[Shubley]

Let's compare this to height as you did.

Because of how gaussian distributions work, the differences at the outlying values are huge even for small differences in the average.

For example, men are taller than women in general, but of course some men are taller than some women.

But at the extreme heights, the ratio between men and women becomes stagger. At 6 feet tall, men outnumber women 30:1. At 6'3", 2000:1.

Now apply this to tech. If we imagine a relatively elite job, like being a professional programmer with significant responsibility, that is the kind of thing only someone at the extreme end of the bell curve of tech-proclivity is going to do. But because of the way gaussians work at extreme values, even if women on average have similar tech-proclivity to men, at that high level the ratio of men:women could be huge.

All this could just be due to a simple misunderstanding of the math of a bell curve distribution and how it works at the ends of the spectrum.

[geofft]

Sure. But the point of the height analogy was that, first, you don't need to be 6'3" to reach the top shelf of a kitchen, and second, there are so many skills required to cook, not just reaching the top shelf.

In the same way, even if we assume that women have a lower bell curve than men for, say, ability to focus on some algorithmic problem, first, you don't need to be the equivalent of 6'3" at algorithms to do a good job as a professional programmer, and two, algorithms is only a very small part of what you do.

For the bell-curve hypothesis to work for cooking, women would also have to be weaker at using knives, worse at visually comparing the volume of liquids, worse at keeping track of time, worse at tasting things for flavor balance, etc. etc. - every single one of these axes would need to have a bell curve lower for women than for men. That's just implausible. Same with programming.

Of course, if you set up your hiring / recognition processes to look for the 6'3" algorithmists and not for the 5'10" algorithmists or the people who do any of the other work than algorithms, you'll see a 2000:1 ratio. But I think that's a sign of the hiring and recognition process failing to find truly qualified candidates of any gender, and just using the easiest metrics instead of the best ones.