I agree, they aren’t exactly the same. Being disagreeable makes you that much worse at almost every job. Having curly hair doesn’t. Why would a rational employer want to pay worse people better on average?
What I can tell you is that being disagreeable by itself isn’t enough to get more money. Any ******* can walk into their boss’s office and demand more money. Actually convincing your boss to pay you more requires additional masculine displays. Always confidence, often stubbornness, sometimes even boorishness. So is it really that disagreeable people dig their heels in and demand more pay, or is it that the bosses admire them for having the balls to do it?
This is, again, the problem with an overfit model. Also with multicollinear variables. Saying things like, “but my model says you’ll close 5% of the gap by being less agreeable!” is wrong because that isn’t how statistical models work. If disagreeability correlates with masculinity, and masculinity correlates with at all with confidence, your model says literally nothing about whether disagreeability or confidence is important. It says either of them are important, or both of them, or neither of them if another variable is collinear. Literally the only thing this model says is that acting like a man means you’ll get paid more.
Because statistics.
Basically this is what they did. They didn’t like what the univariate analysis said, so they picked a whole bunch of behaviours that are well known to correlate with gender and added them to the model. So now you have some 20 variables each contributing 5%, and this little flappy vestigial t-rex arm gender variable that still explains a little bit, but only because they couldn’t think of any more variables to perfectly describe the average woman. Such a model doesn’t actually say anything - it is meaningless, obfuscatory garbage. A statistical dog whistle.
Look up overfitting and multicollinearity.
Detecting merit being the dual problem to detecting privilege.
Also the fact that women are under immense social pressures during childhood and adolescence to behave certain ways and adopt certain preferences. For example, female students exhibit both more skill and more interest in mathematics during mandatory education, so where the **** are all of the female mathematicians, physicists, and engineers?
I can’t speak for other fields, but I can for software. It’s socially acceptable for a young man to nerd out, stay up all night in his bedroom tinkering with a computer. It’s not socially acceptable for girls. It’s socially acceptable for boys to hang out with other kids to talk about computers, but it’s not appropriate for a girl to hang out with a bunch of boys. Girls aren’t going to be top students in undergraduate, they’re at too much of a disadvantage. Which means they’re also going to be selected against by top employers, so they get shunted into worse jobs. And then the sexual harassment starts.
I can easily imagine other industries working out the same way.
Regulating ratios obviously isn’t going to work, but maybe making it so people with feminine behavioural traits can make as much money as men will show that women are at least welcome in the industry? idk.