Census 2000 Gateway |
Glossary![]() |
![]() |
![]() ![]() ![]() |
Press Briefing -- December 5, 2000 Director Prewitt |
![]() |
STEVE JOST: Okay, I think we're all here. Thank you for joining us
this afternoon. My name is Steve Jost. I'm with the U.S. Census
Bureau. And this is another in our standard operational press briefings
where we report to you on the status of Census 2000 ongoing operations.
I think most of you have been part of the exercise before, but
I just want to make clear that we have a group of reporters here
in the room and a group who are joining us by phone. After the
director has some opening remarks, we'll go to questions, and
we'll alternate from the room to the phone. And, with that, I
give you Director Kenneth Prewitt.
DR. PREWITT: Thank you, Steve, and friends of the Census. As
we've just mentioned up here, the census is the longest
continuous scientific project in our American democracy. And we
are very pleased in Census 2000 to be part of that tradition and
to report again to you today that we are where we need to be at
this stage in the census process. We have finished all of our
major operations on schedule. We are now doing the data
preparation work that's necessary before we release the
apportionment numbers.
A word or two about that data preparation work, to give you
some idea of its complexity. Every data file that we're
preparing for the apportionment numbers is passed through what we
call our disclosure avoidance procedure; that is, it's our
procedure to make sure that no confidential data can be even
inadvertently released. And that's a big project, and we're
doing that right now. We've finished about 32 of the states.
We've got the rest of them to go. But we will finish that on
schedule.
Another thing that we're doing in getting the data ready is
collating what we call the collection geography with the
tabulation geography. The collection geography is the geography
that we use when we send enumerators out in the field to try to
get responses, and that geography does not perfectly map against
what we call tabulation geography, which is, of course, the
geography of state lines and county lines, city lines and so
forth.
So I'll just give you a couple of examples of why we're still
working very hard to make sure the data are exactly as they
should be before we release the apportionment numbers. The
apportionment numbers will be released the last week of December,
in all likelihood on the 28th or the 29th. That is ahead of our
obligation, of course, to have them released by the end of the
year.
A word or two about what you will see when the numbers are
released. There will be three tables that will be shared with the
press and the public at the same time they're prepared for the
president. One of the tables is the basic apportionment table,
and that takes the population of the 50 states and the overseas
military and diplomatic corps and allocates that, of course,
across the 50 states.
That does not include the population of Washington D.C., nor
of Puerto Rico nor the other island areas. So the first major
table is a table that simply describes the apportionment-based
population total. Then there will be another table that will be,
in effect, the resident population of the United States, which
will exclude the overseas military and diplomatic corps but
include the District of Columbia, and as a separate line, the
population of Puerto Rico. So that will be a different
population count.
And then we'll also produce for you a population count of the
overseas population by state, so you'll see exactly how we came
up with the apportionment number itself. So those will be the
three major data products that will be released, as I say, the
last week of December.
Now, a word or two about the census, and then I'll really
allow you to pose whatever questions you would like to. We
continue to feel very good about census operations. We've now
done a lot of quality control work, re-examined our data files,
compared them with what we expected. And we are convinced that
Census 2000 will have an unusually strong coverage; we say, that
we manage to get to a very, very high percentage of all the
households in the United States, and we think we've got forms
from a very, very high percentage of those households.
One of the ways that the numbers will be looked at when the
population count is released -- the press, other people will be
wanting to compare our population estimate with the estimate that
would come from demographic analysis. Demographic analysis, as
you know, is an independent measure of the population, which
comes from vital statistics. It comes from birth, death records,
in-migration, out-migration.
That is an estimate of the population. It has its problems.
For example, it's very hard to get a good accurate count of the
magnitude of the undocumented population, of course, in the
United States. But also, interestingly enough, it's hard to get
a good estimate of out-migration. The assumption in the United
States, of course, is that if anyone moves here, comes here, they
want to stay; nobody ever leaves.
But, of course, that's not quite true. People do leave. But
we don't have a good statistical measure of emigration as against
immigration. And so that's one of the things that makes the
demographic analysis number itself imperfect. But, nevertheless,
it is a good national estimate of the population. And we will be
comparing and the press will be comparing the census count to
that estimate.
Historically, when we've talked about the net undercount, it's
almost always been a comparison of our count to the demographic
analysis count. And when you saw that the undercount, for
example, in 1990 was 1.8 percent compared to 1980 at only 1.2
percent compared to 1970 at 2.1 percent, all of those differences
were comparing the census count to the demographic analysis
estimate.
We are hopeful, strongly hopeful, that our population
estimate, based on the census data, in 2000 will be closer than
any time in history to the demographic analysis estimate. And
that is one indicator to us of the success of the census.
However -- and this is the point I need to stress -- that is a net
national estimate. That is, a net national estimate can include
two types of what we call coverage error, two types of problems,
one of which is persons missed and the other is persons
erroneously included or persons counted twice. So a net number
can still include both an undercount and an overcount.
When we release the apportionment numbers at the end of
December, we will not yet have a measure of that undercount and
overcount. We will not have that until we complete the work on
our Accuracy and Coverage Evaluation. That work is also
proceeding exactly on schedule.
