2000 Census to Make Limited Use of Sampling


The Census Bureau announced Wednesday that it will conduct the 2000 census using a traditional door-to-door head count, but that it will be supplemented by a controversial plan designed to estimate through sampling techniques the numbers of those it might miss.

The decision is designed to comply with a Jan. 25 Supreme Court ruling that barred the use of sampling techniques to reapportion congressional seats, while addressing complaints by critics that the head count often misses many segments of the population.

Kenneth Prewitt, the census director, said the head count would be used to recompute congressional apportionment among the states, while the sampling system--which he insisted would prove more accurate--would be used by the states to redraw their congressional districts and to allocate federal grants to states.


Republican lawmakers who oppose sampling immediately assailed the bureau’s decision. It is expected to intensify fighting between the GOP-controlled Congress and the Clinton administration, which favors the use of sampling.

Although California is expected to gain congressional seats from the 2000 census because the state’s population has grown substantially since 1990, the plan announced Wednesday is considered likely to give the state fewer seats than it would get if sampling were employed for that purpose.

California now has 52 representatives in the House. If statistical adjustments were used in the next census, most experts have assumed that the state would gain at least one more seat than it would otherwise.

Besides reapportionment of congressional seats, the census figures are used as the basis for computing some $200 billion in benefits under federal programs from aid to welfare mothers to urban renewal. Officials said that allocation probably will be based on the sampling scheme.

The apparent undercount in 1990 has made a major difference to some states and cities. Los Angeles estimates that it has lost $12 million a year in federal funds because of undercounting, while Los Angeles County loses an estimated $20 million a year.

The battle over how to conduct the 2000 census has been raging for several years. Studies of the 1990 census--a strict head count--found that it missed millions of members of minority groups, who often are hard for census workers to reach.

The shortfall in past censuses also has occurred because some segments of the population--minorities and the homeless, for example--are suspicious of strangers and reluctant to divulge information to government census-takers.

The Clinton administration had been pushing for use of a single census plan that would yield a population figure based on both a head-count and the sampling technique, which employs statistical projection to count parts of the population that are more difficult to track. But the Supreme Court ruled that only the head count may be used for dividing House seats among the states.

Part of the conflict stems from a widely held belief that conducting the census using sampling techniques most likely would benefit Democrats, since it would produce higher population counts in neighborhoods that tend to vote Democratic.

Reacting sharply to the bureau’s announcement Wednesday, House Speaker J. Dennis Hastert (R-Ill.) called the decision “a flip-flop” from the agency’s earlier insistence that producing a single set of numbers as a result of the census was essential.

Rep. Dan Miller (R-Fla.), chairman of the House Commerce subcommittee on the census, said the plan would be “a recipe for disaster” that would “only confuse and confound the American people.” He urged the bureau instead to step up its efforts to reach everyone in its head count.

Meanwhile, Prewitt warned that with the census scheduled to begin in a little more than a year, it is time for politicians on both sides to stop arguing and let the bureau do its job. “We are up to the task, but only if we are allowed to do [it],” he said.

Republicans stopped short Wednesday of threatening to block funds for the sampling effort, but both sides said the battle between Congress and the administration is likely to continue. Under a 1998 agreement, funds for the Census Bureau run out June 15.

The two-pronged census approach also is expected to cost substantially more--between $6 billion and $7 billion, compared to the $4 billion estimated for an earlier version of the plan. By comparison, the 1990 census cost only $2.6 billion.

Under the plan outlined Wednesday, the bureau will follow up its traditional head count with a survey of 300,000 households to project how many people it may have missed. Population totals then will be adjusted to reflect that disparity.