Press Briefing -- September 5, 2000
|STEVE JOST: I'm Steve Jost from the communications office, and we are here, as most
of you know, for our standard operational update on the status of Census 2000 as of this moment.
I think most of you know the ground rules. The director will have a few opening remarks, then
we'll open it up to questions. We'll alternate between the room and those reporters who are with
us by phone. If you will, before your question, identify yourself and your affiliation, and we
have folks with microphones. If you can, wait for them so that we can get good audio on your
question for the listeners and viewers.
With that, I give you Director Kenneth Prewitt.
DR. PREWITT: Thank you, Steve. While you were at the beach, or wherever, we did not
hesitate or slow down our work on the census. Indeed, during the month of August, we believe
we made a good census a better census. However, we're not finished, and we know you'll be
needing to file stories over the next several months. So, we want to make certain that you know
the amount of work that's yet in front of us.
There really is much yet to be done, but we are on schedule. The press kit includes a list
of our major field operations. All of those now have been included, with one very, very tiny
exception. Also, you've expressed interest in the past, of course. We're interested in what we
talked about as the final response rate. It turns out, we will not be able to calculate that for
another couple of weeks. This will be our final analysis, if you will, of the percentage of the
housing units that mailed back the questionnaire, or filed over the Internet, or completed a form
over the telephone or used the "Be Counted" form. Those rates will be published probably
within two weeks. I will have a press conference at that time, and that will take the response rate
down, as you know, to the state and smaller areas.
I can report provisionally on that. We did receive many more questionnaires and forms after our April 1st deadline than we had in any prior census; a result, we believe, of the continuing impact of the advertising campaign.
Let me quickly update you on the current operations. As I just said, during the latter half
of August, we completed nearly all of our quality count field operations as listed in the press kit.
Throughout these operations, quality count has been our motto. It's a very major part of how the
Census Bureau does try to conduct its census. It's really business as usual. One of those
operations, we confirmed whether we had accurate information for housing units identified as
vacant, and we added new housing units that came in after our Master Address File was
completed. This was our Coverage Improvement Follow-Up operation, a quite extensive
We called all those housing units, whether it was a question about the number of people
supposedly living in the housing unit, or where there was a large family - that is, more than six
persons in the unit. We did a boundary validation program. We provided local officials with
maps showing the boundaries of their jurisdiction, asked them to correct them. About half the
jurisdictions did reply, and we're now out doing that work, that annexation boundary correction
The total number of housing units that we visited after our major field operation Non-Response Follow-Up was close to 10 percent. Those units got a revisit during the latter parts of
the summer as part of our quality count operation. All those operations are now complete.
We're beginning to close our local offices. We closed 10 of them toward the end of last week,
and about 160 more are scheduled to be closed the middle of this month. We close the offices on
a flow basis, and all of them will be closed in the mid- to late October period.
As I've said on numerous occasions, the census is an estimate. It's extremely difficult to
know the true count of the American population on any given date. What is at issue, then, is how
good is this estimate? And we believe that the quality counts operations have taken the initial
estimate that was based only on the mail-back and Non-Response Follow-Up and basically
improved it. That is what I mean when I say that we think we've taken a good census and made
it a better census. We've gotten the estimate closer to the truth.
We're not finished. What do we do next? Well, obviously we turn to our data capture,
processing and tabulation effort. Here I want to pause for a moment, just to put it in some sort of
context. For more than a century, the census has relied on technology, of course, to make a large
job manageable. Not only has the Census Bureau used state-of-the-art technology throughout its
history, but it has promoted technological improvements. Just for your edification, for fun, we
thought we'd show you a bit of that today.
If you go back to 1880, we only had 50 million people to count, but the first results from
that count took well over a year to produce. And indeed, most of the data dribbled out over the
entire decade. It was widely described, at the time, as a major problem that the Census Bureau
took so long to generate the results. In that environment, a young mechanical engineer named
Herman Hollerith was concerned about the delays in producing the count and began to work on
tabulating machines. Indeed, the census office used the first tabulating machines in the 1890
census. This cut the tabulation in 1890 by a year. Hollerith was a census employee. He
continued to innovate. Indeed, what you see on the far right is a picture of the first Hollerith
assembly of machines that were used in the 1900 census.
This is part of our apparatus, if you're interested in looking at it. This is the 1900
version. The 1890 version, there's a photo over here of it, is even more primitive. This was also
used in the 1920 census, this calculating machine. If you put this machine and this machine
together, if you marry them, you begin to get the rudiments of computers because you begin to
get both tabulation and data processing. Indeed, Mr. Hollerith, in 1904, left the Census Bureau
and went into private business producing the tabulating machines. He then partnered with
another firm that was producing business machines, and in 1920 that became IBM. So it really is
the linking of these two types of technologies that created what we now know, of course, as the
We continue to innovate at the Census Bureau, improving these technologies. This is
1920, and even if you take a quick look at the picture, you'll see on the desk a somewhat
improved version of the tabulating machines. And, of course, over here is 2000, one of our
current work stations, data capture stations. In 1950, for example, we shifted to Univac. Univac
was the first commercial computer, and it was developed according to the Bureau's
specifications in order to process the 1950 data. It's now on display at the Smithsonian.