We have now finished all of the field work. We're now getting
those data files in shape. And when they're in shape, we'll
begin to compare the responses from those 300,000 households, or
including Puerto Rico, 314,000 households, to the census record
we have for those households as of Census Day. And it's that
comparison, of course, that allows us to make an estimate about
the undercount and the overcount.
So I do want to simply emphasize at this stage that when we
release the apportionment count, one of the things that it will
be compared to is other estimates of population -- demographic
estimates, our own intercensal estimate program. It'll be
compared with other benchmarks. And we are convinced that when
the data are out and those comparisons are made, it will
underscore what we have been saying, which is that Census 2000
has been an operationally successful census.
But then, to be redundant, let me say it again. That does not
mean that there will not be an under- and an overcount, and we
will not know that until we've completed the work on the Accuracy
and Coverage Evaluation. The time schedule for that, of course,
is through January and February.
We anticipate by the end of February to have finished that
statistical work, and on the basis of that, will make a decision
about whether we can improve the census by using the adjustment
factors. And we've gone over that before. We can go over that
again. We'll certainly have an opportunity to do it between now
and the latter part of February, when that decision will be made
and, of course, widely announced.
I should say, just to revert to the first point, that we are
very, very proud at the Census Bureau of having gotten this far
in Census 2000 as successfully as we believe we have. We're very
proud to be part of this, as we say, the longest continuous
scientific project in American democracy, starting in 1790 when
we were simply apportioning, of course, to only 16 states at that
time, the initial 13 and then three that were added: Kentucky,
Tennessee and, I think, New Hampshire. Is that right? Oh, no,
that's wrong. I've really caught myself now.
STAFF: Georgia.
DR. PREWITT: Georgia was added after the first 13. And then,
by 1890, you can see the migration of the forms. By 1890, we're
apportioning 357 seats in 45 states. Not yet states were
Arizona, New Mexico and Oklahoma, and, of course, Hawaii and
Alaska; and then, of course, in 1990 the full 50-state
complement, 435 seats.
And, of course, what we will be doing in that last week of
December is sharing with you, sharing with the public, the
numbers that the Secretary of Commerce will deliver to the
president that will have counted the total population, spread it
across the 50 states, and identify the allocation of those seats
in the House of Representatives at that time.
So, with that, I'm happy to take your questions. Genaro?
Q Genaro Armas, Associated Press. If Governor Bush becomes
president, how does that affect how the Census Bureau proceeds
with the Accuracy and Coverage Evaluation?
DR. PREWITT: We would hope not at all. The Accuracy and
Coverage Evaluation, of course, has been built into the census
plan from the beginning. It is part of making sure that Census
2000 is as accurate as it can be. There's absolutely no reason
to presume that a change in administration should alter the
Census Bureau's plans with respect to the use of the Accuracy and
Coverage Evaluation.
Just to emphasize, for those who may not know, the Accuracy
and Coverage Evaluation is an evaluation of the census. It
itself is also evaluated. That is, the Census Bureau will look
at the quality of that piece of work, the quality of the basic
census, and then ask itself, "Can we make that initial count more
accurate by applying the dual-system estimation from the A.C.E.?"
If the feeling of the Census Bureau is yes, we will then
adjust. If the feeling is no, we won't. And there's absolutely
no reason for that scientific decision to be made differently if
Mr. Bush is in the White House or Mr. Gore is in the White House
or if Mr. Clinton were in the White House. This has been
conducted as a scientific project, and we hope it will continue
to be so conducted.
MR. JOST: Okay, we'll go to the room again. If you're on the
telephone, if you want to get in the queue of reporters, please
press "1" on your telephone. Over here.
Q Tom Hargrove, Scripps-Howard News Service. When can we
expect to see household data brought down to the county and Zip
code level?
DR. PREWITT: The next major data product from the Census
Bureau is, of course, the redistricting data tape, which comes
out on a flow basis starting in early March -- on a flow basis,
state by state. By the end of March, all 50 states will be out.
That data tape is down to the block level. It's the most refined
-- that is, geographically refined -- data tape that we produce.
That will have a limited number of variables on it.
It's the variables that are needed to enforce civil rights
laws -- I mean, the Voting Rights Act -- make sure that boundaries
are drawn, one person, one vote. That will have the count. It
will have the age over and under 18, because, of course, that's
the voting-age population. And it will have race and ethnicity
and gender.
So as early as the beginning of March, you'll begin to see
very refined population data with those very limited variables
available down to the very, very smallest geographic level,
obviously block-level data. Then you get aggregated back up to
counties, cities and places, and so forth. Other data products
then roll out on a fairly steady basis, starting almost
immediately, as early as June. We have a handout -- I think it's
in your packet -- of the data products and the dates.
What we do is we tend to bring them out earlier at higher
levels of aggregation. That is, each stage of more refined data
down to lower places simply requires more work, and so we'll
start at higher levels of aggregation and then simply keep
driving it down for all of our variables. But a lot of the data
will be out in 2001; not all of it, but a high percentage of it.
