APRIL
29-30 2004 MEETING OF THE DECENNIAL CENSUS ADVISORY COMMITTEE
Gloria
Gutierrez (the Census Bureau’s designated federal official), called the meeting
to order, and announced that with the resignation of chairman Kim Koon, Vice
Chair Sonny Flores would serve as chair for this meeting. Flores reviewed the agenda, and responded to
early interest in discussing the Census Bureau’s hiring of Hispanics, last
meeting’s vote in support of a mandatory American Community Survey (ACS), and
proposals for counting prisoners at their pre-incarceration address.
Census Bureau Updates
Hermann
Habermann (Census Bureau Deputy Director) relayed Director Louis Kincannon’s
regret that he was unable to attend.
Looking to the 2005 budget, Habermann commented that Census has the
support of the administration, and is confident of congressional support. The greater concern is that a budget
impasse, with continuing resolutions limiting agencies to 2004 funding levels,
would disrupt the launch of full ACS implementation.
Habermann
then told of a new OMB initiative requiring that federal agencies provide
“impact analyses” for data collections that could impact privacy. He noted that Gerald Gates (Chief of the
Census Bureau’s Policy Office) and other Census staff are well out in front of
this, having completed more work than the initiative requires. Asked how the OMB initiative might impact
the dissemination of data products, Habermann said he was unsure, but
acknowledged that it has caused them to think further about the issue of
privacy vs. dissemination.
Following up on the question of Hispanic recruitment, Habermann affirmed the Bureau’s commitment, and noted that they have met on the topic. The Census Bureau has made strides in hiring African Americans, but needs to improve with respect to Hispanics because, as Habermann put it, the Census Bureau needs to “look like America to do its job.” And in response to a request for his view of the test of counting Americans overseas, Habermann said only that they are seeing large problems and no easy solutions, and deferred to Jay Waite for the fuller assessment.
Reengineered
2010 Census
Jay
Waite (Census Bureau Associate Director for Decennial Census) described 2004 as
the “year of the tests,” and noted that further tests are planned for 2005, and
2006 in advance of the 2008 dress rehearsal.
The 2004 site tests (in rural Georgia and northwest Queens) are testing
questions that could improve census coverage by identifying households with a
high probability of missing persons or persons counted in error. They are also testing ways to better
identify and count group quarters populations, and the use of hand-held
computers for non-response follow up.
Waite noted that the tests are on time and on budget, even though mail
response has been a bit lower than expected.
Waite
then offered his view of the overseas test—noting that since a canvass approach
cannot be taken, it is difficult to measure the completeness of response. The degree of uncertainty is suggested by
estimates of American citizens living in Mexico, which range from 100,000 to
several million. Responses so far
number about 1,800 for France, 1,100 for Mexico, and 200 for Kuwait. Although still in progress, Waite
characterized the response as low relative to any denominator, and not
defensible for apportionment. Waite
also noted that Spanish language forms are available in the Mexico test, but
that so far only 12 have been returned.
With
respect to the counting of prisoners, Waite explained that the Census Bureau
views its mandate as that of counting people at their usual place of
residence. He noted that they would
count prisoners at their pre-incarceration addresses if directed by legislation
to do so, but otherwise, they will continue counting prisoners where they
usually live. At this point there was comment
from the Committee on the importance of this issue to the urban neighborhoods
where many prisoners resided before incarceration.
Turning
to the ACS, Waite explained that, with the launch of full implementation, the
ACS budget will jump from $64 million in 2004 to $164 million in 2005, and he
reiterated concern with the potential impact of a continuing budget
resolution. He also noted that Census
staff have been meeting with congressional district offices to prepare them for
the questions they will get when ACS data collection begins in their
areas. Waite also noted that the ACS
will begin collecting group quarters information in 2005, and that a National
Academy of Sciences panel is now looking at issues including language, small
area data, and multi-year averages.
Waite
concluded by noting that MAF/TIGER enhancements have focused on spatial
accuracy—improving consistency with GPS.
He described the work (being done under contract with Harris Corp.) as
involving a lot of pavement pounding, with 311 counties completed so far, and
850 due to be completed by September 2004.
In response to a question, Waite confirmed that a Federal Register notice on proposed ACS data products would be issued soon. That notice was released May 14, and an APDU alert was distributed that day. The notice specifically solicits feedback from “public data users,” with written comments due by July 14, 2004. Details and example data products can be found at www.census.gov/acs/www/product_review/.
Race
and Ethnicity Working Group
Because
of the broad interest in the race and ethnicity questions, the Race and
Ethnicity Working Group meeting was held as a general session with Nancy Gordon
(Census Bureau Associate Director for Demographic Programs) leading the
discussion. Gordon described the broad
goal of collecting quality race and ethnicity data, and the specific goal of
maximizing self-response to the five OMB race categories. Put another way, the objective is to
minimize the imputation of race required by either non-response or the
reporting of “Some other race.” Gordon
cited the increase of (mostly Hispanic) persons marking “Some other race,” and
expressed concern with the validity of imputing race based on persons who
reported a specified race.
