all
the neighbors told you that six people lived there? She said,
we would count it zero.
Then Justice Breyer asked, what if the lights go on, off
and on, every night and you see the lights going off and on
every night and you know people live in that home? How would
you count that home? And she said, zero.
And that really clarified in the starkest and really
simplest of terms why we need to adjust for the undercount when
we know that people live there, when we know that people are
there. We are being dishonest and unfair and unjust not to
count the six people that we know live there. And on the count
and the issue----
Mr. Miller. Let Dr. Prewitt answer that.
Mrs. Maloney. May I continue? I do not believe I've
interrupted you. May I continue?
Mr. Miller. Yes.
Mrs. Maloney. And on the issue of the double count, many of
my friends, because I am a mother, happen to be the parents of
daughters; and I can't tell you--and my daughter is in college.
I did not count her. She is going to be counted at her
university. But I can't tell you how many of my friends who
have similar children my daughter's age at school either told
me that they counted their child or literally called and asked
me whether or not, because they know I'm working on the census,
whether or not they should count their child. So I'm giving
these as just practical examples of why we need this.
Now, I have a question that--Dr. Prewitt, you mentioned
that you would use the 1.6 percent if you had to, but you also
said that the tool that you had for the 2000 census was a
better tool than 1990. And could you explain to us why it is
better?
Mr. Prewitt. Well, obviously, we've drawn on our experience
from 1990. We also have a sample of approximately twice the
size of what we had in 1990. That was 150,000. Turned out to
be, finally, 175,000 households. In 2000, 314,000 households.
We do think the construction of our post-strata is drawn upon
research of over 10 years about all of our matching procedures,
how we're handling movers, our software development work.
There's no end of ways in which we try to improve it. That's
true of every census.
1990 was better than 1980, but 2000 is much, much superior
operationally, just like the census itself is superior
operationally to the 1990 census thanks to the U.S. Congress,
that they allowed us to front load our recruitment staff.
That's why we can say we're near 85 percent complete today. A
lot of the improvements that we put into the census we've also
put in to our ACE design.
I just--for a moment if I could refer us all to this--I
brought this chart before--because I think it's important. Each
of those peach boxes represent the moments in the census when
we can miss people, but it also represents the moments in the
census when we can erroneously enumerate, that is, double count
such as the college students. So all the ACEs is nothing more
complicated than this, all it is, is to go to try to find those
persons who returned the form but didn't completely mail it
back. They left some people off.
It's the people who--for whom we never got an address. We
think there won't be many of those, but there are some. It's
the people we got in nonresponse followup, but we didn't get
the complete household. And it's the people we got in yet
another process called coverage improvement followup.
All of those are processes to try to get everyone. Every
one of those processes can leave someone out, and all the ACE
is, is a way to go back and find out the percentage of people
in those various boxes when we missed them, how we missed them
and what their demographic characteristics are. It's not a very
complicated thing. It's a very straightforward thing. If it
works operationally, we think that we should give to the
country the better data, the adjusted data, the more accurate
data, the more corrected data.
That's all I can really say about it. It's an attempt to
find the people we missed or to find the people we erroneously
included, that is, the double count and make certain they are
not represented in the final statistical records.
Mrs. Maloney. Thank you very much.
Mr. Miller. I just want to ask you to clarify one thing.
Mrs. Maloney I think knows the answer to this. But when the
lawyer spoke about not counting someone if the lights come on
and off, that's not the way the Bureau would handle that; is
that correct?
Mr. Prewitt. In that particular instance we would try to
get a proxy interview.
Mr. Miller. Correct. You would get proxy data. And,
hopefully, you could find someone that would know who lives in
there, and it would not necessarily be zero. You'd try to do
everything you can to get some type of data from someone else
nearby.
Mr. Ryan. Like a neighbor or something like that.
Mr. Miller. Right. So I think the attorney was not as clear
on the procedures as your process would show; is that correct?
Mr. Prewitt. That is correct. There are many things about
that Supreme Court ruling that were not accurate.
Mr. Miller. Mr. Ryan.
Mr. Ryan. I have a couple of procedural questions.
It's not always the inner city. We have to focus on the
rural areas, too. So I'd just like to put in a word for
Orfordville, WI, if I might.