And I should say that, unlike previous censuses, all of the
data, of course, will be out electronically, be on the Web. It
will be a much faster data dissemination than back in the old
days, the really old days, but even the recent days, where it all
came out in tabular -- you know, in bulk form, and you had to wait
until we could print and distribute our huge books full of
tables.
Of course, back in the 1790, 1800, 1800s and on, sometimes you
had to wait a year or two before you could see census data. It
took a very long time to tabulate it all. But now we'll be
talking a matter, even though it's a very complicated tabulation
process, we'll be talking a matter of weeks and months to get all
the data out.
MR. JOST: Okay, we'll move back to the telephone. Judi
Hasson with Federal Computer Weekly.
Q Hi. It's Judi Hasson with Federal Computer Week. How
are you, Dr. Prewitt?
DR. PREWITT: Very well, thank you.
Q I wanted to ask you about security. Did you experience
any hacking that was successful? And, if not, why not?
DR. PREWITT: Well, why not is because we may be smarter than
the hackers. No, we did not have any hacking problems. As I
think I reported to you before, we actually have contracts with
experts to try to get into our system, and then we use that to
probing for weaknesses. We then use that to identify if there
are weaknesses and tighten them up. And we believe that. Of
course, you never know. You're in the business; you would know
better than I. But you can't know definitively, of course. But
we believe that no one got through any of our firewalls with
respect to any census data.
As you know, we also allowed Internet filing. We did not use
that heavily because we wanted to test its security. We've now
begun to evaluate that. And early indications are that it easily
met all of our security tests, and we will be wanting to use the
Internet much more heavily in 2010 than we were able to use it in
2000.
So at least, based on the information we have, no hackers were
able to get through our firewall. The Census data are completely
protected. And we believe that will continue to be the case.
MR. JOST: Back in the room, over here.
Q John Mercurio with Roll Call. I know you have a
schedule for releasing information on reapportionment. I was
wondering, though, if there's any way to get any sense of
preliminary estimates from the Bureau on any sort of demographic
trends for that process.
DR. PREWITT: Well, you mean, between now and December 28th?
Q Yes.
DR. PREWITT: No. (Laughs.) No, there are no preliminary
data that are available at this stage. I can tell you, we keep
the apportionment count data very, very, very secure, which is to
say the director will not know it until the very last minute. Of
course, there are many things about the Census Bureau that
probably the director doesn't know. But what happens, quite
honestly, is that we do quality check and quality check and
quality check these data to make absolutely certain that there
are no mistakes or errors.
The apportionment count itself doesn't exist until the
overseas military and diplomatic staff have been added to those
counts. And that's the last thing we do.
Q [Inaudible.]
DR. PREWITT: No, absolutely not. And that will not happen
until we're just ready to produce them, because we realize the
level of public interest in these data and we are very determined
to have no preliminary data out. It would create mass confusion
in the American public if some data kind of slipped out
inadvertently earlier and they turned out not to be the most
accurate data. So we will hold very tight to those data until we
have finished all of our work, and then we are right on schedule
to do that. And that's really the good news.
You know, you don't know, when you start this kind of
operation, whether you will finish all the operational steps.
You've got a schedule, but it's not necessarily self-evident that
you can meet all of those scheduled requirements. As I just
quickly alluded to in the beginning of my comments, a lot of very
difficult stuff goes on after the data are in. For example, I
didn't mention, but the data actually come in, as you know, as
census forms, individually household-responding census forms.
But then you've also got all this data of addresses that are on
your geographic data file.
Well, the geographic files have to be brought together, of
course, with the census files. That's one of the early things we
have to do, and in the process of doing that, make certain there
are no mistakes, errors and so forth. So all of that is the work
that's going on now. But as of this morning, when I was given a
briefing on exactly where we were before this session, I just
feel very, very confident that we'll make our deadlines.
MR. JOST: We'll go back to the telephones. Paul Johnson with
the Bergen Record.
Q Yeah, hi. I had a question on the apportionment data
you guys are releasing in December. Will you be releasing that
by age or race as well, or is it just going to be aggregate
numbers for each state?
DR. PREWITT: The only thing that will be released in late
December is just a population count.
Q And that's not just voting age, it's all ages?
DR. PREWITT: It will be just how many people live in Maine,
New Hampshire, all the way to Hawaii, so forth; no, just a
population count. That's our obligation. And we simply stagger
our work to meet the current obligation, and that is our first
one. We want to make sure we get that right. So, no, we won't
be able to tell. And you won't know anything about within a
state, you know, suburb versus city, any of those things, until
about six or eight weeks later.
Q Okay, thanks.
MR. JOST: Over here in the front row.
Q I'm Elizabeth Marchak, Cleveland Plain Dealer. I'd like
to know when you're going to be releasing what was known in 1990
as the STF-3 tape file. And what will you be calling it this
time?
DR. PREWITT: We're going to call it --
Q It's my favorite.
DR. PREWITT: We're going to call it Summary File 3, because
we don't use tapes anymore. And I believe the schedule in your
packet has the month that's scheduled for release. Summary File
1 will be released in June 2001. So Summary File 3 is in line
after that. And if it isn't answered, somebody will give you the
exact correct answer at the end of this session. I want to make
sure I didn't give you the wrong answer.
