CENSUS DECISION BEST LEFT AT CABINET LEVEL

The new commerce secretary in the Bush administration has moved quickly to put off -- if not undo -- a leftover plan to count Americans on the basis of a computer formula.

This counting method, called statistical sampling, was favored by Democrats during Bill Clinton's days because it allowed for census guesstimates of mostly poor and minority Americans who might not have been counted in the official 2000 census.

Following the 1990 census, there were estimates that several million Americans weren't counted. As a result, some areas of the country, mainly in large cities, felt they were shortchanged by anti-poverty programs whose federal funding is based on the number of needy residents.

The intent of statistical sampling was well-intentioned: By using computer models and sophisticated statistical analysis, census officials would be able to estimate how many people weren't counted around the country. Even Congress backed the plan instead of insisting that the U.S. Census Bureau do a better job of head counting.

What Congress approved was funding for the regular head count for legislative redistricting as required by the U.S. Constitution. But it also appropriated funding for the statistical sampling, whose numbers would be used to determine eligibility for key funding programs.

What advantages there are in statistical sampling are more than offset by the pitfalls.

For one thing, it is too easy to get the wrong numbers when relying on statistics alone. Look at the results of the presidential election last November. The winner flip-flopped several times during that long night, thanks entirely to faulty statistical information.

Perhaps the most compelling reason to take another look at the need for statistical sampling is the fact that preliminary estimates show the 2000 census missed a much smaller percentage of Americans than the 1990 census, although Democrats contend some 3 million individuals may have been uncounted.

Commerce Secretary Don Evans has decided against leaving the decision on statistical sampling up to the Census Bureau and a group of career statisticians. Instead, Evans himself will decide whether statistical sampling is even needed.

The potential for manipulating figures by any method, of course, is great. And the results would have a huge effect on how lines are drawn for legislative districts and on how federal funding is allocated across the nation. This is an important decision, and it's one best left to a Cabinet member.