WASHINGTON (AP) - Critics angered by a Bush administration decision to designate the raw 2000 census head count for congressional redistricting are calling for firm numbers of the undercount - not just a percentage - to be released to better determine the tally's accuracy.

In ruling against an adjustment of the raw count Tuesday, Commerce Secretary Don Evans said that initial tally was the "most accurate in history" and would be sent to states for redistricting.

The Census Bureau was to send redistricting data to governors and state legislative leaders to 11 states this week. The public was to see the data after those state officials tell the Census Bureau they received it.

Those states are: Arkansas, Indiana, Louisiana, Mississippi, New Jersey, Pennsylvania, South Dakota, Texas, Vermont, Virginia and Wisconsin.

In explaining his decision, Evans cited Census Bureau estimates of a net undercount of about 1.2 percent, or 3.3 million of the nation's 281 million people. That was down from 1.6 percent, or about 4 million of the 1990 population.

In both years, though, the undercount was primarily made up of minorities, and could have been made up through an adjustment, critics said.

The net undercount figure in 1990 is based on Census Bureau sampling data suggesting that 8 million people were missed and 4 million people were double-counted.

The Census Bureau said comparable numbers for 2000 were "slightly lower" than 1990, and would be officially released soon. Democrats said that until those numbers are released, it is impossible to say how accurate this year's count was.

"If the benchmark Secretary Evans is using is that the net undercount was reduced, we won't know until we get the total number of people missed in the census, and the total number of people counted twice," Rep. Carolyn Maloney, D-N.Y., a critic of the decision who called for adjusted data to be released regardless of redistricting, said Wednesday.

Chip Walker, spokesman for Rep. Dan Miller, R-Fla. chairman of the House Government Reform Committee's census panel, called the Democratic pleas "sour grapes."

Evans agreed with a Census Bureau recommendation that raw numbers were more accurate than the adjusted numbers.

Though the political debate has quieted for now, all sides agree that the issue will again arise in court. Some supporters of adjustment want all data to be released to help bolster their legal arguments.

The Census Bureau "made clear that right now the most accurate set of numbers is the actual head count, which some consider the most accurate in our history," Commerce spokesman Jim Dyke said.

Census officials have said they are still studying the adjusted numbers and are unsure to what extent that data would be released.

A consultant to members of the U.S. Census Monitoring Board appointed by former President Clinton also said Wednesday there were roughly 44 million "errors" in the raw count last year, up from about 35 million in 1990, though the population also increased in that time frame.

Among other things, errors include people who could have been omitted, counted in the wrong place, or double-counted. The error number, however, is not a comparable measure of the actual overcount and undercount numbers being sought, said Gene Erickson, who had access to the census numbers in his capacity as the Democratic consultant.

"Still the fact that it is bigger tells us that overall level error was just as great in 2000 as in 1990," Erickson said.

Because of the size of the population, both sides agree it is nearly impossible to have an error-free census. But Walker said even more error could have also been unfairly injected into the count using adjustment.