Roanoke Times
                 Copyright (c) 1995, Landmark Communications, Inc.

DATE: FRIDAY, November 18, 1994                   TAG: 9411180113
SECTION: NATIONAL/INTERNATIONAL                    PAGE: A-15   EDITION: METRO 
SOURCE: Associated Press
DATELINE: WASHINGTON                                LENGTH: Medium

PANEL TELLS CENSUS BUREAU TO TAKE 'NO' FOR ANSWER

The goal of trying to physically count every person has proven costly and unrealistic, the National Research Council said in a new analysis. Diminishing returns set in when repeated efforts are made to reach people who don't respond, yielding only small gains at great cost.

But the panel endorsed the basic census process of mailing out forms and even defended the long form that has been criticized as being so complex it could discourage people from cooperating.

New Census Director Martha Farnsworth Riche welcomed the study, noting that a test census is planned for next year to evaluate some of the survey and sampling systems recommended by the panel.

``They are basically saying that we have gone to the end of that avenue'' in sending people again and again to the homes of people who haven't been counted, she said. ``The great cost is doing the job and getting physically to every American.''

Instead, the research panel's 460-page report ``Modernizing the U.S. Census'' calls for increased use of surveys and statistical estimates to round out the numbers compiled in the basic head count.

Pure counting ``has been pushed well beyond the point at which it adds to the overall accuracy,'' said the report.

``Statistical estimation can be used, in combination with the mail questionnaire and a reduced scale of follow-up of nonrespondents, to produce a better census at reduced costs.''

``Correctly designed, this approach reduces the need to spend large amounts of time and money following up on people who don't return the questionnaire,'' said Charles L. Schultze of the Brookings Institution, who headed the research group.

The census is required every 10 years by the Constitution, but rising costs have brought demands for more efficient methods at the same time government agencies and others require more detailed information about the country.

``If we really want to do the two things Congress wants - be more accurate and spend less - we will need to use more statistical tools rather than more workers,'' Riche said.

In 1990 census takers made as many as six personal visits to the homes of people who did not return the mailed forms, yet millions of people still were missed. The cost of the 1990 count was about $25 per housing unit.

Under the research panel's plan, mail questionnaires would still be used and workers would visit homes that did not return them - but the follow-up period would be shorter and less intense.

After a reasonable period statisticians would estimate those not counted for inclusion in the final total.

Alternative proposals ranging from a rolling census over a 10-year period to compiling a count from records such as tax files and Social Security information were rejected by the panel.

While politicians have criticized the long census form, contending that it discourages people from cooperating, the study defended it. It said the length of the form had little influence over whether or not people responded.