The Obama administration’s continuing to tout the widely discredited claim that women still earn only “77 cents on the dollar” that men earn has been much in the news lately. For reasons I explained in this old, long post (see the last two thirds or so of that post) discussing my involvement in a big case (at the time the largest that ever went to trial) involving pay (and other) discrimination against women, this current controversy gives me a strong feeling of déjà vu. Thus my title doesn’t refer to the familiar (to us old people) Sears, Roebuck catalog but to the voluminous record of EEOC v. Sears, Roebuck and Co. 628 F. Supp. 1264 (1986), 839 F.2d 302 (1988).
One of the most insightful comments on the current debate comes from Megan McArdle. (In fact, McArdle is one of the most insightful commentators on just about everything economic, especially Obamacare; you should read her regularly.) I do, however, have one quibble with one thing she says, along with almost all other commentators on this issue. “The residual gap that’s left after you control for age, experience, work hours, choice of profession and so forth, is small,” she writes. “But it’s not zero. That residual most likely represents sexism.”
Not quite so fast, please. The EEOC case against Sears began, as do most similar arguments, with the assumption that pay gaps are caused by discrimination. When that assumption is tested by controlling for the obvious variables, the gap is reduced for each variable tested. In Sears, the statisticians controlled not only for all the obvious variables — age, experience, hours, etc. — but an almost endless list of every variable on which we could find data, and every single one reduced the pay gap incrementally. Now, it may well be the case that any remaining gap is the product of discrimination. But it also, perhaps just as or even more plausibly, may be the case that all variables are not known, or not measurable, or that our measurements are imperfect.
What this amounts to is thus what it amounted to (to no avail) for the EEOC: if you can’t disprove my accusation, my accusation is correct.