From Anne Phillips's “Defending Equality of Outcome” [pdf]:
When differences in outcome are explained retrospectively by reference to differences in personal preference, this assumes what has to be demonstrated: that individuals really did have equal opportunities to thrive. In many case, moreover, these explanations reproduce ideologically suspect stereotypes about particular social groups: that 'women' for example, care more about children than men, or have less of a taste for political power. When outcomes are “different” (read unequal), the better explanation is that the opportunities were themselves unequal.
I haven't read the whole paper yet, but this strikes me as forcing a false alternative. Phillips makes an excellent point about not just assuming that differences in outcome reflect a simple difference in preference. This point cuts especially deep when we acknowledge that preferences are not formed in a cultural vacuum, but can and often do reflect entrenched prejudices and social expectations at odds with a decent measure of individual autonomy, a requirement of equal opportunity. (Which is why, if you care about autonomy, it is insane to think sexism and racism are non-problems just as long as there are no legal barriers.) I think this is pretty well undeniable.
That said, it may also be the case that men and women, say, would tend to have differences in preference even when their preferences are sufficiently autonomously generated and reflectively endorsed — even in a climate of truly equal opportunity. Indeed, I think it would be pretty surprising if they didn't. And these no doubt will tend to affect patterns of outcome. Furthermore, unless we assume, bizarrely, that autonomous preferences must be either uniform or random, then there will be autonomous, reflective, culturally correlated preferences, again generating different patterns of outcomes for different groups.
Whether inequality in preference or inequality in opportunity is a better explanation for inequality in outcome is an empirical issue. And we should try to figure out what's going on in various cases. There is lots of evidence for heteronemous social pressure and unequal opportunity in all kinds of cases. And there is also plenty of evidence for sex differences and benign cultural variation that could effect outcomes. In most cases, it's not clear how much weight to put on either consideration, but it really is something we ought to just go ahead and dig into and fight about. That's how you find out. (There is also a matter of sex-, population-, and culture-based variations in skill, which is whole other question that bears on differences in outcome.)
Maybe this is the big complication: If there are cross-culturally robust, non-culturally-constructed sex-based tendencies in preference, for example, then cultures may not only reflect but harden and exaggerate the differences in tendency, creating cultural expectations that exert pressure on men and women at the tails of their respective distributions to form preferences not fully “theirs.” Another way to put it: “naturally” emergent cultural expectations that reflect the average “natural” preferences of a group may raise the social cost of expressing statistically deviant preferences past the point that most people are willing to pay. So you get broad conformity in preference expression, but lots of people would have expressed non-conforming preferences had the price of expressing them been lower. If the problem of empirically determining whether or not different preferences determined different outcomes requires distinguishing between preferences that were supressed by the price of non-conformism, then we're probably in trouble.
Now, by “equality of opportunity” do we mean that everyone faces the same cost of expressing their “natural” preferences? I don't, since people with statistically deviant preferences will almost always face a higher cost of expressing them, unless they're also statistically deviant in not having a preference for conformity. Further, some preferences just aren't compatible with a desirable social order, so we need to keep the cost of expressing them high.