MR. JOST: We'll go back to the phones. Ariel Sabar with the
Providence Journal.
Q Dr. Prewitt, I was wondering whether the Census Bureau
has made any determination on whether it would release some of
the more detailed data to the media on an embargoed basis.
DR. PREWITT: I'm sorry, do you mind saying that again?
Q Yeah, I was wondering whether the Census Bureau had made
a decision on whether it was going to release some of the data to
the media, some of the more detailed data that will require some
computer knowledge on our part --
DR. PREWITT: Oh, right, right, right.
Q -- on an embargoed basis.
DR. PREWITT: Yes. No, that's a very active conversation.
That will not be true for the apportionment count, by the way.
That will not be released in an embargoed environment. But some
of our subsequent data products, which do require, as you know, a
lot of computer preparation in order to do quick analysis; we are
working right now with -- I don't know. Steve, you may want to
fill this in. Do you want to talk about it in more? Yeah, go
ahead.
MR. JOST: We're actively meeting with the Associated Press
and major broadcast and print media to talk about timetable for
the data releases for especially the redistricting data, but more
importantly, I think, the summary file releases, which are big
numbers, a lot of crunching. And we're very close to working out
arrangements for that data to be made available to the media a
little bit in advance.
I'm not sure we're going to label it as an embargo process.
And the Associated Press will be packing it for their customers,
and any other major media outlets will be getting access to it.
And you all will be getting access to it through a major FTP
pipeline, which we don't think will cause anybody any delays or
backlogs or wait times. But we're probably a couple of weeks
away from getting the details of those arrangements out to you
all.
Okay, we'll go back to the room; right back here.
Q Carl Weiser with Gannett News Service. What recourse is
there for states, when you release your apportionment data, if
they object or don't like the numbers that you've come up with?
What can they do?
DR. PREWITT: The Census Bureau does have a process where, as
a jurisdiction, a state or even a city or county, any
jurisdiction can appeal its number. That process kicks in, I
think, starting -- not immediately; it kicks in later in the
spring, after the numbers are out. And if it's required, if we
then examine the nature of this appeal, the appeal period is, I
think, about two and a half years, a very long appeal period
during which jurisdictions can, as I say, make an appeal that the
numbers are wrong in some respects. And we then make adjustments
if necessary.
Sometimes it's a geographic problem. For example, I said
we're still looking hard at the data to make sure it's right. I
won't get the example exactly right because I just don't have all
the details. But my understanding is, for example, in one of the
states we had a prison that was out on an island. And its
address was someplace on the mainland, and so initially all of
the forms got coded to that address on the mainland, which put it
in a certain block, in a certain city. When we discovered that,
of course, we had to redistribute those back out to where they
really belonged, where the people really lived.
That kind of thing can happen. And what we've done this year
to try to minimize those problems, we have brought in state
demographers from every state that wanted to cooperate. And I
think eventually 40, 41 states did cooperate in this program. Is
that the number? That's the number that I remember. A very high
percentage of all the states cooperated in this program. Their
demographers came in and they looked at the data, as we were
beginning to prepare it, to look for anomalies. They would say,
"Well, that doesn't make sense." And then we'd go back and look
at it to try -- and the prison example is an example of where they
detected something.
We will bring them back in before we do the redistricting data
tape. We will ask them to look at anything that they think looks
strange from what they know about their state, their cities. And
these are people very, very well-equipped. So we hope to
minimize those kinds of errors by correcting them before the data
are already out. Nevertheless, there is this appeal process.
If people do successfully appeal, and we come to the
determination that we did make an error of some sort of that
type, it's almost always a geography error, we then make the
correction, and that correction then goes into all subsequent
census products, intercensal estimate programs and so forth. At
that stage, of course, we could not go back and correct the
apportionment numbers or the redistricting numbers. I mean,
those data would have already been completed.
MR. JOST: I think the details of that appeal process will
likely appear in the Federal Register in January with the kind of
proof that would be required on the part of anyone who would make
an appeal.
DR. PREWITT: Sorry, you had a quizzical look, so I didn't
quite answer.
Q So they won't be able to appeal the apportionment
numbers?
DR. PREWITT: No, the apportionment number itself, once it's
given to the U.S. Congress, it is the number. Yeah. And they
can appeal the number. And if the appeal is successful, we're
not going to go back and move a seat from one state to another.
That's what I meant. But that doesn't mean the number for that
state wouldn't be changed. And the population of that state
would recognize the success of that appeal.
This happens after every census. The adjustments are really
very marginal. They're the kinds of things I just mentioned.
It's almost always a geographic problem. You know, we thought
the boundary was here, but the city comes along and says, "Well,
actually, our boundary -- you know, we forgot to tell you, but it
moved about two blocks. Therefore, those people that you have
out in the county really belong to the city." And so we then
make those adjustments.
MR. JOST: We'll go back to the phones. Tom Ginsberg of the
Philadelphia Inquirer.
Q Hi, thanks. Just one quick question I may have missed.
When do you expect, after you've made the decision in late
February whether to use the adjustment, when will we get the
adjustment?