Today, of course, as you know, we made a huge investment in data capture technologies,
and particularly we're interested in digital imaging -- optical character recognition to recognize
handwritten answers instead of just filled-in ovals and boxes. As you know, this all works very
well. Our data capture centers process about 3.3 million forms a day, 23 million a week. A huge
operation: 7 days a week, 24 hours, of course, trillions of processing operations. All these
materials have now been shifted to Bowie, Md., which is our central processing unit.
We have prepared about 8.5 terabytes of capacity to process and tabulate the Census 2000
results. About 2 terabytes are actual data. The remaining capacity is for all the file merging and
quality checking that is required. Now to give you some idea of what 8.5 terabytes of capacity
consists of, I'll give you a couple of examples. The printed collection of the U.S. Library of
Congress is 17 million books, and that takes about 10 terabytes. So, our capacity will be
approximately, or slightly short of, the capacity of the Library of Congress. NASA and the
National Science Foundation produced a survey of the nighttime sky, and that takes about 4
terabytes of storage. Or the entire capacity of the Web in 1990 was estimated at about 15
terabytes, so our capacity is roughly half the size of the total World Wide Web.
These are very big data sets. It gives you some idea of the magnitude of what we're
about to do. Our processing system has total redundancy. The clusters, the CPUs, the power
systems, the controllers, the network between the sites, the air conditioning and full backups of
all of our work are done weekly. Incremental backups, of course, every day.
We've come a long way since 1880, when it took almost a year to process the census and
release the first data. It no longer takes that long, but you can understand why, nevertheless, it
will take us several very busy months to tabulate the characteristics of 275 million people, 120
million housing units, and so forth.
We've gotten mail-back forms, enumerator forms, blank forms, late-mail forms,
undeliverable as addressed forms, "Be Counted" forms, group-quarter forms and more. And, of
course, we do quality checking work on all of that.
Now, part of what makes it difficult over the next several months is that we really are
doing, as I said before, two separate operations that have to get linked. One operation is the
count -- that is, the data on the questionnaires, and the other is the address file. That is, we have
to put together the people that we counted with a location. It would be a vastly easier task if it
were just the count. But the decennial is in the Constitution for a very specific reason. Our
representative form of government is geographically based. That is, we do not elect
representatives from estates or social classes or demographic groups. We elect representatives
from geographic areas. Therefore, to fulfill our constitutional obligation, we have to take each
one of our 275 million people and put them some place. And that's the work that will be going
on over the next several months. It's extremely difficult and complicated work, as you can
We've started it now; it's on a flow basis and will continue through the months of
September and October. What we then produce is something we call our Census Unedited File.
It's our census count file. It's unedited because we are still doing edits on some of the individual
questionnaires, but we don't then change the count. This Census Unedited File, as we call it, is
the basic file from which we produce the apportionment counts, and it is the basic file that is the
platform from which we then do the Accuracy and Coverage Evaluation.
With respect to the Accuracy and Coverage Evaluation, using 8,800 field representatives,
we've now contacted 314,000 housing units that fell into the A.C.E. sample survey. And we, of
course, do a computer match of the census file on the A.C.E. to detect differences; and this is
followed by a check conducted by groups of increasingly expert matchers. If we find
differences, we then send workers into the field.
Thus, although the data collection and follow-up operations are complete, there will be
some additional follow-up work in the field that will continue well into November. But the point
I want to stress, then, is the census is far from over, and our legal deadline is approaching.
Thousands of programs have to be run on these enormous data files before we can produce the
state population numbers.
As you know, in late December we release the first results of Census 2000: the total
population of the U.S. and for each of the 50 states, plus the District of Columbia and Puerto
Rico. And these numbers, of course, have to be delivered to the president prior to December 31st.
Then three months later, in March, on a flow basis, we release the number that states will use for
redistricting state legislative districts. These, of course, are the total population numbers
classified by race, gender, age and ethnicity -- all the way down to block group and block levels.
It's all of this work that's been completed and now being data processed that makes us
feel quite confident about the quality of the census operation. But I do want to stress that we are
working with huge data sets, running thousands of programs. So, there are still challenges to
make sure that we get the work done by our targeted deadline; that is, by the end of the year.
I want to conclude with just one other, more general, comment. This is the electoral
season and, of course, there is a lot of attention being given to the relationship of the census to
the election. I think, sometimes, the larger way in which census data connects to the democratic
system is not fully appreciated. Democracy rests upon the notion of electoral accountability, and
electoral accountability is a very simple idea. What it really says is that there are competitive
elections, that offer to the electorate alternative portrayals of how well the government in power
has performed or will perform against future challenges. So, challenges for political office
present themselves in terms of their past achievements, and their promises of future
accomplishments with respect to managing the economy, protecting the nation's security,
enhancing the national well-being and so forth. Voters then elect, re-elect, or evict. It's a very
simple model: the theory of electoral competition and democratic accountability.