The
2003 census test showed that dropping “Some other race” increases the
self-reporting of OMB categories. It
also increases total non-response to the race question increases, but there was
a net reduction in “Some other race,” and the need for imputation. Gordon described the Bureau as “very
encouraged” that this approach is “worth pursuing.”
The
next step is the 2005 test, which will test alternative versions of shortened
race and ethnicity questions accompanied by an ancestry question. Although differing in details (such as
wording), all versions involve a series of three questions. An abbreviated Hispanic ethnicity question
is asked first, followed by an abbreviated race question—providing only the
five OMB categories as response options.
A third question (similar to long form ancestry) invites respondents to
describe their ancestry, ethnic origin or tribal affiliation. There is space for up to three write-in
responses.
There
was the usual discussion of details (wording, instructions, etc.), and concern
that this approach might reduce the number of persons reporting some (Asian)
categories. Some representatives
expressed interest in providing written comments, but the lack of immediate and
strong objections was notable for a subject area so prone to immediate and
strong objections.
Within
two weeks, however, stronger concerns were expressed in the promised written
comments. There is concern that the
abbreviated race and ethnicity questions will reduce response, and impair the
Hispanic and Asian data. So it appears
likely that there will be pressure to restore some nationality detail to the
ethnicity and race questions. The
written comments also expressed concern with the elimination of the “Some other
race” category.
Back
to the meeting, your APDU representative inquired about the prospect for short
form ancestry data, and the implications for data products, but little thought
has yet been given to such issues.
However, Jay Waite took the opportunity to comment that the 2010 (short)
form has a serious “real-estate problem,” and that there would only be room for
ancestry on this form if the race question is held to the abbreviated
form.
At
this point, the other three working groups met separately, then reported back
to the whole committee at the end of the day.
Ken
Hodges (Association of Public Data Users) reported that Dave Hubble of the
Census Bureau updated the group on the progress of the ACS. Hubble confirmed that the ACS is on schedule
for full implementation beginning July 2004, but described the achievement as
“nuanced” in that CATI (Computer Assisted Telephone Interview) follow up does
not begin until September, and CAPI (Computer Assisted Personal Interview) follow
up is deferred until October-December, with training of field staff going on
during this period. Hubble’s
explanation that the first full implementation data products would be based on
2005 data is consistent with comments from other Census staff who have
described fourth quarter 2004 as more about ramping up than about data
collection. Hubble indicated that data
collected during this period would not expand the 2004 Supplementary Survey,
but might be used to enhance some of its estimates.
Hubble
noted that the ACS would start collecting group quarters (GQ) information in
2005, and explained some details. All
GQ data would be collected in person from a targeted 2.5 percent of the GQ
population per year (the same rate as household population). Some types of group quarters—such as soup
kitchens and street locations—would be “out of scope,” and concern was
expressed that this would be a problem for those working with these
populations. Sampling would be with reference
to the size of the GQ facility. Data
would be collected for all persons in a sample of facilities with fewer than 16
residents, while only a sample of persons (perhaps as few as 10 per facility)
would be interviewed in larger facilities—with the probability of the
facility’s selection depending on its size.
Questions
remain, and there is concern with how stable ACS GQ data would be from year to
year. When asked how the GQ facilities
list will be updated, Hubble managed our expectations by carefully stating that
“there is a list,” and acknowledging that the update process is still being
worked out. He also noted that
procedures for weighting the GQ responses are not yet determined. The Census Bureau expects administrative
data to contribute to GQ data, but it is unclear how far such data can be
pushed.
Hubble
also explained that ACS sampling for non-response follow up—originally an
across-the-board 1 in 3—will now apply higher rates to tracts where mail-back
and participation rates are expected to be low. Without quantifying thresholds, Hubble reported that tracts with
the “lowest” response would sample 1 in 2 non-respondents (50 percent), those
with “low” rates would sample 2 in 5 (40 percent), and those with average
response would sample 1 in 3 (33 percent).
Tracts expected to have the highest response rates would also have a 1
in 3 follow up, but the initial mail out would be reduced by about 8 percent to
keep the process cost-neutral with the original plan. Variable rate follow up is expected to improve data quality for
the African American and Hispanic populations.
Hubble
concluded with a brief description of the special challenges associated with
the collection of ACS data in remote Alaska and Puerto Rico.
Henry
Der (National Coalition for an Accurate Count of Asians and Pacific Americans)
reported that the group examined 2003 test results suggesting that that a
side-by-side, or “swim lane,” English/Spanish census form would work better
than one that presents questions sequentially—in English then in Spanish. The group recommends that the “swim lane”
approach be adopted.
