From: Census2000 <Census2000(a)ccmc.org>
Sampling Controversy Resurfaces in Congress As Census Director Stands By
Plan to Adjust Counts
Plus: New Jersey Assembly Approves Anti-Sampling Bill;
Legislation Introduced for Count of Overseas Americans; and more.
The controversy over the use of statistical sampling to adjust the
initial census counts reemerged at a congressional hearing last month,
as the Census Bureau began its quality check survey to measure
undercounts and overcounts in Census 2000.
At a hearing of the House Subcommittee on the Census on May 19, Census
Bureau Director Kenneth Prewitt described the agency's plans for the
Accuracy and Coverage Evaluation survey (A.C.E.) of 314,000 housing
units nationwide. Telephone interviews with households in the A.C.E.
sample that had returned a completed census form by mail began in late
April. By May 19, the Bureau had completed 60,000 interviews by phone,
more than its projected 10 percent completion rate for that point in the
process. Census takers will start in-person interviews with A.C.E.
households on June 19 in areas where census takers have finished all
non-response follow-up visits.
Subcommittee Chairman Dan Miller (R-FL) reiterated his long-standing
opposition to the Census Bureau's plan to produce adjusted population
numbers based on the A.C.E. results for purposes other than
congressional apportionment. "[T]here is no guarantee that this plan is
even viable," the chairman said. The Supreme Court ruled 5 to 4 in
January 1999 that federal law prohibits sampling to compile the state
population totals used to allocate the 435 seats in the U.S. House of
Representatives among the 50 states. However, the Census Bureau plans
to transmit adjusted census figures to the states early next year for
use in the redistricting process. A provision of the Bureau's fiscal
year 1998 funding bill also requires the agency to make available census
counts down to the block level that are not adjusted based on the
quality-check survey.
Rep. Carolyn Maloney (D-NY), the panel's senior Democratic member, said
the 2000 census "may well be the best, fairest and most accurate census
ever," not only because of operational successes so far but because "it
incorporates modern scientific methods into its design." The 1990
census, she noted, had a 10 percent error rate, including who are
missed, counted twice, or counted in the wrong location. "The closer
the Census Bureau has gotten to developing a way to fix those errors,
the harder the opponents of a modern census have worked to stop them,"
Rep. Maloney said, referring to the disproportionate undercount of
racial minorities and the poor in urban and rural communities.
Rep. Miller said Republicans question the constitutionality, legality,
and feasibility of a census plan that incorporates sampling and
statistical estimation techniques. He called the decision to issue two
sets of census counts "a political move, clearly against the best
interest of the Nation" by the Clinton/Gore Administration. Pointing to
last year's Supreme Court decision, he cautioned that supporters of the
plan "have yet to win a court case." "Census estimation," the
chairman
said, "is not a system that lends itself to trust and integrity."
Rep. Miller described statistical estimation as existing "only in a
virtual world," where people are added to or subtracted from the counts
without verification. This means, he said, that some people who have
filled out and returned a census form "will be counted as less than a
whole person." Rep. Miller also criticized an element of the Bureau's
statistical methodology that he said assumes people of the same race or
ethnicity "act alike and have the same tendencies," by assigning them to
one of nearly 450 demographic groupings (called 'post-strata') for the
purpose of measuring the under- and overcount. The chairman asked the
director for "assurances... that [the adjusted] numbers will be fully
scrutinized by the Bureau and the scientific community, at large, prior
to their release for public use."
Countering the chairman's criticism of the sampling methodology, Dr.
Prewitt said the A.C.E. survey was no more difficult than other large
Census 2000 operations. He called the census "an estimation of the
truth" and said its accuracy should be measured by how close the results
are to the nation's true population. The director noted that census
data are "unstable" for very small areas (such as census blocks)
regardless of whether the figures are adjusted using statistical methods
or not. Adjusted census numbers are "unquestionably" more accurate at
the block level, Dr. Prewitt said in response to a question from Rep.
Miller, than numbers produced from direct counting methods alone. He
also defended the use of demographic 'post-strata' to estimate under-
and overcounts, saying the groupings were based on scientific knowledge
about people with similar probabilities of being missed or
double-counted in the census. A panel of experts convened by the
National Academy of Sciences (the Panel to Review the 2000 Census,
chaired by former Bureau of Labor Statistics Commissioner Janet L.