Orfordville is a town of about 700 people. Hopefully, it
will be a town of 700 people after the census is done, but the
interesting thing about Orfordville is they all have post
office boxes. That is just the way it works there. They all use
P.O. Boxes. So when they didn't get the forms they were very
much alarmed.
I think we followed up with your Chicago office, and I
think we're doing a very good job of getting some enumerators
over there to handle that situation. But what about the other
Orfordvilles throughout America, small farming towns at the
intersections of rural county trunk highways?
My question is, if we didn't intervene on the Orfordville
situation and there are other towns like that who have P.O.
Boxes who, because we don't have post census local review,
didn't catch that and these are not included in the master
address file, how do we catch these mistakes? Can the
adjustment help a neighborhood where no people are counted
essentially?
Mr. Prewitt. Yes, sir. But before I get to that question,
we do have procedures to find those areas where there was
this--where we thought it was city style, but it turned out to
be----
Mr. Ryan. We're working on that right now. But what if it
doesn't work?
Mr. Prewitt. Absolutely, we would find this in the ACE.
Here's how we would find it. Since the ACE is a random sample
of all blocks in the United States--as I say, about 12,000
blocks are in the ACE design--one of those blocks or some set
of those 12,000 blocks would be exactly those areas by
definition and proportionate to how many such areas they are.
When we go to that area in the ACE interviewing process,
we'll knock on the door. They'll say, I never got counted; I
got left out. We will then determine how would that have
happened, and we will then detect exactly that problem. Indeed,
it will show up because our address file won't work.
We will have independently listed--as I say, we have
independently listed every address in our ACE sample block; and
we are now saying, my goodness, something has gone wrong.
Because we have a listing of this household, but it's not in
our master address file. How could that have happened?
Then we'll determine how that happened, and when we do the
adjustment we will be able to adjust for exactly those
population groups who fit into this top upper right peach box.
This is missed housing units. Our ACE design is as focused on
making sure that we account for missed housing units as missed
people in known housing units.
Mr. Ryan. I'm going to go back to this Orfordville example,
because I think it's an interesting one. Not only do they all
use mostly P.O. Boxes--but let's take Footville, which is the
town just up the road. Footville, for some reason, your master
address file, even though the LUCA tried to change this, the
change was not incorporated. You included everybody who lived
in Footville, WI, a town of about 600 as if they were
Janesville, WI, residents. So the names were correct. The
addresses, however--the street addresses were right, but the
cities were--were the larger city in that county. And they all
had P.O. Boxes.
So when the enumerators came around to collect the data,
they knocked on the door. The people would say, I never got a
form; I was never counted; I was worried you wouldn't come by;
glad you're here. And the enumerator then had a Janesville
address.
Now it's up to the person who answered the door to change
that address, I assume, from Janesville to Footville, but what
if that didn't take place? What if an enumerator didn't make it
to the house that was a P.O. Box and the address for the entire
small town was lumped into another city and those people
weren't counted? That means in a town of, say, 600 people you
missed 200 people. That's a third of the city of the town. How
does the adjustment fix that?
Mr. Prewitt. I want to make certain that we give you a full
answer to this. So what I'm saying may not be completely
responsive.
It's my understanding that what would happen--the important
thing is the addresses are all geocoded, which means whatever
the kind of the denominator is, what the town is called----
Mr. Ryan. Footville.
Mr. Prewitt. Footville is called----
Mr. Ryan. Janesville.
Mr. Prewitt. Janesville, that the important thing is to
make sure that when we count the person in that household they
occur on the block. From the census point of view, the name of
the community is not what's important. The unit of analysis for
us is the block, and they will be geocoded to that block. So
they will appear in the right place.
Now, the process by which we make sure that all of our
blocks get attached to the right place, but now we've got to
make sure to connect to the right denominator. That wouldn't be
a problem of the adjustment. That would be a problem of our
geographic division working with the local community. It would
be easy to fix because we know where they are.
Mr. Ryan. I see my time running out. You mentioned in my
earlier questioning that you're going to be giving the States
the redistricting numbers kind of on a State-by-State basis
starting maybe March 5 and then moving out, but your post-
strata adjustment is based on a national scale, correct?
