Census Sampling Alternates Sought

As the question of how to count the nation’s population heads to the Supreme Court for a likely fall hearing, the Clinton administration is scrambling for a last-minute plan to improve the 2000 census without using its preferred method, statistical sampling.

Legal scholars say the conservative high court is likely to uphold Monday’s ruling by a three-judge panel that blocked sampling, a decision that found that such extrapolation violates federal law governing the census.

While the Census Bureau and independent demographers agree that sampling is not only the most cost-effective but perhaps the only way to get an accurate picture of the population, these experts say there are a few expensive, narrowly focused programs that could slightly improve the data in hard-to-count places.

Such programs--involving highly trained interviewers with high-tech equipment in the field and better outreach among non-English speakers--could be crucial to California, which stands to lose about $1 billion in federal funds and two additional congressional seats if the undercount that plagued the 1990 census is repeated.

“It’s going to cost a tremendous amount of money to do any alternative to sampling, and you’re not going to be as accurate as sampling is,” said Tony Coehlo, a former California congressman who now represents Democrats as co-chairman of the Census Monitoring Board, which Congress set up to oversee the upcoming tally. “But if sampling is rejected by the court, are Republicans prepared to put the money where their mouths are? Are they really sincere about counting the undercounted?”

Chip Walker, a spokesman for the House census subcommittee, said Congress is “willing to go the extra yard” in appropriating additional funds but will not provide the Census Bureau “a blank check.” Some have suggested adding $1 billion to the $4-billion price tag on the 2000 count.

Robert Shapiro, the Commerce Department undersecretary responsible for the census, said the administration will not even have a ballpark price tag for a count without sampling until late September.

“If the courts direct the Census Bureau to conduct the census without sampling, [we] are currently studying 20 different supplementary operations . . . to have at least marginal improvement in the result,” Shapiro said.

Monday’s ruling came after a year of political squabbling over the decennial count. Republicans say sampling is guesswork that Democrats want to use to puff up numbers of minorities and poor people who typically support them.

Sampling supporters contend that it is the GOP that is politically motivated in opposing the method. They say that because of a head count’s high error rate, the insistence by Republicans for an actual head count presents a biased view of the nation by undercounting Latinos, African Americans, children and renters much more than others.

“We’re not doing an ‘actual enumeration’ now, because we’re not counting everybody,” Louis M. Seidman of the Georgetown Law Center said, quoting the language of the Constitution that Republicans have insisted prohibits sampling.

Seidman added: “What we’re doing now is a kind of sampling, but it’s an especially invidious kind of sampling--it’s systematically undercounting certain groups.”

In 1990, census officials believe that they missed nearly 9 million people, including 1 million Californians, and double-counted about 4 million. The government believes that it missed 4.4% of the nation’s African Americans, 5% of Latinos, 12.2% of Native Americans living on reservations and 0.7% of whites.

Some suggest targeted programs to address the problem, including more and better trained census-takers for inner cities and far-flung rural areas. But those can be problematic. Stephen Fienberg, a Carnegie Mellon statistician who is co-authoring a book on the census, cited a 1990 attempt to reach parolees that resulted in significant double counting.

The question of accuracy may not be the focus of Supreme Court review. Monday’s ruling by the three federal judges--two Republicans, one Clinton appointee--was based entirely on statutes that said the Census Bureau shall use sampling “except” in cases of determining representation in Congress--one of the key uses of the census. The key question is whether that “except” means sampling for that purpose is prohibited, or can be used at the government’s discretion.

“I just cannot see a conservative Supreme Court overturning this decision,” said Jamin Raskin, a law professor at American University.

But Walter Dellinger, Clinton’s former solicitor general who now teaches law at Duke University, said that if the justices find ambiguity in Congress’ legal language, they probably would defer to the experts--who back sampling.

If the court does get to the constitutional question, legal experts say, sampling has a better shot. Though the Constitution mandates an “actual enumeration,” scholars said it leaves the method up to Congress.

“The idea of ‘actual enumeration’ was to try and have an honest count,” said Jonathan Entin of Case Western Reserve University in Ohio. “It’s implausible to say that the . . . language was to prevent sampling. Sampling was just not there, it wasn’t even thought of.”

Times staff writer Jenifer Warren in Sacramento and researcher Robin Cochran in Washington contributed to this story.