[geofft]

That interpretation is misleading for a few reasons.

One is that "to the point of vanishing" is still very nonzero. There's a report by Consad Corporation prepared for the Department of Labor that reaches the figure 5-7% after accounting for differences in jobs (their website is down at the moment), even though it concludes that "there may be nothing to correct." That's also the number Sommers herself writes in her HuffPo article she cites in that Time article: "After controlling for several relevant factors (though some were left out, as we shall see), they found that the wage gap narrowed to only 6.6 cents." So the Cornell claim of 8% is basically the same. I don't think there's an argument to be made that a 23% gender pay gap is a scandal but a 5-8% gap is totally fine.

The other is that occupations, positions, education, job tenure, or hours worked is affected by existing wages. I was unemployed by choice for several months last year, learning Rust and working on a few open-source projects, and I became a more valuable job candidate as a result of that. I could not have done that without the savings to take those months off. If my previous job had paid me anywhere from 8% to 23% less, I would probably not have decided to take those months off, and in return I would have been a less competitive candidate and my job would probably have been one of lower position.