Well, this very simple model presumes that there is some way of assessing government
performance. If you think about it for a moment, there are really only two ways to assess
performance. On the one hand, anecdotes or personal experiences; and on the other hand,
statistical trends and social indicators. A healthy democracy uses the national number system to
indicate whether the economy is growing or stagnating, whether education or health or housing is
improving, whether crime rates are down, or the environment is being protected, and so forth and
so on. These trend lines tell the voter how well the country is doing. Upward and downward tics
in a statistical series become grist for assertions and rebuttals about who can take credit for
improvements, or who should be blamed for failure.
That is, the statistical portrait of society really is a public window into government
performance. That is part of what makes it possible to have a democracy in a large, complicated
country like ours. A good census is actually basic to the nation's number system, and the
nation's number system, in turn, is the underpinning of our system of democratic accountability.
So what I say to you today, and I'm comfortable saying it, is that we have had a good census.
We're continuing to work to make it a better census.
The stakes are very high because the stakes are not just for a democracy, just in terms of
the most basic issue of drawing boundaries, but they are also how well the country can
understand how well it's being governed; and that is a picture and a portrait and a set of analyzes
that extend into the entire next decade.
So I am very pleased to be able to report that we did use the latter part of the summer to
improve the basic census that we finished earlier on. We do think we took a good census and
made it a better census, and we think that our data tabulation, our data processing work over the
next several months, will further strengthen the quality of the numbers that we can give to the
So with that, I'm delighted to take your questions.
Q What exactly are the findings that are contained in the census that are going to tell
people how the economy has performed over the past 10 years?
DR. PREWITT: Well, fair enough. I did use the word "nation's number system." The
nation's number system, after all, is a series of ongoing surveys -- unemployment surveys,
government productivity, housing starts, the whole array of things. All of those in turn use the
census data as their sampling frame. The census is a platform to produce ongoing surveys about
economic behavior, social behavior, housing conditions, health conditions and so forth. And
that's why I say the nation's number system is really what's at stake here.
Without a good census you could not have a good number system. Is that fair enough?
MR. JOST: We have a call on the telephone, Evan Osnos of the Chicago Tribune.
MR. OSNOS: Dr. Prewitt, two questions, if I could, please. First, do you expect to issue
preliminary numbers any time this fall? And two, if you could, give us an update, please, on the
status of the various investigations that have been requested by the subcommittee.
DR. PREWITT: With respect to the first question: No, we will not produce preliminary
numbers. As many of you know, because many of you wrote the story, we do produce what we
call intra-censal estimates, population numbers, demographic characteristics. We did that
recently, based upon Current Population Survey data as well as on our inter-censal estimate
program. But the Census 2000 numbers themselves will not be produced in any way, shape, or
form prior to the apportionment counts. As I say, that will be in the latter half of December.
The reason for that, of course, is that we do take a lot of care and attention with those
numbers, and we simply do not want any of them out until we're as certain as we can be that we
have done everything we can to make them accurate.
With respect to the investigations, I'd be happy to have a more specific question on that
because there are a number of them, as you know. Is there anything particular in mind, or do you
want me to just go down the whole list?
MR. OSNOS: I suppose the most prominent ones would be the GAO's investigation of
the e-mail. I'd be curious to find out how that's progressing. And also the investigation of the
event sponsored by Tavis Smiley in L.A.
DR. PREWITT: Yes. With respect to the e-mail inquiry: Those of you who are not
familiar with that may recall that there was a congressional hearing some time in the middle of
the spring where a particular e-mail that had been sent by someone in the Los Angeles office
included a sentence suggesting that these results not be shared with the GAO. The reason for
that was because he was including in his e-mail, information from other local offices, and the
rules were simply that no data should be shared about any other local office by a particular local
office. That is, all of those data had to be shared at one level higher.
As you know, we have constantly given GAO all kinds of information. Indeed, as I've
testified, the total amount of information given to GAO is a terabyte. Now we all know what a
terabyte is. It's very large. In that testimony, I think we made reference to the fact that it was
roughly 50 million D.C. Yellow Pages worth of information. And we were very pleased that
GAO did not feel like that there's any problem whatsoever with access to census operation
Nevertheless, because of this particular phrasing in one e-mail by one employee,
Congressman Miller felt like perhaps there was something of a broader nature that had to be
investigated. As you may recall, he made reference to the fact that there were dangerous people
in senior positions at the Census Bureau. That then started this conversation about an e-mail
investigation. This is four or five months ago, that all of this conversation started.
We have been approached by GAO about the possibility of running our e-mail traffic.