Mr. Prewitt. Yes, sir.
Mr. Ryan. How do you take that into consideration as you're
releasing State redistricting data on a State-by-State basis--
when your post-strata is national, how does that jibe or
correspond with--say you put Vermont's out in March and then
you put Wisconsin's out in April, then California's out in
later April, how does that correspond and how does that take
into account the fact that the post-strata is national but
you're sending out individual States earlier?
Mr. Prewitt. All of the work that will have been used to
create the correction numbers on a national basis will have
already been done across all 50 States, the District of
Columbia, and Puerto Rico, and the actual mechanical process of
actually now creating the right products takes a while, and it
takes a couple of days or whatever.
I better be careful, though, have to correct me, but it
takes a period of time, and we will simply turn first to those
States which have earlier redistricting deadlines as best we
can. I should say the entire--I'm not making a promise about
March 5. I'm only saying that, in principle, it will be a flow
basis to try to respond to your question about January 20.
But, basically, we could wait until the last day of the
month, March 31, to mail them all out the same day. But I
think, as a courtesy of the States, we would want to get them
out where we can get them out sooner where possible, but all of
the work that has to be done in terms of making the correction
numbers from the post stata will have already been done.
Otherwise, the implication of your question is correct. We
couldn't do one State and so forth.
Mr. Ryan. One more question. The apportionment data--and
correct me if I'm wrong, the apportionment data will be done
before the ACE adjustment is completed?
Mr. Prewitt. Correct.
Mr. Ryan. How will you be taking into consideration
Orfordville and Footville, WI? If the apportionment data is
done before the adjustment and if those towns aren't fixed and
counted for, will they not be lost in apportionment but--may be
caught up in redistricting but won't they be lost in
apportionment if they are not fixed with the adjustment
beforehand?
Mr. Prewitt. Not if they are in Wisconsin. The
apportionment number is nothing but a State total.
Mr. Ryan. You're saying because it is this block, but what
if an enumerator didn't hit a door and no one got answered?
Mr. Prewitt. No. If they are not captured in the census and
we have not done the ACE, then the Wisconsin number will be
deficient by that amount.
Mr. Ryan. Because we have a lot of reports about rural
areas who are subsisting mainly of P.O. Boxes and the
enumerators just don't catch them--you know, we're planting
right now in Wisconsin. People are in the field. They are not
in their homes right now.
Wouldn't a post census local review make sense? Wouldn't a
1-month post census local review--let the county clerks, let
the local county board supervisors take a look at the data and
say, gosh, you missed half the town of Orfordville because,
during the time you were coming around with enumerators, they
were out in the field planting. Wouldn't post census local
review make sense for these cases?
And these cases I appeal to you are not unique. They are
all over the place. Our Governor, Tommy Thompson, is saying
he's getting it from the entire State of Wisconsin. Why
wouldn't we want to do post census local review for those kinds
of instances?
Mr. Prewitt. Obviously, a post census local review would
have to be done for 39,000 jurisdictions, not just yours, which
means you're asking us to redo the census starting sometime in
October or November. That's impractical. If you wanted an
apportionment number by December 31, we can't start redoing the
census based upon 39,000 different mayors or county
commissioners saying we would like you to come back and count
again because we don't think everybody got included.
Mr. Ryan. What about a voluntary post census local review,
like localwise?
Mr. Prewitt. This gets into a very complicated thing having
to do with the nature of distributive accuracy and numeric
accuracy. It's really what the court case went to and so forth.
And I can get into this if we have time, but any kind of
voluntary process like that that was used in some places and
not other places would have all kinds--at this late stage in
the census would have all kinds of implications for the final
quality of the data.
I can't imagine if we made this voluntary that the only
State that would be interested would be Wisconsin. I think
every State in the country would say come back and count us
again. We may find a few more people. That is what the ACE
does.
Congressman Ryan, I'm not trying--you are making a case for
the ACE. You are making a case for why we have to do this
quality process to go out and determine if we miss people,
where they live, and then correct for that.
Mr. Ryan. Actually, I'm making a case for post census local
review for apportionment and everything else because LUCA was
designed to fix this. It didn't fix it, though, in some of
these towns. Some of these towns did participate in LUCA, did
send their data, and they still--we still have the problems.
