[gyardley]

Yes, it would be nice to see the raw data, but note that they've bucketed Level 1 through Level 6 employees, and have still found gender discrepancies.

Are you suggesting that different jobs might occupy the same level but have widely differing compensation? That might be true - I don't know Google well enough - but that's not how salary bands usually work.

[guitarbill]

It's exactly how salary bands work - they're designed to be opaque.

If you have women who largely choose to work in Foo with a compensations which puts them at the top of the band 1 and at the bottom of band 2, and men who work in Bar that starts with compensation at the bottom of band 1, but a promotion gets you to the middle of band 2, then you'll get that in band 1, women earn more, but in band 2, men earn more. This is why you must know the actual distributions (which includes weighting and deviation/variance), and the job role.

Indeed, in level 2, seems like women earn more - maybe? might not even be significant though.

In any case, I'll bet job role correlates far more strongly with compensation than gender.