We don't yet know the magnitude of it. The outer boundaries, of course, will be roughly 60
million e-mails -- because they are talking about a year's worth of e-mails. And if you were to
do roughly 4.5 million e-mails a month, that would be the outer boundaries. But we have no idea
at this stage what portion of the e-mail traffic they'll want to investigate. We're waiting. We've
been waiting for some time, since the charge that there were dangerous people is now, as I say,
somewhat dated. We've been waiting for some time for more specific instructions, and we're
more than happy to cooperate with the GAO as best we can in this investigation. But I can't give
any details because we're simply waiting for the next step.
We certainly met with the investigators whenever they've wanted to, and we talked to
them about what we have in our files and what could be made available.
With respect to the investigation of the incident in Los Angeles, we have not, ourselves,
been directly contacted by any of the three investigatory agencies. That's the Inspector General,
Department of Justice and the GAO. We have, ourselves, written a letter, of course, to all three
of those agencies explaining what happened, and what happened was completely innocent, had
nothing to do with partisan activity. But we don't know what sort of an investigation they have
in mind, if any. And as soon as we know that, we will be happy to share that with you.
Q I'd be curious to hear any assessment you might have on the whole issue of sampling
and whether your inability to use sampling because of the Supreme Court decision has had a
negative impact on the census.
DR. PREWITT: Well, the issue about sampling is, of course, a long, complicated story;
and I won't tell the entire story. It starts on the assumption that it's impossible to count everyone,
that there is an undercount, or undercoverage, as we call it. And this undercoverage is not equal
across all geographic areas and all demographic groups. We've been working on this for about a
half century. Over the last three censuses -- that is, 1980, 1990 and now 2000 -- we have
gradually perfected what we call dual system estimation, which is our way of using a sampling
survey, Accuracy and Coverage Evaluation, in 2000 to improve on the basic count.
It's the Census Bureau's position, as I said just a minute ago, that this census is an
estimate of a true count, and every operation you can do to move that estimate closer to the truth
improves the census. We believe that employing the dual system estimation moves that estimate
closer to the truth. That is, it makes it a better census. The Supreme Court ruling had to do with
an interpretation of the statutory language suggesting that we should not use sampling
methodology for the apportionment count. It did not then rule out the possibility of using it for
other purposes. The Census Bureau, therefore, believing that it will make a better count, intends
to use the Accuracy and Coverage Evaluation, dual system estimation, for subsequent census
products after the apportionment count; unless in its own judgment, it believes it doesn't make
the count better. That's a decision that we will make in February/March, when we're actually
looking at the results of our Accuracy and Coverage Evaluation.
To specifically answer your question, the Supreme Court decision did not materially
affect how well we did the census. It will reflect the quality of the apportionment numbers. We
believe the apportionment numbers will be less accurate than they would have been had we used
the dual system estimation for producing those numbers. On the other hand, we know that a very
big, important task associated with census numbers goes beyond the apportionment counts, some
of which I just alluded to, a fundamental underpinning of ourselves as a society over the next
decade, and we very much hope that the 2000 census will reflect that improved technology to
make those numbers stronger than they otherwise would be. But in terms of operations, the
Supreme Court decision did not, in any serious way, impair our capacity to do a basic good
MR. JOST: We'll go to the phones again. Judi Hasson of Federal Computer Week.
MS. HASSON: This is Judi Hasson, from Federal Computer Week. Dr. Prewitt, I have
two questions for you. What is your prediction for e-filing in light of your experience this time
around; and two, how would you assess Lockheed-Martin's performance?
DR. PREWITT: With respect to electronic filing in future censuses, we think we will
make much, much heavier use of it than we certainly did in the 2000 decennial. Indeed, some of
you are already following our work on what we call our ACS, or American Community Survey,
which is taking the long-form data and rolling that over a monthly sample survey. So we're
collecting long-form data on a continuous basis through the decade. And we have major work
connected with that survey going on now with e-filing.
As you probably know, we do heavy e-filing already in our economics statistics, but we
haven't used it as much in our population-based survey work. But our experiences were certainly
positive enough. We learned a lot. We learned something about encryption, which we had to
learn, of course. We were very, very concerned with confidentiality of the data, and we learned
something about how to present e-filing to the American public. So we feel confident now that
we can greatly expand our e-filing work.
With respect to the Lockheed-Martin work on our data capture centers, we were pleased,
extremely pleased. They, after all, got accuracy rates better than 99 percent. That was higher
than our target. They did it on time, on schedule. Lockheed-Martin, as well as a number of
subcontractors -- TRW, was a major part of the data capture system as well. Another major
contractor was EDS, which was doing our telephone work.
In general, we had extremely good experiences with our major contractors, including
Lockheed. Lockheed itself, of course, is expressing its own judgment about how well it did, and
I'll direct you to their own work, but I have no reason whatsoever to call into question their own
statements about their levels of accuracy and timeliness.
