[geofft]

> Let's assume ability is independent of sex, completely measurable, and completely random, and remember that 80% of engineers are men (a fact an individual employer can't change in the short term).

> Under those conditions there's an 80% chance the best candidate is a man,

Hold up, I think you skipped a step there. I claim that 80% of engineers are men because of sexist employment practices, not because 80% of people qualified to be an engineer are men and employment practices are entirely unbiased.

[leereeves]

> I claim that 80% of engineers are men because of sexist employment practices

Only if requiring a relevant degree is a sexist employment practice.

The 80/20 ratio begins in college, or earlier. Again, not something an individual employer can correct for in the short term.

[geofft]

> Only if requiring a relevant degree is a sexist employment practice.

One, there's a difference between necessary and sufficient. You can get a degree with a 4.0; you can get a degree with a 2.0. It seems entirely plausible to believe that (especially if, as argued by James Damore, men are attracted to tech because it's a high-status field) the number of men who are good at their job is a small subset of those who have a CS degree.

Two, if we believe that the 80/20 ratio begins in college and is reflective of biases and not actual ability, then it's at least reasonable to argue that requiring a degree is a sexist practice. (It might not be, too, but I think it's not 100% clear that it isn't.)

Here's a report from HBR on why women interested in studying engineering in college leave: https://hbr.org/2016/08/why-do-so-many-women-who-study-engin...

If there is a social effect from a sexist culture that discourages women who are struggling with their engineering classes from staying with their original choice of degree, but doesn't equally discourage men who are struggling with their engineering classes, we'd definitely expect to see the phenomenon I suggested above where more men receive engineering degrees than women, but that's not reflective of men being better at the work on average.

(Also, the HBR report points out that internships and job prospects have the effect of pushing apparent sexism back down the pipeline into college. People are smart enough to tell that they're better off picking a different field if they'll be unhappy in their current one. So causality can certainly be such that sexist practices in industry hiring create visible gender imbalance in college degrees.)

Three, plenty of tech companies have quite loudly proclaimed that a CS degree is not something you need to get into tech. Peter Thiel literally runs a fellowship encouraging people to drop out of college. Lots of famous founders are college dropouts. It makes little financial sense to go back and get a CS degree if you didn't make the decision to get one at 18; requiring one doesn't seem like a high-quality hiring practice, even leaving demographics and discrimination (in whatever direction) entirely out of the question. You'll have tons of false negatives.

[leereeves]

This is all quite irrelevant to the question at hand: an employer picking the best candidates for the job immediately available.

Whatever the reason women drop out of, or don't enter, the field, an employer looking for people to hire will find that 80% of the candidates are men.

[geofft]

No, it's very relevant.

I am claiming that if 80% of the candidates are men and 20% are women, that's quite likely to be because that pool has far more unqualified men than unqualified women. Maybe it's 5% qualified men and 5% qualified women. Maybe it's 20% qualified men and 10% qualified women, which is still only 33% women, but very different from the original pool.

There is zero reason to believe that a process that specifically looks for the most qualified applicants is going to look the same as a process that chooses blindly and randomly from the applicants!

The only reason you'd expect that outcome to be guaranteed is if your hiring process is literally broken and has been picking people at random -- which is a good reason to move to one that actually tries to hire the most qualified people -- which is literally what the employer said it was doing.

[leereeves]

Then you're making a sexist argument that ability is correlated with gender among the candidates.

That may be true but it's not a neutral assumption. You'll have to provide evidence.

[geofft]

> Then you're making a sexist argument that ability is correlated with gender among the candidates.

I don't know what you mean by "sexist argument," but yes, I am making an argument that because of pervasive industry-wide sexism, ability is not independent of gender in applicants to tech jobs.

I agree that I need to provide evidence for this (and I'm sort of intending to write up something arguing this, but I'm at a conference this week). But I Googled for 'average gpa men women computer science', and the first paper I found shows that retention rates in college are worse for women than men year-over-year, but GPAs are worse for men than women year-over-year, which matches the hypothesis I came up with earlier.

https://engineering.purdue.edu/MIDFIELD/Papers/2005-graduati...