Usually, we generalize from other research: there is no reason to suppose CS is miraculously free of discrimination against women, considering the variety of areas in which this kind of discrimination has been shown exist.
There is also no reason to assume that all or most of the gender gap is caused by discrimination. If we generalize, we discover that lots of other fields (e.g., medicine, law) have sexism, but they also have 50% women.
Remember, we are asking what makes CS special.
Why do you have to outperform others before you can be discriminated against?
Suppose you are biased against green cars. You want to buy fast cars, and you generally only buy a car if it drives at least 80mph. However, since you hate green cars, green cars must drive at least at 100mph before you will consider buying one. Therefore, you will have plenty of slow (80-90mph) red cars, but your green cars will all drive at least as fast as 100mph.
Substitute "women" for "green", "men" for "red", and "productivity" for "speed", and you'll see the same conclusion applies to workers.
If we generalize, we discover that lots of other fields (e.g., medicine, law)
have sexism, but they also have 50% women
Firstly: in neither medicine, nor in law, are 50% of the graduate students women. Secondly: even when the number of students in the field is 50/50, that doesn't mean there is no discrimination. Here's a nice study about discrimination among students of medicine[1], the consequences of which you can guess.
Remember, we are asking what makes CS special.
No, that's what you are asking, because you assert that gender discrimination does not influence gender ratios in other studies and occupations, which is simply not true. Solely based on experimental observation of cognitive biases, it is extremely unlikely to be true. Women are always, everywhere, underrepresented in the top echelons. Even in nursing, there are more males in leadership roles than you would expect from the number of male nurses.
[..] and you'll see the same conclusion applies to workers.
Wait, what? The women have to perform better to be even considered and only if they are then rejected, then they are being discriminated against? It's the fact that they have to perform better in the first place that is the problem. You are already discriminating against green cars. That's the point.
(BTW, I did an experiment here: I upvoted your comment some time ago. I have a feeling that in a discussion of 'score 1' comments, bumping one of them to '2' precipitates upvoting of the '2' and downvoting of the others)
The women have to perform better to be even considered and only if they are then rejected, then they are being discriminated against?
You've completely misinterpreted what I said. Women overperforming at a given level is a consequence of discrimination, not a precondition. Discrimination is equivalent to a higher standard. If women (or any other group) are held to a higher standard than men, then they should overperform men.
Another (non-emotionally loaded) example: colleges have a higher academic standard for regular students than for football players, equivalent to discriminating against non-football players. As a result, non-football players tend to perform academically better than football players.
For this reason, I'd like to see a study that measures performance rather than opinion.
Oops, my mistake. The actual numbers are 49% and 49.4%.
Those are the numbers of applicants, not the numbers of students actually finishing their education, where women used to have a larger dropout percentage. But that doesn't matter: even if the same amount of men and women start and finish law school, that doesn't mean the women aren't being discriminated against. It may mean they are just working harder to overcome the disadvantages. That works in school, but it doesn't work in the workplace, as research shows, when it once again turns out that women in the exact same position as men earn less than their male counterparts.
If women (or any other group) are held to a higher standard than men, then
they should overperform men.
They probably do. However, it doesn't show, because their performance isn't rewarded equally. They may work harder work or have better brains, but they will not be rewarded accordingly, so there is no way to show, by position or salary, that they are better.
I'd like to see a study that measures performance
I feel that's rather like saying you'd like to a study that actually measures how a lack of oxygen causes death, instead of accepting that all kinds of laboratory experiments show that various part of the body can't do without oxygen. Men have discriminated women for centuries and it's impossible that we suddenly stopped doing that during the past century, as many experiments show. The only things that have disappeared are the egregious cases, where it is easy for anyone to point out there is discrimination. Now we can only show it through statistics.
I have the distinct feeling that I'm misunderstanding the point you are trying to make, because we seem to agree on many points. Must be some fundamental assumption about what we are discussing, but I can't put my finger on it :/.
They probably do. However, it doesn't show, because their performance isn't rewarded equally.
That's why I'd like to see studies comparing (women's performance)/(women's reward) to (men's performance)/(men's reward). If discrimination harm's women, the ratio should be higher for women.
I don't really trust opinion polls (except as a measure of opinion), particularly when they ask questions like "is your failure to get promoted/paid/etc caused by $YOUR_FAULT or $NOT_YOUR_FAULT?"
There is also no reason to assume that all or most of the gender gap is caused by discrimination. If we generalize, we discover that lots of other fields (e.g., medicine, law) have sexism, but they also have 50% women.
Remember, we are asking what makes CS special.
Why do you have to outperform others before you can be discriminated against?
Suppose you are biased against green cars. You want to buy fast cars, and you generally only buy a car if it drives at least 80mph. However, since you hate green cars, green cars must drive at least at 100mph before you will consider buying one. Therefore, you will have plenty of slow (80-90mph) red cars, but your green cars will all drive at least as fast as 100mph.
Substitute "women" for "green", "men" for "red", and "productivity" for "speed", and you'll see the same conclusion applies to workers.