Use sampling for census

The Register-Guard
December 17, 2002

Last week's reluctant release of computer-revised data by the U.S. Census Bureau on the 2000 census underscores the need for the use of scientifically approved statistical sampling in the decennial enumeration of Americans.

The computerized update showed that the bureau's regular head count missed about 1.2 percent of the nation's population, or about 3.3 million people. In Oregon, the undercount - due to census forms being junked by recipients, the bureau's inability to find certain addresses, etc. - was pegged at about 43,000 residents, or about 1.3 percent of the state's population. Of Oregon's 36 counties, Lane County led in the undercount with 5,730 people. Multnomah County and Marion County came in second and third, respectively.

California had the largest undercount, with 509,012 people, but Alaska led the nation in percentage of population missed by the census, with 2.4 percent, or 15,136 residents.

Unfortunately, it took a Freedom of Information request by Oregon state Sens. Susan Castillo and Margaret Carter to get the Census Bureau to even release the updated, computerized count. Even then, the bureau called the updated version as inaccurate as the original count.

That says a lot: How can Americans trust in the accuracy of the census if the Census Bureau itself doesn't believe in it?

The reason behind the confusion is political. Despite the accepted use of scientifically approved statistical sampling by governments, universities and the like, Congress refused to allow the Census Bureau to use the technique. Why? Because most of those missed in the regular head-count are members of minority groups. The Republican-controlled Congress feared that adding to the numbers through statistical sampling would boost Democrats' chances of picking up congressional seats. Callous? You bet. Wise? Not in the least.

An accurate census is needed to ensure that congressional and other political districts are properly apportioned. Accuracy further ensures that $200 billion in federal funds for Medicaid, foster care, block grants and a wide array of social services is fairly shared among the states.

While the next census is just over seven years away, Congress should allow the Census Bureau to employ the statistical sampling used on such a wide scale throughout the country for other purposes. Indeed, Congress should insist upon it. The whole point of the census is to obtain an accurate count of all Americans. Sampling will come far closer to achieving that goal than will the old-fashioned head count, the inadequacies of which have been demonstrated once again.