roj - the raw census data don't control for a whole host of confounding factors, like age, experience, education, the fact that women might prefer to trade off career advancement or higher earnings for a job that offers flexibility to manage work and family responsibilities (and like it or not that tends to fall to women. Although in my family, I've chosen the flexible life while my wife has chosen the 8-5 executive life - and she outearns me. QED!).
So it's in fact wickedly difficult to determine whether the lower wages are due to discrimination or some other factor not included in the model. Still, there aren't any studies out there claiming that women earn more than men, controlling for as much as can be controlled for!
it will be interesting to have this discussion in 20 years (should we all still be here) given that US university enrollment in most well-paying professions is currently skewing at least 60/40 women to men.