So that's why I'm saying, why not exhaust every effort
possible? I still contend that there may be a chance, there may
be a small timeline, a small window to do a voluntary post
census local review so these rural towns who are having these
problems can make sure they are counted. There is a lot of
anxiety out there over this. I just appeal to you to take a
look at that.
Mr. Miller. We have a vote coming up, but we have time for
Mr. Davis.
Actually, I'm glad, Mr. Ryan, you're on this panel. Because
rural areas, as we all know, have problems of their own. And we
keep focused on large metropolitan areas and the migrant
population, immigrant population, but there are unique problems
in rural America, so I'm glad you can bring them up.
I agree with you, by the way. It's too bad we don't have
full census local review which the House of Representatives
passed but was opposed by the Census Bureau and by the
Democrats.
Mr. Davis.
Mr. Davis of Illinois. Dr. Prewitt, are there any post-
strata groups that we've conclusively determined to absolutely
be the most difficult ones to enumerate or count?
Mr. Prewitt. Well, based on our 1990 experience, we would
expect a group that was made up of young African American males
in inner cities who rent and who live in irregular housing,
thats unrelated to each other and so forth in that housing,
that's likely to be a particularly hard-to-count population
group.
Also, based on 1990, though we think we've done a lot of
work on this, Indians living on reservations with other sets of
characteristics were more difficult to count.
Age is actually a big factor in how well we count people.
That's also true in the rural areas.
By the way, the post-strata structure, of course, includes
a special post-strata just for rural areas and especially for
rural areas where they rent, which we know to be a hard-to-
count population group, highly mobile and so forth. So we are--
in that sense, the design takes care of--it sweeps across all
of the problem situations in the country, not just fixed on the
one.
But, yes, sir, we will have a particularly difficult time
with that particular population group.
Mr. Davis of Illinois. I thank you very much. I think it's
been a very productive hearing, and I certainly want to thank
you for your responses.
And, Mr. Chairman, I want to thank you; and I'm going to
yield to the ranking member here.
Mrs. Maloney. I tell you I'm getting tired of all of this,
of this constant effort really to disrupt the efforts of the
professionals at the Census Bureau from doing their job and
from correcting the undercount.
I would like to put in the record an article from the
``Washington Post'' written by David Broder entitled, Playing
Hardball on the Census; and I think he clearly puts into focus
what's going on.
He says, in preparing for the showdown on the census,
Republicans reshuffled the leadership of the House Census
Subcommittee and hired its new staff director, Thomas
Hofeller--this was back in 1998--a Ph.D. Professor and battle-
tested GOP strategist in redistricting. And he talks about
meeting with Hofeller and how Hofeller goes to a blackboard
analysis of the Census Bureau's plans stressing the risks they
see of serious miscalculation with untested techniques and a
tight timetable.
But as I was leaving, Broder says, Hofeller offered a
decidedly non-academic comment; and he said, ``someone, he
said, should remind Bill Daley, the Secretary of Commerce and
overseer of the Census Bureau, that if he counts people the way
he wants to, his brother, Chicago Mayor Richard Daley, could
find himself trying to run a majority-minority city.'' This is
Hofeller talking. And Broder then explains this blunt reference
to racial ethnic realities is not uncommon on either side of
the fight.
Among the thick file of scholarly papers Hofeller gave me
was a memo entitled, Why Conservatives Should Be Opposed to
Census Sampling; and it went on and said and warned in these
papers--again a direct quote from the Republican papers--a
census that uses sampling and statistical adjustment will be
the biggest victory for big government, liberalism since the
enactment of the Great Society. These statistical techniques
will be used to add millions of virtual people to big-city
population centers, thus increasing the political power and
levels of Federal Government funding in those jurisdictions.
Then came two pages of answers of how this outrage can be
stopped. And it outlines the courts, in Congress, the
grassroots, and they are trying to do this.
We have been to the Supreme Court. I understand there has
been another suit filed in Virginia against scientific
sampling, and I remind my colleagues that two budgets have been
held up. Anti-sampling language was attached to a disaster
relief bill, and yet we have pages and pages of testimony that
there is an undercount. We either correct it when the
professionals have told us how to correct it or we deliberately
don't count people. That's what this hearing is about, whether
we correct for the undercount or whether we do not and
therefore deliberately not count people.