DR. PREWITT: Well, as soon as we make the decision, we will
start rolling out the data on a flow basis. And the actual work
that it takes to prepare a state's redistricting file is
considerable. And depending on the size of the state, it can
take as long as a day or two to just do the work, and again,
quality-check it and so forth.
So we will put, at the head of the queue, those states which
have early redistricting requirements by state law, and states
that aren't going to redistrict for another sort of six or eight
or 10 months or a year or whatever, they'll be later in the
queue. We fully expect to have every state out by April 1st, our
statutory deadline. And the earliest state will come out the day
we can get it out. And our schedule calls for us to make this
decision -- there's not a fixed date, but sometime in the latter
part of February. So by March 1st, we are starting to produce
data.
Is that adequate for your question?
MR. JOST: In the room. Yes.
Q Sean Higgins, Investors Business Daily. You said
there's no reason to presume a Bush administration would change
the policy on using adjustments. But it would be within their
authority, wouldn't it?
DR. PREWITT: The census plan calls for the collection of the
Accuracy and Coverage Evaluation to be conducted and then to be
used. Like all of our quality control procedures, we do our
quality control procedures. We then use them as best we can.
There is a federal reg that was introduced or went into law, I
think, December 4th, somewhere in that period, early December, if
I'm correct.
MR. JOST: September.
DR. PREWITT: No, earlier than that. September. No, sorry.
It's been a fast fall. Excuse me, December 4th was just now --
anyway, a federal reg that we can get you if you would like, that
stipulates the following: that the decision about whether to
adjust or not is a Census Bureau decision. There is a standing
committee that meets almost every week. It's examining all of
these data as they come in.
The amount of data to come in before you make this adjustment
decision is enormous. It has to do with all kinds of things --
response rates, match rates, housing match rates and person match
rates. When we're trying to compare the responses from 300,000
surveys to the census records, there's just a lot of stuff that
you have to look at. And that's what I say we'll start doing in
about another two or three weeks, and that takes a couple of
months to do that.
The Federal Register notice stipulates that this standing
committee, which is identified, will evaluate those data and then
make a recommendation to the Census Bureau director. And the
Census Bureau director will then make the final determination
about whether to adjust or not. This is a change from 1990, but
it replicates what was true in 1980. In 1980, the Census Bureau
collected these kinds of quality surveys, called a
Post-Enumeration Survey then, and determined that the data were
not strong enough to use, and itself decided not to adjust. That
was litigated, but the position of the Census Bureau prevailed.
In 1990, the Census Bureau did recommend that the adjustment
procedure be applied. That recommendation went at that time to
the Secretary of Commerce, and he overruled the Census Bureau.
And what this Federal Register notice does is it changes the
policy back to the 1980 policy, which is it is properly a
decision of the Census Bureau, not a decision of the Secretary of
Commerce or any other political appointee.
For an administration to change that -- I guess they could
change that -- I am not an expert in statutory law. But certainly
the Census Bureau is proceeding on the assumption that what is
now the expected practice will continue, irrespective. There is
absolutely no reason for this. As I say, it's a project. It's a
scientific project. We will complete it on schedule, do our job.
And I would hope that there will be no disruption of that.
MR. JOST: We'll go back to the telephones. Sherry Sylvester
of the San Antonio Express News.
Q Thanks. You got my question earlier.
MR. JOST: Okay. And to the room -- I think we saw a hand over
here.
Q Obviously the figure you were seeking to be at in the
1980s, was a 1.2 percent undercount. Can you give us a hint of
what the undercount appears to be this time?
DR. PREWITT: Certainly. Carefully. The hint is as follows.
The Census Bureau feels very, very good about the quality of its
coverage programs; that is, the programs that try to go out and
find everyone it could in the United States. And we believe that
the differential between the count that we'll report at the end
of the month and the demographic analysis estimate will be
smaller than at any time in history. That's what we believe. We
could be surprised when we finish the numbers, but that's what we
expect. That does not itself tell you anything about whether the
magnitude of the undercount and the overcount, and whether those
under- and overcounts are differentially experienced by different
demographic groups in geographic areas. You could have a net
number which is very close to the demographic analysis estimate,
and you could still have in that net number persons who were
missed and persons who were overcounted, and they netted out to
be close to the demographic analysis estimate. But we will not
know until we have done the Accuracy and Coverage Evaluation,
which is for us the gold standard. It is the real measure of, in
our judgment, the population size of the United States.
The demographic analysis is an important benchmark, but it has
its own flaws of the sort I just mentioned. It doesn't have an
accurate count of out-migration, in-migration. It doesn't have a
fully accurate account of course of the illegals. We are
counting the undocumented people in this country, as you know, in
the census, because we don't even bother to ask. And, therefore,
we think we get a higher response rate from them than maybe some
other ways in which the data are collected on them. If we did a
good job with those population groups they will be in the census,
so that in effect the census could have a number that's even
higher than demographic analysis. If we manage to reach further
into the population than that particular statistical procedure
does.
What we are hopeful of is that our own count -- and we have
said many, many, many times and we will say it again -- any large
statistical operation such as a census is an estimate of the
truth -- it is not the truth. We want to get that estimate as
close as possible. And one of the ways that we will evaluate it
is how close did we get to the demographic analysis estimate.
