>>Yup. I've read quite a few discrim studies, and I know that you have to take all kinds of factors into account - job titles, education, experience, industry, etc. etc. You're right that just using raw numbers like that is misleading. But accounting for differences in education, experience, job titles, etc. assumes away a lot of the "pre-market" discrim that tracks people into occupations by gender (e.g., voc ed classes that encourage men to be plumbers and women beauticians), or the lack of child care options that hit women much harder than men, etc. That's discrim too, but of a harder to measure sort.
My problem with many of such discrim studies is that they approach discrimination as "residual variance" - that is they explain the variance in income by first adding the "human capital" variables (education, skills, experience, industry, etc.) to the regression equation and then introducing sex or ethnicity and seeing if it explains some additional variance (if any). The problem is that this "additional variance" explained depends on what variables are already in the equation. It is possible that "sex" or "ethnicity" is simply a proxy for another "human capital" variable not included in the equation - so its effects (variance explained) would disappear if that not included variable were introduced. Or maybe not - it is guessing game.
Thus the inclusion of all relevant factors in the model is essential. But as you correctly observed, some of these factors are "pre-market" i.e. diffused in social fabric and thus difficult to conceptualize, let alone measure. For example, to what degree is the tracking to different occupational paths a "discrimination" or an individual choice designed to take advantage of opportunities created by societal expectations. The study which I cited in my previous missive (Reskin & Roos) attempts to address that issue by incorporating both elements. Their argument is that societal expectations create different perceptions of the potential value of the prospective employee that are based, in part, on factors such as sex or race - but these perceptions affect both employers and employees. Thus a prospective employee may perceive herself as "less qualified" (according to social stereotypes) and scale her job expectations accordingly (e.g. by applying for positions for which she thinks she has a greater chance of success). How would you call that behavior - "self discrimination?" Sounds too damn close to "false consciousness."
In essence, the problem is much more complex than con or lib tropes have it. My main beef is that such tropes are used to excuse all kinds of undesirable behavior - a black man or a woman did not get a job - must be the discrimination; a white man did not get a job - it must be "reverse discrimination" - AA that gave "his" job to some "unqualified" minority or a woman.
wojtek