Gender bias taints unemployment rate

What were you doing most of last week? Were you working, looking for work, or keeping house?

The question may sound innocuous enough, but because of its final three words, the U.S. unemployment rate was substantially underreported for nearly three decades until this January.


That is because pollsters for the Bureau of Labor Statistics asked the question, in that form, only of women. For men, the question left off "or keeping house."

"A lot of women chose to say that they had been keeping house, even though they might have worked part time or spent a few hours looking for work," bureau economist Peter Cattan said.


"That wording apparently proved to be a form of gender bias," he said.

So it did: In test runs last fall, bureau surveys showed that by removing the "keeping house" option and asking women the same question that it asked men, they got a national unemployment rate of 7.3 percent, compared with 6.8 percent in the official polls using the old question.

So the unisex wording went into a revised questionnaire that was put in use in January.

What's at stake here is more than just the wording used by bureaucrats on an employment survey.

We are talking about tools of decision-making that businesses and government agencies use in calculations that affect people's pocketbooks every day. Scarcely any American would be unaffected by the calculations that would flow from a 0.5 percentage-point difference in the unemployment rate.

The Federal Reserve Board never leaves the U.S. unemployment rate off the Ouija board that it consults before raising or lowering the rates Americans pay to borrow money or to use credit cards. Congress looks at the rate whenever it considers extending unemployment compensation benefits.

At election time, politicians blame each other for the unemployment rate. Trade negotiators use it to make the case that foreign "dumping" is hurting American workers.

The tale begins with a Labor Department commission that recommended a broad range of changes in the Bureau of Labor Statistics Current Population Sur


vey -- in 1979.

With all deliberate speed, the department put a team right on the question -- in 1988.

"Maybe it took that long to get all the people they needed to initial it," Maureen McDevitt Greene, of the bureau's Philadelphia office, said with the knowing snicker of one who has spent years in the labyrinth.

The research team recommended a new questionnaire in 1991, and by last fall the new format was tested. It went into use in January.

The revised survey is making the U.S. unemployment rate more accurate -- that is, higher -- but more cumbersome for economists to work with, because it is not easy to compare this year's figures with those of previous years.

It also has demonstrated that unemployment among women is substantially higher than previously believed, almost as high as it is among men. In test runs, unemployment ran to 6.4 percent among men and 5.8 percent among women using the old questionnaire. The new form put it at 6.6 percent among men and 6.3 percent among women.