Supreme Court to consider arguments on legality of sampling in 2000 census Scope of technique's use will affect representation

November 26, 1998|By KNIGHT RIDDER/TRIBUNE

WASHINGTON -- After years of racial, ethnic, geographical and politically partisan conflict across the nation, the Supreme Court is about to enter the debate over how the American people will be counted in the year 2000.

The U.S. Census Bureau, in pursuit of the accuracy that has eluded census takers throughout American history, has touched off a nationwide struggle over power and money by planning to use modern statistical sampling techniques on an unprecedented scale.

On Monday the justices will hear arguments in two cases asking whether federal law or the Constitution bars the use of sampling in determining how many people live in each part of the United States.

Politically, the court's ruling -- and the congressional decisions to follow -- could affect the distribution of House seats among the states, the political division in the House and the election districts of members of Congress, state legislators and local lawmakers.

Financially, a ruling for statistical sampling could alter the distribution of $100 billion for environmental, social and community development, health, education, transportation and other federal programs.

Census figures are used for other purposes, too. Every day Americans make decisions that rely on census data: where to build roads, hospitals and schools; where nutrition and job-training programs are needed; how to market products; where to expand businesses.

The Census Bureau's proposed statistical sampling plan would be similar to the technique commonly used in sophisticated surveys and polls. It would use data from part of the population to extrapolate information about the whole population.

If approved by the Supreme Court and financed by Congress (both big ifs), the plan is likely to benefit states and communities with the largest numbers of people who are historically missed in censuses: city dwellers, recent immigrants, children, poor people, renters, and ethnic and racial minorities.

While awaiting the justices' ruling -- and there is no guarantee they will meet the March deadline -- the Census Bureau is preparing to improve its rate of response in 2000 through an aggressive advertising campaign, 800-number call- ins, questionnaires in Spanish and walk-in census places.

Republicans, fearing that the Census Bureau plan will hurt them, say permitting the census to be statistically adjusted would open the door to new errors and the manipulation of census figures for partisan advantage.

Democrats and minority organizations, convinced that statistical sampling will help them, defend the Census Bureau plan as the most scientific way to count hard-to-find people.

From the beginning, there never has been an error-free census.

Thomas Jefferson, who supervised the first census in 1790, recognized that his head-counters had missed a significant part of the nation's population, especially those living in remote areas.

The net undercount in 1940 was 7 million. But for the next four decades, new Census Bureau procedures improved the accuracy of each census.

Then in 1990 the census takers spent a record $2.6 billion -- yet found later, after examining other data, that they had fallen short of an accurate count by 4 million people, or 1.6 percent of the total population. It was the first census in a half-century to be less accurate than its predecessor.

As the Census Bureau gears up for the 2000 count -- which it says will be hampered unless the justices issue their decision by March -- a two-step plan is being contemplated:

First, 200,000 census enumerators would go to randomly selected addresses of people who didn't respond to the mailed questionnaires, knocking on just enough doors to raise the response rate in each census tract to 90 percent.

The results of that random selection process would be used to estimate the size of the remaining 10 percent of the population.

Second, the bureau, to check the accuracy of its estimate, would survey 750,000 households -- 2.5 million people -- specially chosen to be representative of the entire population in all 50 states and the District of Columbia.

By comparing the results of the two sampling phases, statisticians would arrive at final tallies for each state and political district.

Without statistical sampling, the Census Bureau said, the error rate would rise to nearly 2 percent in the next census. Sampling would save money and reduce the undercount from 1.9 percent to 0.5 percent, according to John H. Thompson, the assistant director for Census 2000.

Lawsuits challenging the Census Bureau plan were filed by House Speaker Newt Gingrich on behalf of the House of Representatives and 16 residents of states and counties that anticipate a possible loss of political representation.

Pub Date: 11/26/98

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.