As the 1980s became the 1990s, the U.S. Census Bureau ran a TV advertising campaign that musically urged Americans: "Answer the Census, we're counting on you." But some in government, and the legal and public interest communities argue that for the upcoming 2000 census, it may be more appropriate to say, "we're sampling you."
"Our goal is to provide the best [count] of persons living in the country," says Rajendra Singh, mathematical statistician for the Census Bureau in Suitland, Md., near Washington, "But our past experience, from 1960, 1970, 1980 and 1990, has shown that we are missing a certain portion of the population, and that portion varies by demographic group."
The 1990 Census missed 1.6 percent of the U.S. population, bureau studies have concluded-a total of 4 million Americans. Moreover, the number of people overlooked varies by ethnic group: Just 0.7 percent of white America wasn't counted in 1990, while 4.4 percent of African-Americans, 5 percent of Hispanics and 12.2 percent of Native Americans living on reservations were missed by the constitutionally mandated enumeration.
To avoid another under-count, the Clinton administration and the Census Bureau propose expanding the use of statistical sampling techniques, similar to those used in opinion polling, to estimate the number of people missed by or unresponsive to census survey methods.
The stakes are large. All sorts of businesses use census numbers to explore new markets, track income and demographic trends, and plan housing, retail and commercial developments. The federal government distributed $125 billion to state, local and tribal governments in 1990 alone, about half of that based on formulas using census data. State legislatures use census figures to draw congressional districts, and congressional seats are apportioned between the states based on the decennial data.
Which explains in part why congressional Republicans took the Clinton Administration to court over sampling. The GOP suspects that sampling would increase the count of poor people and minorities in the inner cities, shifting congressional seats out of the South, and tilting the balance of congressional power to Democrats. In August, a special three-judge panel of the U.S. District Court in Washington threw out the Census Bureau's sampling plan, ruling it did not meet the constitutional standard of "enumeration" (meaning a one-by-one headcount) intended by the founding fathers. That case is now on appeal to the Supreme Court.
But according to K. Tia Burke, an associate at the law firm Christie, Parabue, Mortensen and Young, P.C., in Philadelphia, "No census in the past century has been an actual enumeration." She notes that the Census Bureau has long used sampling to clarify and expand data gathered by census-takers during phone, mail and one-on-one interviews. Still, Singh notes, sampling has never been used for congressional apportionment, and even Burke admits, "Sampling is not an actual enumeration; it can't be."
The Census Bureau remains convinced that it can directly reach 90 percent of the population, then use sampling techniques to scientifically estimate the rest. But David Almasi, a policy analyst at the National Center for Public Policy Research in Washington, says you can't be sure you've reached 90 percent of the population if you don't know how many 100 percent will be. "Of course they will reach 100 percent," Almasi cracks, "because that's their goal."
Almasi proposes instead of sampling that Congress appropriate whatever funds are necessary to finish the job (at least $600 million more than the current $4 billion now estimated), and that the federal government inspect welfare, Medicare and Medicaid rolls, birth certificates and tax records to ensure the most complete enumeration. "I think you can come up with a more accurate count that ever before," he says. Burke counters, "It doesn't seem that any amount of money can do the job. The 1990 Census was the most expensive ever, and at the same time the least accurate."
The Supreme Court is expected to rule on the Clinton administration's appeal of the district court ruling sometime next March. Another case filed by local jurisdictions, voters and a private legal foundation is also on appeal. "Essentially right now, [sampling] is considered illegal" for census purposes, Almasi notes. If the Supreme Court overturns the August ruling and reinstates sampling, Almasi predicts, "You're probably going to have 435 different lawsuits"-one for every congressional district, from voters claiming that sampling missed them or invented others. "And then," he adds, "we're talking about some real money."