|
Accuracy in the 2000 Census
What the Experts Say About the Undercount
Population undercounts, and the "differential undercount" of minorities, have been with us for
decades, and were worse in the 1990 census than in 1980. In 1990, one-third of all households failed to
return their census forms, so hundreds of thousands of census enumerators were dispatched to find and
interview the nonrespondents.
The job was far from easy. Because of changing living and work patterns, such as shrinking
household size (more places to visit) and the rise of two-earner households (fewer people at home during
the day), filling out a questionnaire sometimes required half a dozen visits. In wealthy gated communities,
and in poor neighborhoods where apartment buildings have locked entries, it was challenging even to get
close enough to try. Paying people to go door-to-door was also an expensive undertaking; the 1990 census
cost a total $2.6 billion.
With follow-up research, the Bureau determined that despite these exertions, it had missed about
8.4 million people -- largely poor people and minorities -- and double-counted 4.4 million mostly affluent
whites. If you did not own a home you were more likely to be missed and if you owned more than one
home you were more likely to be counted twice. That net undercount of 4 million represented about 1.6
percent of the total population. For some minority groups, the undercount was far worse: for blacks,
about 4.4 percent; for Latinos, closer to 5 percent. That means about one person in 20 in these groups was
omitted from the federal government's authoritative population count.
A Census for the 21st Century
In February 1996, the Census Bureau unveiled its plan for a re-engineered census in 2000. Based
on a mandate from Congress following the 1990 census, the overarching goals of the plan are to improve
accuracy and hold down cost, while working to eliminate the persistent, disproportionate undercounts of
minorities and the poor. The Bureau placed over $100 million in paid advertising to build awareness of
how communities benefit from programs that hinge on census data. It sent out a notice before the census
forms were mailed, and follow-up mailings to remind people to send their forms in. Then the Bureau began
its follow-up efforts with those who had still not responded.
After all of the non-response follow up efforts were completed, a key element of the 2000 census
plan began -- the use of modern statistical methods to complete a quality check on the raw field counts.
This second "Super-Census" -- known as the Accuracy and Coverage Evaluation(ACE) -- targets selected
census blocks around the nation in order to double the previous work.
For the ACE the Bureau is sending its best enumerators out to interview 314,000 households
throughout the country to check for undercounts and over counts. This type of statistical method is well
known to quality assurance managers in the private sector. The ACE will provide the basis for a correction
of undercounts and over counts before the final numbers are tallied and published.
This statistical process -- dual system estimation -- has generated controversy in some quarters.
But in fact, statistical methods have been part of the census since at least since 1940. For example, in
1990, the Bureau used imputation formulas to assign occupants to the identified but unreachable
households, based on established patterns for that vicinity.
On September 28, 2000, the Commerce Department issued a final rule delegating to the Census
Bureau the final decision as to whether to issue corrected Census data following the 2000 Census. The
draft rule was proposed on June 14, 2000. At that time, in addition to the release of the proposed
regulation the Commerce Department and the Census Bureau released two other documents, a legal
analysis supporting such a release, and a document prepared by the Census Bureau detailing the rational
for a preliminary determination that (1) it is feasible both operationally and technically to produce
statistically corrected data with in the time frame required by law and (2) the statistically corrected data
will be more accurate.
All three documents were published in the Federal Register, and placed on the Census Bureau Web
site for review, and can be found at http://www.census.gov/Press-Release/www/presskit.html.
The rule was subject to a 45 day comment period. The final rule can be found at: http://frwebgate.access.gpo.gov/cgi-bin/getdoc.cgi?dbname=2000_register&docid=fr06oc00-10
What the Experts Say
Despite the disagreement in Congress over the proposed use of statistical methods in the census,
the methods have widespread support in the professional and scientific communities. Here is what some
expert and independent observers have said about the Census Bureau's plans and its decsion making
process:
"As former directors of the Bureaus of the Census we wish to go on record in support of the
proposed rule regarding the procedure for release of the corrected data file for Census 2000 published in
the Federal Register on June 20, 2000.
