FAIR USE NOTICE. This document contains copyrighted material whose use has not been specifically authorized by the copyright owner. The CHANCE project is making this material available as part of our mission to promote critical thinking about statistical issues. We believe that this constitutes a `fair use' of the copyrighted material as provided for in section 107 of the US Copyright Law. If you wish to use this copyrighted material for purposes of your own that go beyond `fair use', you must obtain permission from the copyright owner.

Copyright 1996 The New York Times Company
The New York Times

March 5, 1996, Tuesday, Late Edition - Final

SECTION: Section A; Page 22; Column 1; Editorial Desk

LENGTH: 432 words

HEADLINE: No Need to Count Every Last Person

When the Census Bureau counts heads in the year 2000, it will stop trying to track down every household. Instead, after census-takers contact about 90 percent of the population, the bureau will use sampling techniques to round out the total.

The idea of using sampling to supplement counting may trouble many Americans. But the bureau's proposal is fiscally necessary and statistically sound. Sampling will cut the costs of conducting the census and should improve its accuracy -- especially for hard-to-locate minority families.

The cost of the census has spiraled upward because Congress has demanded more detailed data and because fewer families -- only about 60 percent in 1990 -- return their census forms by mail. The bureau sends census-takers knocking on the doors of non-respondents -- an expensive process that can require six or more visits.

The Census Bureau hopes to save money by knocking on only 1 in every 10 doors after the 90 percent threshold is reached and using these sample results to infer what the nine other households look like. The procedure will not only be cheaper, but probably more accurate. The bureau also proposes a second round of sampling to solve the problem of undercounted minorities, who are often hard to find. That should help cities with large minority populations, like New York, that now get shortchanged in representation in legislatures and distribution of Government grants allocated on the basis of population.

Critics charge that sampling could worsen estimates for small areas, like 50-household census blocks. But even if true, those mistakes would not be important. Sampling would improve estimates for larger areas, like Congressional districts, which are the ones that matter for reapportionment or other important government decisions. Two panels of the National Research Council have recently concluded that the bureau's proposals are sound.

The reality is that door-to-door census-taking has become too costly to be worthwhile. The last census cost $2.6 billion. Using traditional methods, the next is projected to cost $4.8 billion, but with sampling that cost can be cut to $3.9 billion.

Statistical sampling, employed judiciously, allows a small, highly skilled staff to collect relatively error-free data rather than sending thousands of less skilled census-takers out in an erratic, unreliable effort to canvass everyone. The 2000 census would be the first that did not even try to contact everyone. But it would also more accurately account for all Americans, especially minorities, than any of its predecessors.