Mr. Miller. Thirty seconds.
Mr. Ryan. Mrs. Maloney, I just wanted to put for the
record, where I come from it's not a Republican-Democrat issue.
Democrat--liberal Democrat politicians from my home State--
Senator Herb Kohl, Mayor Norquist of Milwaukee--are also
opposed to sampling. We think it's bad for our State. We think
the scientific community is out on this one. So I just wanted
to say it's not a conservative-liberal thing, Republican-
Democrat thing everywhere. In some places, it is. It's just
wrong to paint that very broad brush.
With that, I yield.
Mrs. Maloney. May I respond? Because my name was mentioned.
Mr. Miller. Let me make my statement, please. We are
running low on time. We can come back if you want.
I want to put in the record an editorial by Peter Skerry in
last Sunday's Washington Post. It was titled, We're Overstating
the Importance of the Undercount. I think it's a good
explanation of the fact we really are overstating the
undercount.
What this hearing was about was whether we were going to
use statistical methods and adjustments to a census. There is
real, legitimate concern that the method will not work at the
block level and to use it for the redistricting purposes--I
think it was a good hearing. There are still a lot of questions
to be answered. We'll be discussing this, I'm sure, for the
next months ahead, but there is a real debate within the
statistical community that the method will not succeed, and
that's the reason we've got to be careful. As Justice Scalia
said, there is a potential political manipulation.
On behalf of the subcommittee, I would like to thank you
for appearing here today.
Mrs. Maloney. May I respond?
Mr. Miller. We have a vote. We'll come--if you want to come
back----
Mrs. Maloney. I would like to respond to what was stated by
Mr. Ryan.
Mr. Miller. We will then recess and come back after we have
a vote.
Mrs. Maloney. I would like to respond now for 2 seconds.
Mr. Miller. If you can do this in 15 seconds, go right
ahead. Otherwise, I'm going to adjourn it here.
Mrs. Maloney. My dear friend and colleague from Wisconsin
mentioned that it was not a division between the Democratic and
Republican party, and he mentioned names in Wisconsin that
supported his point of view. But there is a clear distinction
between the two parties on a national level, from the President
who supports the use of modern scientific methods to the entire
leadership on the Democratic side.
And I would like to put into the record statements that
have been reported by the press quoting the Republican
leadership that they will not let it go forward. Newt Gingrich
called it a dagger in the heart of the Republican leadership,
and Linder said even if the court approved it he will stop it.
When you say it is not a division between the two parties, it
is----
Mr. Miller. Mrs. Maloney.
Mrs. Maloney [continuing]. It is clear. It is in the record
not from my lips but from the independent press. I would like
to put those statements in the record.
Mr. Miller. Mrs. Maloney, present the records right now.
Come on. We're trying to get a vote together. You talk about
all this away from the partisanship, and all you want to do is
go back to Newt Gingrich who left Congress over 2 years ago, 3
years ago. This is 2000. We're in the middle of the census.
On behalf of the Census Subcommittee, I want to thank you
for being here today.
I ask unanimous consent that all Members' and witnesses'
opening statements be included in the record. Without
objection, so ordered.
In case there are additional questions that Members have
for our witnesses, I ask unanimous consent that the record
remain open for 2 weeks for Members to submit questions for the
record and that the witnesses submit written answers as soon as
practicable. Without objection, so ordered.
Meeting adjourned.
Mrs. Maloney. I put the quote in the record from John
Linder, the head of the RNC, now.
[Whereupon, at 12:04 p.m., the subcommittee was adjourned.]
[Additional information submitted for the hearing record
follows:]
[GRAPHIC] [TIFF OMITTED] T1389.055
[GRAPHIC] [TIFF OMITTED] T1389.056
[GRAPHIC] [TIFF OMITTED] T1389.057
[GRAPHIC] [TIFF OMITTED] T1389.058
[GRAPHIC] [TIFF OMITTED] T1389.059
[GRAPHIC] [TIFF OMITTED] T1389.060
[GRAPHIC] [TIFF OMITTED] T1389.061