And if it's close we'll feel good. That doesn't make it right.
It doesn't make either one of them right. And what will give us
our final measure of the quality of the census will be the
results of the A.C.E.
MR. JOST: Back to the telephones. Michael Simmons with
Gannett in New Jersey.
Q Hello, Dr. Prewitt. I have a question regarding the
redistricting data being released early in March. Ten years ago,
it was released earlier than that. At the end of January was
when New Jersey received it. There's some concern --
MR. JOST: Mike, I'm sorry, you're breaking up.
Q I'm sorry. There's been some concern expressed in New
Jersey that the redistricting data will be received too late this
year for the state to be able to keep its schedule regarding its
legislative primary election in June. Is there any chance of the
schedule being moved earlier, or can you explain why it has to be
released in part?
DR. PREWITT: Yes. Obviously the Census Bureau is well aware
of New Jersey's particular needs. I should say, of course, that
-- the federal law under which we are governed obligates us to
provide the redistricting data no later than April 1st. If an
individual state, of course, for its own reasons has a law,
regulation or practice which requires the data to be available
earlier, there is not much we can do about that. We can only
meet our own statutory requirements. We are well aware of New
Jersey's needs, and they clearly will be close to the top of our
queue in terms of when we process the data. But there will be no
way that we can process any individual state's data until we have
made the decision about adjustment to all of the work that has to
go into that. This is a national census, and is not just a
state-by-state census. We have to get the nation's numbers right
before we will release them. We will do everything we can to
accommodate New Jersey's needs, but we are really finally only
governed by the federal law that obligates us to meet the April 1
deadline.
MR. JOST: Genaro?
Q Along the same lines of an earlier question, is there
any early indication of just the quality of the work done on the
A.C.E. program, and how likely it would be used?
DR. PREWITT: Yes. Well, two questions in that, Genaro. Yes,
there are some early indications of the quality of the A.C.E
survey itself, and these are very positive. These have to do
with response rates. They have to do with whether there is a
match between the housing units, for example. We are already
beginning to look at that data. We haven't done the person
match, but we can do some housing unit matches. We know that it
was done in a timely fashion. That is always better, because
that means you have higher quality data.
As you recall when we were talking about it in the earlier
period, we managed to complete a very high percentage of the
Accuracy and Coverage Evaluation survey on the phone. That is a
computer-assisted instrument, and that gives you, again, high
quality data, because all the bridging rules are right there in
front of the screen. We also use the CAPI instrument --
computer-assisted personal interview instrument -- in the field.
Everything worked well is the point. All of our field worked
well. The data have come in, they look clean. At this stage,
everything is preliminary. At this stage we are not running into
any big operational problems with the A.C.E survey.
And let me just pause for a moment. This is one major task.
It's only in the environment of a census that we don't talk about
a successful completion of the survey of 300,000-plus households
in a very short period of time as anything other than a
magnificent accomplishment. And the fact that our field people
managed to do that at the same time that they had to complete the
census was, we consider, a quite significant operational
accomplishment.
Now to your question. The fact that the A.C.E. is
operationally robust, and the census appears to be operationally
robust, means that you are well-positioned to make the decision.
If A.C.E. had collapsed, if we had a low response rate, if we had
interviewers -- we thought the data quality were no good -- then we
would have a very hard time using it. But we think it is going
to be strong enough to use. That does not mean we will adjust.
The decision about whether to adjust has to do with when we
actually look at the data are we able to significantly improve
the census. And if there are undercounts -- people who were
missed -- and overcounts, and these are differentially distributed
across different geographic areas and population groups, then we
will conclude we should adjust, because that will make it a more
accurate census. But we really cannot make that determination
until you actually look at the results of the match itself. And,
as I say, that really is not available until basically early
February.
But the good news is the survey is strong enough to sustain
that. And that -- you know, I wouldn't have known that three
months ago. It doesn't happen until it happens.
MR. JOST: Okay, a couple -- anyone back there? Oh, sorry.
Two -- Ellyn and the gentleman over here.
Q Ellyn Ferguson, Gannett News Service, and I guess I have
two questions. Dr. Prewitt, how long do you think you will stay
on as census director? Do you think that you will remain in your
position through April 1, through the releasing of the
redistricting data? And, two, do you think that there is going
to be -- do you expect to have that rule changed that shifted the
authority from making the decision on the fiscal adjustment? Do
you expect that to perhaps be put on hold through legal
challenges? Have you gotten any indication that anyone is going
to be filing a lawsuit?
DR. PREWITT: Let me take the second one first. There has
been to our knowledge, to my knowledge anyway, no legal questions
raised. There were, as you know, early when the reg was first
issued, there was interest on the part of the Congress whether --
and Chairman Miller said perhaps it was an illegal regulation
because it delegated too much authority to the Census Bureau
director. Examinations of that subsequently suggested it's very
hard to understand why this particular delegation could be called
illegal. But I leave that finally to the lawyers.
