Politics

Counts And Undercounts

By CBSNews.com staff CBSNews.com staff

February 15, 2001 / 9:35 AM EST / CBS

The U.S. Census Bureau says the 2000 census tallied the country's population more accurately than the 1990 count, most notably in the traditionally undercounted black, Hispanic and American Indian minorities.

"I could not be more pleased with the strength of these numbers," said Commerce Secretary Don Evans, reacting to preliminary estimates showing the number of people not counted was reduced by at least 50 percent in most groups.

The news cheered civil rights groups but was unlikely to stave off a new round of political wrangling between Democrats and Republicans over the use of statistical sampling to "correct" undercounts.

The heightened political tensions arose at a House hearing Wednesday that reviewed Census Bureau operations.

The issue is a perennial political hot potato as the census numbers are used both to determine how many seats each state has in the House and how to redraw the lines of the districts that elect delegates to those seats.

The Supreme Court ruled in January 1999 that population counts statistically adjusted to compensate for overcounts and undercounts cannot be used to allocate seats for the House and that only the raw data should be used for this purpose.

However, the court said it was up to individual states to decide which numbers they would use when redrawing congressional and state legislative lines next year.

The Census said preliminary figures showed the undercount rate for the total population ranged from 0.96 percent to 1.4 percent, compared to 1.6 percent in 1990. The bureau estimated it had missed some 4 million people in the 1990 census, mostly people of color and the poor, and double-counted 4 million others.

While exact figures were not available, the 2000 undercount rate for Hispanics ranged from 2.22 percent to 3.48 percent compared to 4.99 percent in 1990.

For non-Hispanic blacks, the rate was 1.60 percent to 2.73 percent compared to 4.57 percent in 1990.

The undercount rate for American Indians living on reservations was between 2.77 percent and 6.71 percent compared to 12.22 percent in 1990.

The bureau said it would recommend by the end of February whether adjustment was technically feasible and whether it would improve the accuracy of the counts.

But the final decision on whether to issue adjusted figures rests with Republican President Bush, who will be under pressure from both sides in the debate.

"The resounding success of the 2000 census casts serious doubt on the need for an adjustment plan that may be illegal, unconstitutional, inaccurate and have serious social costs associated with it," said Chip Walker, a spokesman for Republican Rep. Dan Miller of Florida, who heads a census subcommittee.

But civil rights leaders urged Mr. Bush not to make a hasty decision.

"Any decision on the release of corrected numbers made before the Census Bureau professionals issue their report will bear the taint of political interests over truth n numbers," said Wade Henderson, the head of the Leadership Conference on Civil Rights.

Rep. Martin Frost of Texas, the third-ranking Democrat in the House, warns that the decision on whether to use sampling is "an important test for Bush and for his purported effort to make the Republican Party more moderate and more inclusive."

Sampling generally is backed only by Democrats and civil rights groups. Most Republicans contend instead that follow-up surveys should be used only to test accuracy and not to adjust raw results that could be used for redistricting.

"While it appears that the Bureau improved on the 1990 count, we still don't have all the pieces of the puzzle," says Gilbert Casellas, co-chairman of the bipartisan U.S. Census Monitoring Board. "We still don't know the margin of error, who was counted twice and where the people that were missed reside."

"We must be sure that we are not double-counting whites at the expense of counting minorities," adds Casellas. "What is an acceptable number of Americans not included in the census? Our answer is none."