Census Bureau's success raises questions about need for sampling

House Government Reform Census Subcommittee Chairman Dan Miller, R-Fla., Wednesday pointed to newly released preliminary census data--showing a better count than a decade ago--to argue against using statistical sampling for redistricting. But Democrats energetically disagreed, equating inclusion in the census with the right to vote.

"Why does a good census make [sampling] adjustment hard to justify?" Miller asked during a hearing, providing the answer minutes later: "I think if you've done a very good job ... it becomes harder."

The Census Bureau released estimates Wednesday showing the 2000 census missed between 0.96 and 1.4 percent of the country's population--better than the 1990 census undercount of 1.6 percent.

Rep. Carolyn Maloney, D-N.Y., who had previously served as the panel's ranking member, joined Miller in praising the bureau's work, but said sampling is the only way to count people missed by the 2000 census--people most likely to be minorities.

"We are on the verge across this nation of redrawing every political jurisdiction in every state," Maloney said. "Only those census numbers which give us the most complete accounting of everyone residing in our country should be used for that purpose."

Contrary to reports that House Republicans--who oppose sampling for fear of political manipulation--have a guarantee from President Bush that he will oppose sampling, acting Census Bureau Director William Barron said he has felt no pressure from Commerce Secretary Don Evans.

"Absolutely none," Barron said. "The Secretary is encouraging me to do what we need to do."

Barron's role is important because an administrative rule places authority for determining sampling's accuracy into his hands in consultation with a panel of career census experts.

The special committee is expected to make a decision within three weeks, and a recommendation in favor of sampling would likely force states to chose between sampled numbers or more traditional data in their state redistricting.

Barron--a longtime government employee, not a political appointee--called the bureau's sampling program an "excellent survey" with a sample size of 314,000 housing units.

However, Barron said his committee's decision will depend on how well sampling works in practice. "We have confidence in the concept," Barron said.

Miller argued that while sampling methods may be more accurate for large geographic areas, local accuracy is necessary to satisfy the constitutional "one man, one vote" requirement. "One block at a time is what we work with," Miller said. "And the concern that we have is [that] the accuracy is not there."