U.S. finds undercounting of jobless, revises survey Women excluded from tally if they were homemakers

WASHINGTON — WASHINGTON -- The government has been substantially underestimating unemployment for at least a decade or more, particularly among women, the Labor Department reported yesterday.

As part of the biggest revamping in the 53-year history of the government survey that determines the official jobless rate, researchers at the Labor Department have determined a bias in the survey under which many women who were seeking jobs were erroneously described as homemakers and therefore not counted as being in the work force.


Under the redesigned survey, the nation's average unemployment rate in the 12 months through August was 7.6 percent, a half-point above the 7.1 percent that the department had reported. The rate for women was 6.8 percent, rather than 6.0 percent; for men, the figure went to 6.9 percent from 6.7.

The Labor Department also found that besides undercounting jobless women, it was also miscounting laid-off and so-called discouraged workers.


The redesigned questionnaire, which was tested in parallel with the present system for 12 months ending in August, will be used officially by the Labor Department beginning in January.

If the higher jobless rates persist when the new survey takes effect, they could affect benefit programs, such as state unemployment compensation systems, that are tied to joblessness figures.

Labor Department officials said they had suspected for some time that the survey, whose last major revision was in 1967, was failing to reflect changes in American society, such as the greater participation of women in the work force. But it took years to design the revision and begin to put it into effect.

The effort, begun in 1986 and costing $40 million, includes the collection of new data, rewording of questions, changes in definitions and computerization of the entire process, beginning with laptop computers carried by interviewers.

The first results using the new system will be published in early February; the old survey will be calculated, but with sharply reduced samples.

Marvin Kosters, a labor specialist at the American Enterprise Institute and its director of economic policy studies, said, "It's essential to move toward the use of new technology even if it means we have a little discontinuity [in the statistics]."

But he also said the figures published yesterday acted as a reminder of the "subjective" element in even the most sophisticated data, and he reserved judgment about specific changes.

Labor Department officials also cautioned against making too much of the experimental comparisons published yesterday, since the monthly variations were great in the year they ran their test, the samples relatively small and interviewers were still becoming acquainted with the refined questions and their new technology.


Nonetheless, the differences pointed up in the experimental results are considered important both inside and outside the Labor Department. Private analysts who have followed the redesign process generally applauded the effort, and some expressed little surprise at the results.

"The academic work that preceded this was careful, thoughtful and completely nonpolitical," said H. Erich Heinemann, an economist and student of the figures who works at Ladenburg, Thalmann & Co., a securities house.