Q Dee Cohn, Washington Post. I'd like to be greedy and ask three questions. The first
is, is it too late to call in and ask to be counted if you don't believe you were? Second, on the
issue of the Republican investigation of the local offices that they say were rushing the job, do
you have any update for us on the files from those offices that you thought you might have to
save at a cost of several million dollars? And thirdly, is that investigation, or any other, slowing
or compromising your ability to complete your work on the basic count and stay within your
DR. PREWITT: The first question has to do with whether people can still be included in
the census; and the answer to that is, it depends. And the reason it depends is, we are now
beginning to process state level data on the flow basis. No state has yet been processed, but we
do begin to do it on a flow basis. So if somebody picked up the phone and said "Can I still be
counted?", it would partly depend on where they were.
At a certain point, we do have to stop the census. At a certain point, we have to say "It's
finished." We have now put everybody in it that we are able to put in it, separate, of course,
from the A.C.E. adjusted number. But fundamentally the basic census, April 1st was Census
Day, and there will be a time when we are unable to produce the apportionment numbers unless
we say this is the end of adding new cases.
As of today, no state has completely closed down; but that could certainly begin to
happen very soon now. That is, in a matter of days. So the answer to whether somebody could
still get counted in the census does, in part, depend on where they are. We are still making every
effort. A lot of the things that are listed on this handout sheet in the press kit, a lot of those
operations, did include going back to some 76,000 addresses that we thought maybe we'd missed
on our address file. And we went back to those addresses which we found in different ways. So
there are a lot of ways we went back out.
We get letters and phone calls even now. I ran into someone the other day in New York,
an acquaintance who said, "I wasn't counted." I said, "Oh, I feel awful", and so forth, and so on.
Well, I came back and I checked, and they were counted. People forget they sent it in. They
don't know their spouses did it, or sometimes they never are home so we went to proxy. And I
was very pleased then to be able to e-mail her back and say, "Well, yes, you were." So there are
people who do not think they are counted, who were.
That does not mean, of course, that we found everyone. That's why we want to do the
Accuracy and Coverage Evaluation. So it's a complicated answer, Dee, but because it's slightly
ambiguous, but there has to be a day when it is too late to be in the census. That date is not yet
today. But certain states and certain locales, it will start happening very quickly.
The second question had to do with the inquiry by the Subcommittee of about 15 offices
which they felt had not received all of the procedures that the census has to, indeed, use. We
went back and, of course, did our own investigation of all 15 of those offices, and found out that
in each case all of the procedures, all the steps, indeed, had been done. They were fully easy to
explain, the issues of what the delete rates were, what the other kinds of issues were, and so
forth, and we presented all of that, of course, a month or so ago.
In that process, as Dee Cohn just mentioned, we were asked by the Subcommittee not to
destroy any forms. Our paper forms, once they're electronically captured, we do shred them.
We, at that time, had already shredded all of the mail-back forms, but very few of the enumerated
forms. So we stopped the shredding of all enumerated forms, even though they'd all been data
captured, and they are now in storage in Jeffersonville, which is our major data capture, data
processing center. They will be held in storage until we're told that we no longer have to store
As to the state of the investigation of those, we have not heard back since we reported
back to the Subcommittee, I think now, maybe, three weeks ago. On our own investigation, we
briefed the staff on each of those 15 offices, and we've not heard back, and we will wait until we
hear. In the meantime, we will hold all the forms in storage. A small number of the enumerated
forms had already been shredded because that was the process that we were engaged in, but the
bulk of them are still available. They're on pallets, they're in a warehouse. It's not easy to find
them, any particular one you want to look at, but it could be done at some considerable cost.
The third question really had to do with whether those particular investigations, those and
other investigations and concerns and so forth, did impair the census. Here's the answer, as best
I can give it to you honestly. The Census Bureau spent a lot of time during what was a major
operational challenge responding to the oversight apparatus, to put it that way, to the Census
Subcommittee, to the Inspector General, to the GAO, to the Census Monitoring Board. A lot of
Now, the same people who would have been doing the census were doing that. At some
point, I was spending a third of my time simply preparing materials, otherwise responding to the
various questions and inquiries from one or another of the oversight agencies. I have to tell you
that, in retrospect, I'm glad I did, and I'm glad we did. I think it was a price to pay, has been a
price to pay, to prove that this is a transparent census. It certainly deflected operational attention
from operations. It just had to. We did not staff up. We did not staff up to handle the magnitude
of requests that we got.
As you know, I've testified somewhere between -- I'm not going to count -- 12, 14
times. I don't know how many times General Powell had to testify during his big operation, or
anyone else, but I think probably I've testified as much as any agency has ever had to, while
you're actually trying to do the job. We did lots and lots of letters. We're still getting lots and
lots of letters. We're responding to those.
I do want to say that because of the atmosphere in which Census 2000 has been
conducted, it's extremely important, it seems to me, for the American people and its leaders to
have confidence in the census. And if the price to pay for engendering that confidence is to be as
open and accessible as is humanly possible, then I am glad we did it, even though there were
some tradeoffs with our capacity to continue to do our operations.