The
group also expressed concern for the need to test questionnaires in languages
other than Spanish (such as Asian languages), since such forms will probably be
needed for the 2010 census.
Another
question concerned the use of multi-lingual questionnaire assistance
forms—which is being addressed with a three-panel test. The first panel received only an
English-only census form, the second panel received an English-only form with a
multi-lingual assistance guide, and the third panel received an English/Spanish
form with a single-language assistance guide.
The working group recommends a fourth panel, which would receive an
English/Spanish form with a multi-lingual guide. The thought was that this option could be useful, especially if
English/Spanish forms are widely distributed in 2010.
The
group also asked the Census Bureau to consider approaches for persons of low or
no literacy, and the need for oral language assistance, and how such assistance
would be provided. Der reported that
the Census Bureau expressed its understanding of the group’s recommendations,
and indicated that its seemingly narrow focus on English/Spanish options
reflects budget constraints.
Sally
Hillsman (American Sociological Association) described that the group’s
discussion of proposed approaches to improving census coverage. The Census Bureau’s focus is on populations
that are missed the most (kids, young black males, group quarters residents,
etc.) as well as those most frequently counted by mistake.
Numerous
approaches were discussed, and a series of research and development efforts is
summarized in a Census Bureau paper.
Activities related to coverage improvement include the following.
Communication: Partnerships and other outreach efforts are seen as key
contributors to coverage improvement.
Content: The reduction of respondent burden, and a well-designed short
form with clear instructions are considered critical to maximizing high quality
self-report data.
Language: Areas for research and development include the identification of
concentrations of non-English speaking populations, and improving alternate
language questionnaires as well as promotional and support materials.
Residence
Rules and Coverage Probes: Improvements are sought in
the communication of residence rules, and the use of probing questions to
determine if persons really should be counted at a unit, and to identify households
with a high probability of missing persons.
Address
List Development: Obviously critical to census
coverage. An objective is to determine
how the Census Bureau can go beyond USPS information, and make use of address
information from local sources.
Group
Quarters Frame Development: The objective is to combine
the group quarters and housing unit update process—identifying new group
quarters facilities as part of the housing unit update process. A verification process would have to be
developed, and group quarters definitions and classifications are being
reviewed.
Processing: Opportunities for improving coverage also are seen in the data
processing phase. There are concerns
about confidentiality, but one possibility involves the matching of returned
forms to administrative records to verify household members, and identify
persons possibly missed in a household.
As we started day two of the meeting, Jay Waite announced that the 2006 census test sites are the Cheyenne Indian Reservation in South Dakota—which will provide a good test of the hand-held computers in a sparsely populated area—and Travis County, Texas—which includes areas for testing Spanish language forms and procedures.
Public
Comment
Two
persons spoke during the period for public comment. The first was a
representative from the Federation of American Women’s Clubs Overseas (part of
the Census 2010 Coalition), who spoke in support of the enumeration of
non-federal American citizens living in other countries. Speaking of an “alliance of prisoners and
overseas Americans,” she noted that both are seeking to be counted where they
came from, and not where they currently live.
The argument is that because private sector Americans are not counted in
the census, they are not represented in Congress, and that the areas they come
from are under-represented. She
acknowledged the difficulty of an overseas count, and the disappointing test
results (so far), but expressed hope that this would not deter the Census
Bureau from a serious effort to conduct such a count in 2010.
The second, (who described himself as working in redistricting), noted the poor response to the overseas test, and expressed doubt that data from such a count would be good enough for apportionment, let alone redistricting. With respect to the proposed race question, he commended the incorporation of ancestry, but wondered about the impact on the PL 94 data and their use. In response to this speaker, the first speaker clarified that her group advocates the overseas count for apportionment, and not redistricting.
Congressional Update
Jefferson
Taylor (Census Bureau Associate Director for Communications) noted the
departure of Chip Walker from the house subcommittee, and introduced Ursula
Wojciechowski, who is taking his place.
He then observed that this will be an abbreviated year in Congress—with
the summer recess and the fall campaigns limiting the time to reach agreement
on the 2005 budgets, which will include the requested increase for full ACS
implementation.
Jim
Moore (Staff with Representative Tom Davis (R-VA) of the full committee) noted
that Walker is missed, but spoke highly of Wojciechowski, and the contributions
she will make. Turning to the budget,
Moore said “the leadership has decided that the ACS is a priority,” and is
committed to replacing the long form.
The committee is following the test of the overseas enumeration, and
Moore stressed the importance of viewing it as a test of something never done
before—and keeping an open mind about the results.
