As was also mentioned earlier, there are studies that show that it is indeed the case that there are women being paid less for the same role. Not to mention enough equal pay lawsuits. Personally I think the pay issue would reduce if men were granted equal parental leave. From the employers point of view, a woman is a risk to them - if she gets pregnant, they still have to pay her, AND have to find a temporary staff member to cover.
Those studies are mostly older, and the error is in interpretation. The studies generally take an averaged sample across a particular field/industry, and find the average salary for women is lower than the average salary for men.
However, this is not attributable to the companies refusing to pay the same amount for women; in fact, for most employers the contracts are standardised within the company meaning that there is no difference between what male employees and female employees get.
The difference in average salaries arises because there is a much greater population of men in the upper echelons of these companies; for various reasons (old boys' club, women leave to start families before they spend enough time there to get promoted all the way, etc.). This is a problem (i.e. women are not being given the same opportunities to reach the top, the so-called glass ceiling), but it is not the same as saying there is a systematic bias in wages against women; i.e. a given woman will not be paid less than a given man if they both compete for the same role; a given woman will have less chance of being promoted/hired than a given man for a particular role.
The note about parental leave is interesting, though; I would expect that salary is reduced to incorporate parental leave, however this would mean that the contract value of employment is the same, and so to say the women are being underpaid would be misleading; they're getting the same calculated value from their contract as men, it's just that their part of their base salary is converted into a different form.
@OJ Simpson: Actually, one of the defining aspects of the reasonable doubt in that case was because the defense attorney misrepresented the mathematics behind the statistics of wifebeating.
OJ had a history of violence against his wife, but the defense lawyer pointed out that only 5% of people who beat their wife go on to kill them (and therefore 95% of people who beat their wife do not kill them). This, however, is the wrong statistical analysis. While that statistic is true, what is also true that men who do not beat their wife are 0.001% likely to kill them. Consequently, a man who beats his wife is 5000 times more likely to kill her than if he didn't beat his wife.
However, this argument successfully prevented the jury from considering his past in making their determination of guilt.
So, in summary: Everyone should learn how Bayesian Reasoning works if they are going to be doing any considerations with statistics.