I do not think that it impaired the census in any serious way. I wish we had had more
time to simply focus on the census. But I think at the margins, if it did impair the census, the
payoff of proving that we were transparent and that we did care, that we are committed to an
open process, it was worth the price, and I would continue to pay it. As some of you know, in
just less than a month, we have a major meeting at the National Academy of Sciences. We're
going to lay out all of the processes that we're going to use in the Accuracy and Coverage
Evaluation. We prespecify every one of our steps. That's very unusual behavior. No statistical
agency has ever prespecified every one of its statistical operations months ahead of when it does
It's actually not necessarily the best way to conduct a statistical operation; but, again, if
we're able to do that to prove that you can trust these numbers, then we think we should do it.
Just quickly on that; to go back to what I said earlier, good numbers are really critical to a good
democracy. If the numbers aren't trusted, if the people don't have confidence in them, then you
really do damage. Not just to the numbers, but you damage the capacity to govern and the people
to hold their government accountable. So, I think we have to do everything possible to maintain
the level of trust and confidence in the Census Bureau that I think we've earned over 200 years;
and, therefore, we're willing to be accessible, be responsive and provide the information.
MR. JOST: Back to the telephone. Dave Suto of the Oregonian.
MR. SUTO: Yes. A question about the sampling process. Are you still planning on
releasing both adjusted and unadjusted numbers?
DR. PREWITT: Yes. The way the law reads is that we're obligated to produce these
numbers simultaneously. We have written to all 50 states, saying that it would be our intention,
if the adjusted numbers are as we expect them to be -- superior to the unadjusted numbers --
we've written to all 50 states saying, if you want the adjusted numbers, then here's what you do.
And all 50 states have now replied. They want the unadjusted numbers as well. So we'll be
reproducing those, in effect. As soon as they get one set, they will get the other set. That is by
virtue of the law that was passed in 1998.
If the Census Bureau decides that the adjusted numbers are, for some reason, inferior --
that is, what would have been the adjusted numbers, are inferior -- that they've got their own
problems, this is an issue we'll be talking with the National Academy of Sciences about. Then,
it's less clear what we would do, as a matter of fact. So I don't want to be completely trapped
into saying that, under all circumstances. Because if the Census Bureau feels like any set of its
numbers are deficient in some important way, it doesn't produce them. We certainly don't
expect that to be the case, but I just want to be a little careful about how I answer your question.
But, certainly, if we do produce the corrected numbers, we will also produce the
MR. HARRISON: Charlie Harrison of Hispanic Link News Service in Washington.
What's your latest measurement of success on how well you were able to reach the traditionally
undercounted communities of Hispanics, immigrants and others?
DR. PREWITT: Two things to that question. First, when we produce the final -- what
we call our final response numbers -- which, as I say, is in approximately two weeks, we, to the
best of our ability, will produce those in a way that allow you to see what particular kind of
census tracts did well. Now, the problem is that the information on the ethnic or racial makeup
of a census tract is based upon the 1990 census. We'll explain that when the time comes. The
point is, you can't use the 2000 data to assess the 2000 census until you've finished tabulating
and processing the 2000 data.
But we will certainly be looking at that number, and we expect it to show that the
hard-to-count populations participated in the mail-back part of the census at higher rates than we
would otherwise have expected. And that's a good sign.
Secondly, we have a lot of indirect evidence, anecdotal evidence, if you will, about the
responsiveness when we went back out and did Non-Response Follow-Up. However, and it's
extremely important for everyone to understand, we can have a good census operationally, in
terms of levels of public cooperation, and still have an undercount. The undercount is something
that happens in the last 2 to 3 percent of the population. It's not something that happens in the
first 97 percent of the population. So you can do a very good job on schedule, on budget, with
high levels of cooperation, and count 96, 97, 98 percent of the population. But you're now down
to those last hard cases. And if those hard cases, which are hard to count, are differentially
distributed across different social groups, geographic areas, you will have a differential
So the true answer, the only full answer I can give your question, sir, is that we will not
know that until we have finished our Accuracy and Coverage Evaluation. That is our big
comparative survey work that tells us. So we cannot give you a definitive answer to how well
we covered the population, counted everyone, including the hard-to-count, until we have done
our own big evaluation of our work, which is the Accuracy and Coverage Evaluation. That is not
available, then, until next March.
MR. JOST: Back to the telephones. Maria Vega of El Diario.
MS. VEGA: Dr. Prewitt, I would like to know what's your assessment of the partnership
program and census work with community organizations.
DR. PREWITT: Well, we believe that it was extraordinarily successful and important.
That is, when I say it's been a good census, I do not mean at all to suggest that it's simply been
the result of the Census Bureau's work. What really made it a good census was the level of
really deep engagement and commitment by thousands and thousands of community activists,
social groups, churches, businesses, mayors, complete count committees, and so forth, across the
country. We're now, as best we can, trying to say thank you for the kind of job these groups did.