Delegating the decision on the release of the corrected data for Census 2000 by the Secretary of
Commerce to the Director of the Census Bureau, upon the recommendation of the Director's professional
staff, puts the decision with scientific professionals where we believe it belongs. Our experience ins that
the Census Bureau professional staff is comprised of highly-trained, objective experts in the fields of
census and survey methodology, statistics and demography. They are more capable than others of judging
and recommending whether the corrected 2000 data more accurately reflect the U.S. population than does
the enumerated count released earlier for apportionment purposes.
We emphasize that our position in this letter focuses only on the procedure as contained in the June
20, 2000, Federal Register. We are not taking a position on whether to release corrected census data, but
rather that this decision be delegated to the Census Bureau Director working with the Census Bureau
professional staff" Letter from former directors of the Bureau of the Census: Martha Farnsworth Riche,
Ph.D.; Barbara Everitt Bryant, Ph.D.; John G. Keane, Ph.D.; and Vincent Barabba to John Thompson,
Associate Director for Decennial Census, July 27, 2000.
Commenting on the Census Bureau plan developed in response to the Supreme Court's January 25,
1999, decision, the National Academy of Sciences wrote that it "represents good, current practice in both
sample design and post-stratification design, as well as in the interrelationships between them." Report of
the Panel to Review the 2000 Census, Committee on National Statistics, National Academy of Sciences,
May 3, 1999.
In a subsequent letter, the National Academy of Sciences "commends the Census Bureau for the
openness and thoroughness with wich it has informed the professional community about the kinds of
evaluations that it plans to conduct of the census and the A.C.E. data prior to Marcy 2001. The papers
presented at the panel workshop provide evidence of the hard work and professional competence of Census
Bureau staff in specifying a series of evaluations that can inform the adjustment decision." Report of the
Panel to Review the 2000 Census, Committee on National Statistics, National Academy of Sciences,
November 9, 2000.
"There is compelling evidence from the Bureau's evaluations that adjustment corrects, at least
partially, the pernicious differential undercounts for various minority groups." Sampling-based
Adjustment of the 2000 Census - A Balanced Perspective; By Anderson, Daponte, Feinberg, Kadane,
Spencer, and Steffey
"[P]hysical enumeration or pure 'counting' has been pushed well beyond the point at which it adds
to the overall accuracy of the census. ...Techniques of statistical estimation can be used, in combination
with the mail questionnaire and reduced scale of follow-up of nonrespondents, to produce a better census at
reduced costs." Report of the Panel on Census Requirements in the Year 2000 and Beyond, Committee
on National Statistics, National Academy of Sciences, 1995.
"Sampling households that fail to respond to questionnaires produces substantial cost savings and
should improve final data quality." U.S. General Accounting Office, October 1995.
"The Census Bureau has adopted a number of innovations to address the problems of past
censuses -- declining accuracy and rising costs. One innovation, which we fully support, is the use of
statistical sampling for non-response follow-up." Honorable Frank DeGeorge, Inspector General, U.S.
Department of Commerce, October 1995.
"Because sampling potentially can increase the accuracy of the count while reducing costs, the
Census Bureau has responded to the Congressional mandate by investigating the increased use of sampling.
...We endorse the use of sampling for these purposes; it is consistent with best statistical practice." Report
of the Blue Ribbon Panel on the Census, American Statistical Association, September 1996.
"The planned and tested statistical innovations [in the census] ...have the overwhelming support of
members of the scientific community who have carefully reviewed and considered them. If their use is
severely limited or prohibited, the 2000 Census planning process will be obstructed, and the result could be
a failed census." Douglas S. Massey, President, Population Association of America, June 1996.
"Change is not the enemy of an accurate and useful census; rather, not changing methods as the
United States changes would inevitably result in a seriously degraded census." "Preparing for the 2000
Census: Interim Report II" of the Panel to Evaluate Alternative Census Methodologies, National
Research Council, June 1997.
|