I would only say for sure, Ellen, that if there is a legal
challenge that is emerging, it has not surfaced at the Census
Bureau. I would be surprised -- it could happen, but I would be
surprised. I think that my guess is that that process, which is
a perfectly reasonable process from the point of view of how
scientific projects should unfold, should stay in place. I'd
like to point out that we're obviously going to release the
apportionment numbers as the Census Bureau. That is, the Census
Bureau is going to release those numbers. Nobody else is going
to do it. We are going to hand them over to the president, who
then, in turn, hands them -- well, we hand them over to the
Secretary of Commerce, who hands them to the President, who hands
them to the U.S. Congress. That is, nobody is coming out to
Suitland to look at those numbers before they are released. And
I have no reason to believe that the redistricting data should be
handled any differently. It's one more Census Bureau product.
If people then litigate it after the fact, they litigate it. But
at least from the point of view of our task, we simply will move
through the steps.
You ask as well about my appointment. If people aren't aware,
the director of the Census Bureau is not on a term. It is
presidentially nominated and Senate-confirmed, and serves at the
pleasure of the President. As I understand it, I could be asked
to leave by any president. I could be asked to leave by Mr.
Clinton today; I could be asked to leave by Mr. Gore or Mr. Bush,
who happens to be the President at the time. I will remain in my
position until I am asked to leave. Well, some caution on that.
(Laughter.) I mean, at some point.
But, speaking more candidly, you know, obviously, to get to
the burden of your question, I think that it would be healthy for
the census, indeed, healthy for the American society, if the
Census Bureau director were not thought of as a political
appointee who has to be removed immediately upon the appearance
of a new administration. I can fully understand that a new
administration would like its own director. We all know that
takes time. That person has to be identified, has to be
nominated, has to be confirmed and then appointed. And I would
hope for the stability of the Census Bureau and the stability of
this Census 2000 process that I would be there until my successor
were confirmed. And then I would obviously be happy to depart
immediately. And if that successor happened to be confirmed
early, then I would step aside. If it takes two or three months
for that successor to be confirmed, then it seems to me
appropriate that I would be willing to stay there.
I will do whatever I am told to do, of course, I have not a
whole lot of choice. But my own preference would be that the
Census Bureau be allowed to do its work completely and
independently of any kind of political influence, as it has tried
to do its work, and that the director be left there in order to
create stability in this process. It's no different from if you
were about to do if you are in the middle of a big scientific
project at NIH, the war on cancer or space exploration, why,
there is some value to having stability in the leadership
position while you are completing this big task. So I would hope
that I would be allowed to stay there until my successor were
confirmed.
MR. JOST: Right back here in front of the cameras.
Q Jonathan Nicholson with Dow Jones. Continuing on the
two-question track here, if I may, there are two things -- I am
relatively new to this. Is there a criteria that, a set criteria
you guys will use in deciding whether to adjust that's laid out
somewhere? And, secondly, are there different -- are there two or
three different estimates that we'll have then for population
2000? Our Census counts of A.C.E. and the equivalent of like an
inter-Census estimate, and what's the difference between these
types of --
DR. PREWITT: Surely, surely. Let me just take a word and try
to clarify some of these issues. I appreciate that they are not
intuitively obvious.
Let me go back to 1990. We did a census in 1990. We
recommended adjusting it; that is, there were about 12 million
errors in the census by our own quality work -- an overcount of in
excess of 4 million, an undercount of in excess of 8 million,
that netted to about a 4 million person difference. We
recommended that the census count be adjusted to that number.
This was not accepted by the Secretary of Commerce. Yet about
two years later, a year and a half later, the Bureau of Labor
Statistics in particular said, "Look, we don't want to try to
collect labor statistics, labor force behavior data for the next
eight years based upon what the Census Bureau has told us is an
incomplete count of the population." And so we worked out with
the Bureau of Labor Statistics that all of their data would be
collected based upon the adjusted or corrected Census numbers.
And since that day not just the Bureau of Labor Statistics -- let
me just read to you the American Community Survey, the American
Housing Survey, time-use survey, long-term care, ambulatory
medical care survey, crime victimization, hospital, home hospice
care, Health Interview Survey, prisoner statistics, school
surveys, Survey of Doctorate Recipients, Survey of Income and
Program Participation. I have got a list of that long of all of
the major surveys that either the Census Bureau does or some
other federal fiscal agency does, which are all calibrated back
to the adjusted 1990 number -- not to the unadjusted number.
That's one of the -- so I only mention that because the country
is actually quite capable of functioning intelligently with more
than one basic number. The apportionment redistricting number of
1990 was unadjusted, but all of the statistical work when Alan
Greenspan goes before the United States Congress and tells you
about the state of the economy, he is making his assessment on
the basis of data, all of which have been corrected to account
for the 1990 undercount. So that's just observation one.
And observation two is the first thing I said is the
apportionment number itself is a single number. It is the
population of the United States and the overseas military and so
forth scattered across the 50 states. So that's an estimate.
The other number we are going to be releasing even then is a
different number, because it includes the District of Columbia,
but excludes the overseas and diplomatic corps, and also as I say
has a special line for Puerto Rico if you want to, because after
all the Puerto Ricans are citizens of the United States. And if
you want to make that part of the U.S. total, then the number is
right there. So that's a different number. If you want to leave
Puerto Rico in, take it out, it's a different number. But all
the numbers are there for the press and the public to use as they
will.