Norwood) is independently reviewing the A.C.E. statistical methods and
operations.
Dr. Prewitt emphasized that while the Bureau has decided to conduct the
A.C.E. according to a technically sound design, it cannot determine in
advance whether the direct count or the statistically adjusted numbers
are more accurate. "If the Census Bureau does not have confidence in
the A.C.E. results, we will not use it," the director told subcommittee
members. The Census Bureau director, not the Secretary of Commerce, he
said, should make the adjustment decision.
The director also vigorously challenged the assertion that statistical
estimation could invite tampering with the numbers to achieve partisan
advantage. He said the Census Bureau is "not competent" to design and
implement an estimation plan that produces a partisan outcome. "Where
is the evidence?" he asked, that the Bureau has a partisan agenda. Dr.
Prewitt said it is "inconceivable" that the Census Bureau would know how
statistical adjustments would affect the drawing of political district
lines by 50 state legislatures.
Other legislative news: Rep. Carolyn Maloney introduced legislation to
provide funds for the Census Bureau to begin planning a census of
American citizens living abroad in 2003. H.R. 4568 would authorize $5
million for planning a Census of Americans Abroad. Rep. Maloney
introduced a bill (H.R. 3649) last year to require an interim count of
Americans overseas in 2003 and planning to include this population in
the 2010 census. Senator Jay Rockefeller (D-WV) sponsored a similar
bill (S. 1715) in 1999, which was referred to the Senate Committee on
Governmental Affairs.
Census 2000 will include active members of the U.S. armed forces,
civilian government employees, and their dependents, who are stationed
or living overseas on Census Day (April 1, 2000), in the state
population totals transmitted to the President for the purpose of
congressional apportionment. These personnel and their family members
are counted at their 'home of record,' the state in which they lived at
the start of the their military or federal service.
In a letter to Reps. Harold Rogers (R-KY) and Jose Serrano (D-NY),
chairman and ranking minority member, respectively, of the House
Appropriations Subcommittee on Commerce, Justice, State and The
Judiciary, Rep. Maloney said the estimated three to six million American
citizens living abroad "make enormous contributions to our economy" and
"vote, pay taxes, and enjoy the same constitutional protections" as
other Americans. She urged the subcommittee to include funds for
planning an interim count in the Census Bureau's fiscal year 2001
appropriations bill.
Rep. Lee Terry (R-NE) introduced a bill last month (H.R. 4458) that
would limit the questions asked in future censuses to those included on
the Census 2000 short form. The measure was referred to the House
Committee on Government Reform.
Update on Census 2000 operations: The Census Bureau announced last week
that it has counted, either by mail or through a personal visit, about
92 percent of the nation's 118 million housing units. Director Ken
Prewitt said gathering information from the remaining eight percent of
American households posed the greatest challenge for the approximately
450,000 census enumerators working to complete 'nonresponse follow-up'
operations by July 7.
State legislative activities update: The New Jersey State Assembly
yesterday approved a bill to require the use of unadjusted census
numbers for redistricting purposes. The measure passed the
Republican-controlled chamber on a party line vote of 42-34. On May 22,
the State Government Committee approved A. 1682, sponsored by
Assemblyman Michael Carroll (R), by a 3 - 2 vote along party lines.
Assembly Speaker Jack Collins (R) initially scheduled a vote on the
measure for May 25, but agreed to postpone action temporarily after
Minority Leader Joseph Dorio Jr. (D) said legislators and the public
needed more time to consider the bill's consequences. Last week,
however, the Speaker announced that the Assembly would take up the bill
on June 5. The State Senate has not yet considered the measure.
Questions about the information contained in this News Alert may be
directed to TerriAnn Lowenthal at 202/484-2270 or, by e-mail at
<terriann2k(a)aol.com>om>. For copies of previous News Alerts and other
information, use our web site
www.census2000.org. Please direct all
requests to receive News Alerts, and all changes in
address/phone/fax/e-mail, to the Census 2000 Initiative at
<Census2000(a)ccmc.org> or 202/326-8700. Please feel free to circulate
this information to colleagues and other interested individuals.