But I think there is no doubt, and we again don't have good systematic evidence, though
we are conducting evaluations of our partnership work; and those evaluations will, of course, be
shared fully as they're completed. But there is every indication that the partnership effort on the
part of communities all across the country really did make a difference in the level of
Let me just give you one example. Hialeah is the one area which we have to recount
because of our own concern about the data, whether the data had been compromised in any
important ways. I went down to Hialeah, met with community leaders, met with the staff. This
is a community that had already been counted. Most of those people had been included in the
census correctly; but, nevertheless, we felt like, because of some concerns, we'd better do it all
We went down, and we got extraordinarily high levels of cooperation yet again. In fact,
both mayors representing that part of the country went on television and said: look, you know,
it's really important. We know you've already done it; do it again. And the Hialeah recount
went right according to schedule, high levels of cooperation, and, we believe, an accurate count.
I just give that as an example.
I think that would have happened in most places, if we'd had to have done it.
Fortunately, we only had to, recently, in one of our local offices. But we did a little recounting
also in Chicago. Again, high levels of cooperation. That doesn't mean we didn't run into all the
standard instances where people said, I refuse to cooperate, and they tore it up and they sent us
$100 saying, we'd rather pay the fine than do it. But this is a very small number against the 120
million households in the country.
The overwhelming response of the American people was they believed in this; and they
wanted to do it, and they did it right. And that's why we, with some confidence, called it a big
civic ceremony because it turned out that's what it was.
Q Capitol News Service. Are there any new trends that have emerged with the Census
2000, like the massive growth of minorities, Hispanics and Asians, vis-a-vis the white
population? Any new trends that have been known so far?
DR. PREWITT: Yes. Again, the Census Bureau conducts more than just the decennial
census data, and we've been conducting major population surveys on an ongoing basis. So most
of what you're seeing in the press today about the changing demographic makeup of the
American population, the fact that we reported just last week that California now appears to be a
minority majority state -- that is, there are now fewer non-Hispanic whites in the state than the
rest of the racial groups put together. Those are all produced by our ongoing population surveys.
There are big, big trends out there, very big trends out there that we are already beginning to see.
What Census 2000 data themselves will do will refine those trends. That is, they will
take it all the way down to very small communities. So what we now know to be true of
California, or large cities in California -- Los Angeles, San Jose, San Francisco, and so forth, or
counties in California -- we will begin to know at very detailed levels. That's what Census 2000
data will produce. It will not markedly show us a different picture of the country than we already
know from our survey work, but it will give us a much more fine-grained analysis of those
Look, there is no doubt that Census 2000 -- partly because it is so visible to the
American people -- there is no doubt, when the data come out, people are going to take a harder
look at it than they normally look at just our other survey data, and they are going to see a picture
of a rapidly changing --
It's the first country in world history which is literally made up of all other parts of the world.
That is, we're still getting immigrant flows from all across the world: from the former Soviet
Union, from the Middle East, new immigrant flows from Africa, of course, as well as Latin
America, Southeast Asia. And taken together, Census 2000 is going to paint this picture in, I
think, a very dramatic and important way for the country to deal with as it moves into the new
Q Did I hear you correctly to say that all 50 states requested that they get both sets of
DR. PREWITT: Correct.
Q If a state had declined that, would those numbers have still been made available?
DR. PREWITT: Well, they'd be made available. We wouldn't have officially
transmitted them. We'll put them up on our Web site. The Census Bureau puts all of its data
products out in that sense, publicly. But what I meant is the state -- we, by law, we have to send
I think five sets of the data to each of the 50 states, to the governor's office, to the majority and
minority in both houses, the five sets of data. We simply wrote a letter asking, do all of these
five units want both sets of data? If the answer is yes, then you get them. We would certainly
make the data available.
Q -- OMB definitions of metropolitan areas, when they emerge, come together, with
the 2000 Census data? And the last piece you mentioned earlier, on the long-form data there
might be some elements that you would not be able to report? Has that percolated far enough to
make any statements about that?
DR. PREWITT: Yes. Let me skip the first of those two questions, the metropolitan area
question. I'm not an expert exactly on how that's all going to happen, and let me direct you to
Kathy Wallman. That's an OMB issue, not a Census Bureau issue.
With respect to the long-form data, this is a question that, of course, came up during the
last week of March, when there was a lot of concern about the long form, and we were quite
worried about levels of public cooperation. We did get a differential return rate, as we well
know, between the long and the short form. However, we went into the field and knocked on
doors. You do have to work harder to get people to cooperate with the long form, but basically
they did cooperate.
We do not yet have systematic evidence. Your question really has to do with what I then called item non-response. That is, when we get patterns in the data where they only answered on the long form just the short-form questions. It does not appear as if that's the case. I do not have definitive answers to that, but of the things that I'm most concerned about as we finish the census up, that's simply not high on the list. It does appear that we were able to get enough cooperation.
It's again one of those things where the expression of concern about the long form turns
out to be a very small number of people who made a lot of noise. But it wasn't. They didn't
speak for the American people. When the American population actually had the forms in front of
them or had the enumerator there trying to talk them into it, our best guess is they cooperated at
approximately the same levels that they cooperated in earlier.
One of our indicators of that is, we do have the American Community Survey in the field
at the same time. We'd had to do that in order to compare American Community Survey with
the census data. We're getting very good response rates to the American Community Survey.