Then there are other estimates of the population size. And
one I described with the demographic analysis estimate -- that is
not based upon census data; that is based upon vital statistics
and administrative records. There is also an estimate of the
population of April 1st, 2000, that is the Census Bureau's own
what we call the intercensual estimates program. And what that
program is it takes the 1990 census and updates it every month --
every several months depending -- and says, "How many people now
live in the United States?" That estimates program is based upon
the 1990 unadjusted data, for reasons I won't here try to
explain, but it had to do with legal reasons. The unadjusted
data, which means if you now go to our Web site -- you can go to
our Intercensual Estimate Program and say, "How many people lived
in the United States on April 1st?" And that number we know is
lacking the undercount numbers from 1990. But we have added
since 1990 about 25 million people of the United States
population through that program. Obviously the country has grown
since 1990. From 1990 to 1995, I think -- and 1999 -- I think the
number is roughly 25. So there are different numbers out there.
And there just always will be -- that's the nature of the
enterprise.
Now, finally, to get to your -- what I think the burden of your
question is -- is that there will be a population count of the 50
states that will be not adjusted that will be used for
apportionment. There will be about eight weeks later a new count
of the population of the 50 states all the way down to the block
level that could in principle be adjusted; that is, could account
for over- and undercounts. And that will be a new number. We
think it will be a more accurate number.
Just to be a little bit redundant about all this, the census
is a process of trying to get its numbers closer and closer to
the truth. If we had stopped the census on July 1st, we would
have had a reasonably good count of the American population. But
that's before we did about 16 different quality control
procedures, which improved that count. But we had the mail-back
in, after all, we had gone out and knocked on doors -- we had all
those data. We could have said, "Aw, we're finished -- here's our
best estimate." But we know we improve that with our quality
programs that went on through the summer. The Accuracy and
Coverage Evaluation is nothing more complicated than yet another
quality procedure. It's a particularly ambitious one. It's
statistically complex. But it is nothing in principle other
than, "Look, we have got the time -- let's try to get it better."
And so that's what we will do. We will tell you in February
whether we believe we could get it better with that procedure;
and, if so, that will become the basic number. Sorry to be so
lengthy on all that, but it's an important, big question.
MR. JOST: Okay, Genaro?
Q Just a follow-up. If the decision is made then not to
use the adjusted data for redistricting, how does that affect the
way all these other surveys are collected, how the data is
reported in all those other surveys?
DR. PREWITT: It's again quite important. It depends on how
the decision not to adjust is made. If the decision not to
adjust is made by the Census Bureau because it believes that the
A.C.E. doesn't materially improve the basic Census, then that
number will be the, in effect, statistical control for all of
these surveys. If, however, the Census Bureau believes that that
number is the better number, and some of the process disallows it
from being used, a legal or political process, the Census Bureau
would argue very forcefully that the adjusted number should be
the base number for all of these other surveys for the rest of
the decade. We would not want to cripple the country with what
we knew to be an inadequate number if we could possibly avoid it.
So you are quite right, it is possible that a legal challenge
or some other process will preclude the adjusted number from
being used in the redistricting process. But then we would very
much hope that we would be able to go ahead and do that for all
the rest of our statistical work. These surveys after all are
the basis for the spending over the next 10 years of
approximately $2 trillion, and you know, you just don't want the
fundamental numbers used for expending $2 trillion of taxpayer
money to be out of focus. And we would have out-of-focus numbers
if we believe they should be adjusted and they weren't adjusted.
But I do want to emphasize that the decision about whether to
adjust or not is a, and should be, a Census Bureau decision. It
should be based upon the best science we can bring to bear within
the timeframe that's available to us, which is, as I say, late
February or early March.
MR. JOST: Any closing remarks, director? I don't see any
more hands up.
DR. PREWITT: No, thank you. Well, we'll probably reconvene
right around -- right after the release of the apportionment count
itself. We'll obviously have a press event then to do any
further explanation of the numbers that might be useful. Thank
you all very much.
Q The day of, or are you talking --
MR. JOST: Day of.
DR. PREWITT: Try to do the day of almost immediately. Right.
Q Do you have an actual date?
DR. PREWITT: Well, it will -- we expect it to be the 28th or
the 29th. And I might say this: We are ready -- will be ready --
but we are obviously having the conversation with the White House
about their own judgment about when they would like to receive
these data.
MR. JOST: A couple of housekeeping duties. In your packet,
this is a table that was just cleared and approved by the Census
Bureau -- how the data on that date will appear. The three tables
that will be part of the press release, showing the apportionment
population, the overseas population and the resident population.
Also in your packet is a CD-ROM of historical footage from Census
2000. And I bring these points up -- if you are not with us here
today but you are on a telephone, if you call our decennial media
relations office at 301-457-3691, you can get your own copies or
additional copies of both of these, and the other materials that
were in the press packet here at the Press Club. Thank you very
much. We'll see you near the end of the month.
[END OF EVENT.] |
![]() |