The “gender pay gap”: an exercise in statistical control

Andrew Biggs has an interesting article called “The Gender Pay Gap Is a Media Myth” at the American Enterprise Institute, in which he explores a recent study of said pay gap, and finds its dimensions change rather dramatically as the raw data from the study is “controlled” in various ways.

The study, published last week by the American Association of University Women, offered an initial “headline” result that “new female college grads are paid only 82 cents for each dollar of male earnings.”  Naturally, this headline was reported far and wide.

But even the AAUW researchers knew their headline factoid wasn’t really true.  They did the necessary diligence and controlled the data for such factors as the major fields of study chosen by the graduates – some of which pay much better than others – and the greater number of hours male graduates were generally willing to work.  With these considerations duly accounted for, the number changed to roughly 93 cents on the dollar for female graduates.  Biggs gives the AAUW “credit” for performing these calculations, although it’s curious that this much more comprehensive calculation wasn’t the “headline” conclusion of the study.  (Of course, the reason is fairly obvious: a 7-cent “gender pay gap” isn’t much of a headline.)

Biggs then decided to introduce some further controlling factors, working from Census Bureau data.  He stuck with recent college graduates who remained employed during the entire year, from 2009 to 2010, and “controlled for age, race, Hispanic and immigrant status, detailed geographic location, weekly work hours, college major and occupation.”  He took special care to control for “compensating wage differentials,” which is another way of saying that “dangerous or unpleasant jobs may pay more, while jobs offering flexible hours or more generous benefits might pay less.”  Men and women tend to apply for these “lucrative but lousy” jobs and “awesome jobs that don’t pay as much” in different proportions.

“Including all these controls, the gender pay gap for young college grads drops to around 1 percent,” Biggs concluded.  (Emphasis mine.)  He went on to note that some of this 1 percent gap might be accounted for by the “well known” tendency of women to “negotiate over pay less aggressively than men.”

It’s hard to infer widespread gender discrimination, which can only be addressed by far-reaching legislation, from such a narrow “pay gap.”  Fortunately for the Left, the media is willing to work very hard to make that inference.  Many of the controls Biggs applied to his data are simple common-sense observations to the average working man and woman, but with enough political marketing, many of them can be persuaded to set common sense aside.