According
to Moore, the committee does not have a position on the counting of prisoners
at pre-incarceration addresses, but does have questions and concerns. While many prisoners move back to their home
communities upon release, Moore noted that they are replaced by others—often
from those communities. As he put it,
“the prisons do not empty out,” and the committee is sensitive to the stake
that the communities with prisons have in retaining the count of the prison population.
Moore
argued the urgency of mail security in the 2010 census. The subcommittee has been looking into this
issue (and visiting census facilities), and finds current safeguards to be
inadequate. Noting the disastrous
impact that a bio-terrorist attack would have on the census, he asked the DCAC
to consider whether the Census Bureau should have a lead staff position
responsible for mail security.
Moore
concluded with complimentary remarks about the Census Bureau’s efforts to
safeguard confidentiality, and expressed regret that some press reports and
remarks by congressional critics have given the erroneous impression that the
Census Bureau has not been diligent in preserving confidentiality. He called for care in how everyone talks
about and reports on census activities.
David
McMillen (staff with the Committee on Government Reform and Oversight)
described three reports that we ought to know about – two from the National
Academy and one from GAO. The GAO
report concludes that the 2010 census design introduces new risks, and
expresses concern about escalating census costs. GAO also is concerned with the lack of comprehensive census
planning with measurable performance goals.
The Census Bureau disagrees with this assessment.
The
National Academy report on Census 2000 commends the Census Bureau for stemming
the decline in mail response, and for a successful operation, but is critical
of problems in address list development and the enumeration of group quarters. The report also cites the sizable gross
errors in the 2000 census, is critical of the demographic analysis estimates,
and calls for a 2010 post-enumeration survey (something not planned now). The National Academy report on 2010 planning
describes a process “at risk,” and expresses concern that MAF/TIGER resources
are going to TIGER, and not to address list development (MAF).
McMillen
commented that without an address list, everything else is irrelevant. He recalled the importance of the 1994
Census Address List Improvement Act, but noted that improvements are needed for
2010, and that partnership work with local governments needs to start earlier
this decade. In response to a question,
Jay Waite confirmed that the Community Address Updating System (CAUS) is still
in the works. The program would have
census staff updating rural addresses while in the field for the ACS.
The
meeting concluded with discussion of a vote taken at the October 2003 meeting
in support of the ACS as a mandatory survey.
The ACS is currently mandatory, and the vote was in response to test
results confirming the increased costs and impact on data quality associated
with a voluntary ACS.
Representatives
from the Privacy Rights Clearinghouse and the American Civil Liberties Union
objected that a DCAC letter to the Secretary of Commerce described the vote as
unanimous. The PRC and ACLU
representatives were not present at the October meeting, but had what they
called “proxies” in their place. ACLU
contended that its proxy abstained, but that this abstention was not
recorded. Both called for advance
notice of such votes – PRC noting that its proxy would have been instructed to
abstain, and ACLU noting that its proxy would have been instructed to vote
against support for the mandatory ACS.
It
was an extended discussion, but for purposes of this report, the prevailing
DCAC opinion was against revisiting the mandatory ACS issue. There was little support for the suggestion
that there be advance notice of such votes, and that letters to the Secretary
of Commerce be circulated for advance review by all DCAC representatives. The discussion identified the importance of
recording abstentions in Committee votes, but also served as a reminder of the
importance of having informed representatives in regular attendance at DCAC
meetings.
Questions for APDU Members
As in previous meeting reports, we are posing questions to promote dialogue, and to help ensure that your views are represented to the Decennial Census Advisory Committee.
Please send your thoughts on these or other census data issues to your APDU representative, Ken Hodges at khodges@claritas.com and/or alternate representative Mark Salling at mark@urban.csuohio.edu. Your responses will keep us aware of APDDU interests and insights, and provide us with specific feedback to relay to the Committee.
1. Proposals to count prisoners where they lived prior to incarceration are gaining momentum. The discussion centers on which areas should get the prison population count, but are there data use implications if prisoners are counted where they used to live? Please describe any positive or negative implications you see for your work.
2. Do you have preliminary thoughts on the inclusion of an ancestry question on the 2010 census form? Recall that the plan is for a short form only in 2010.
3. ACS will begin collecting data on group quarters next year, but the plans are not fully developed, and there are concerns about the quality of ACS GQ data. What are your thoughts or preferences on the following options?
A. Collect GQ data in the ACS, where the emphasis would be on characteristics rather than updated counts.
B. Do not collect GQ data in ACS (make it a household survey), and focus the group quarters effort on improved estimates of the group quarters count. This option sacrifices estimates of detailed GQ characteristics.
These options have not been proposed, but it would help to know your priorities. In short, the questions are: 1) Is your priority with group quarters counts or group quarters characteristics? 2) Is the ACS an effective vehicle for collecting group quarters data?
4. Do you have other comments or issues relative to this most recent DCAC meeting?