Census Bureau reports big undercount in rehearsal

Census Bureau Director Kenneth Prewitt Tuesday said three dress rehearsals for the 2000 census turned up substantial undercounts in the three tested sites, while he also reported active local participation as the bureau compiles its master address list.

"There is simply no way to correct for the undercount," Prewitt said in a conference call. "We experienced undercount problems at a very high level."

Prewitt said more extensive undercounts were more likely in the dress rehearsal than in the actual decennial census, which officially begins April 1, 2000. "It's a dress rehearsal," he said. "But it's still fundamentally psychologically a dress rehearsal."

House Republicans oppose the bureau's plan to use sampling methods. Republicans favor standard counting methods, but the bureau maintains that sampling is the only way to improve the 1990 undercount.

The bureau Monday released redistricting data from the rehearsals conducted last year. Using various statistical sampling and adjustment methods, the bureau concluded a total undercount of 3.9 percent in Menominee County, Wis., and 6.3 percent in Sacramento, Calif.

In Columbia, S.C., an unofficial review turned up a 9 percent undercount.

House Republicans last week approved a bill that would allow 39,000 local government entities to review census counts, but Prewitt continued to maintain that the post-census local review is unnecessary.

He said about 19,500 communities, representing 85 percent of all U.S. addresses, were reviewing address lists to be used for the census.

Prewitt said the rehearsals also confirmed bureau suspicions that an earlier plan, relying on postal data and a pre-census local review, was inadequate, and that the bureau will soon finish its physical check of every address in the country. "There were serious address list problems," he said.

Prewitt also repeated earlier statements that the bureau plans to release detailed redistricting numbers on April 1, 2001, compiled with both sampling and traditional methods, and that states, unless restricted by their own laws, could use whichever number they wanted.