Altogether, we're getting mail back. When we first mail it out, it comes back in, and we
telephone, and we send an enumerator and all those kinds of things. Our response rate is over 97
percent with the American Community Survey; and, of course, that's all long-form data.
MR. JOST: Herb Sample of the Sacramento Bee.
MR. SAMPLE: Hi. Herbert Sample, Sacramento Bee. Two questions. The first one is
quick. Can you tell me the status of the proposed regulation making the census director
[inaudible]. And also, you mentioned that -- you touched on this earlier, about the unadjusted
and adjusted data and the comparison of the two. Do you know now how you're going to
compare it, and to what criteria you'll use, to decide one way or the other?
DR PREWITT: Yes. First, on the feasibility document, I can't say much. The way the
feasibility document works: this is put in the Federal Register, as you know, for public comment.
And public comment was -- I forget what the deadline was -- was in August, some time.
Anyway, all the public comments came in. Most of the public comment had to do with the
legality of this action by the Secretary of Commerce, and very few had to do with the technical
The Census Bureau has now prepared its own answers to the technical questions and the
lawyers are still working. The Department of Commerce is still working on their answers to the
legal questions. I would think that we will be releasing our response jointly, the technical and
the legal responses, in the next three or four weeks. I think by the end of the month is the
expectation. It's just one of those things that there's a job to be done, and the job will be getting
done; but there's no reason not to try to get the responses out as soon as possible.
What I cannot tell you today, Herb, is whether when you've looked at all of that, you'll
be modifying the reg in any particular way. From the technical point of view, it doesn't appear
so; but I cannot speak for the Commerce Department lawyers.
The other question really had to do with what are the criteria we will use when we're
actually examining to what extent our Accuracy and Coverage Evaluation does to improve the
basic census. A lot of thought is going into that because it's an extremely important issue for us.
That will be shared publicly at the meeting with the National Academy of Sciences in early
October. We simply haven't finished that work yet. We're working hard on it now, so we're less
than a month away from being able to say that.
Once again, it's our attempt to be as transparent as possible; because even in early
October, we're trying to tell you the criteria we will try to be using in February. And it's just our
way to try to build as much confidence as we can in the Census Bureau's processes. So we'll
simply have more detail on that in about a month.
MR. RUDGE: Ronald Rudge with Black Issues in Higher Education. The question I
have is: Where is the Census Bureau in the process of determining how it's going to factor sort of
race and ethnic types of characteristics, and sort of do an analysis. I understand that was an
DR. PREWITT: The Census Bureau's part of that question is fully resolved. If you
follow this very carefully you'll know that the multi-racial question provides for up to 63
different response categories. The five major races plus other, that's six different categories. Five
different categories, different combinations, five different combinations of four, different
combinations of three, two, so forth, produces a total of 63 categories. The Census Bureau's
major products will report all 63 of those categories.
The issue then is not a Census Bureau issue. The issue is for the enforcement agencies
who have an obligation to apply the nondiscriminatory Voting Rights Act and the other
affirmative action policies and programs and so forth. They will have to do some collapsing of
those data in order to carry out their responsibility. But the basic data will be provided in its full
The way in which OMB has discussed this collapsing is now available in a document. I
don't have it right now. It's reasonably transparent once you see it. You can get it directly from
OMB. It's on the Web site, and I forget what it's called; but it was issued in late July, I believe.
The point is, decisions have been made; and we can help you find the reference for them.
Census Bureau data themselves will be, as I've said, displayed in all the categories.
MR. JOST: Any other questions?
DR. PREWITT: Let me just say a word or two, if I could then, in conclusion. I mean
this quite seriously. The press has been extremely important to the census and helping the
American people really understand its importance, and so forth. We have been through what is
the most visible part of the census, which is to say our field work, where we're actually knocking
on doors and people are interacting with the census, all the community activity and participation
and so forth.
The next phase is much less publicly visible: That is -- the tabulation, the processing.
But it really matters, and we really do want the American people to try to understand that. It
really matters because that's what's going to produce the numbers. And the census is not really
about the operations. The census is about the portrait of America that we will provide with the
census results. So we will try to keep you up to date with our operational procedures, even as
we're going through this less visible part, because we believe it's still very important for the
health of this democracy to understand what goes into a decennial census, how it's produced,
why you can have confidence in the numbers. So we'll continue to try to call these meetings,
even though it's much less, as I say, there's much less visibility from the public's point of view.
The next operational press briefing will be something more public because it will be
about the final response rate, which as I say, we'll try to do that in approximately two weeks.
That will again be interesting because that goes back to the larger question of the American
people's response to these kind of civic obligations. We think, as we've already suggested, the
data already look good. We think they're going to look slightly better when we get the last
numbers out. How many states actually met their '90 Plus Five target -- did any states meet it,
how many jurisdictions did? When we've added all the numbers back in, we think the picture is
probably going to be a quite positive one. Thank you.