You refer to human instinct on an issue where there is no consistent pattern. Discrimination levels, for example, differ considerably across countries. We also know that positive discrimination works very effectively. See, for example, the lower male-female wage differentials in continental Europe compared to Britain (with the latter following a labour flexibility approach that traditionally avoids positive discrimination measures). As I said, you've gone for the shallow in order to avoid the actual evidence.
But how do we know the Oaxaca methodology captures the effects of discrimination? It could suffer from omitted variable bias and overstate discrimination or instead these differences in human capital are DUE to discrimination (I.G. women don't go to school for as long because their returns are less). I guess you could use a couple decompositions and use them as a way to bound the discrimination in an economy. For instance, due to the fact Oaxaca is used in courthouses to settle lawsuits, we could predict it could be abused by economists, trying to bias the results one way or the other. Remember that discrimination lawsuits has the potential to cost millions of dollars.
Its actually more likely to underestimate discrimination. It doesn't take into account, for example, the heterodox approach that acknowledges that all parties suffer (e.g. analysis into divide and conquer, rather than the analysis into nepotism). Testing for problems such as omitted variable bias is, in contrast, a relatively minor test of robustness
Hadn't thought about that. I suppose Oaxaca assumes that the higher paid party isn't suffer any losses due to discrimination. edit: to add something to the discussion, It has always bothered me when economist assume there are "non discriminating" firms out there. I find it hard to believe we don't all discriminate in some way (even if it is small). We might not be aware of it, but we certainly do it. It could be as simple as being more likely to interrupt a women as opposed to a man when she speaks.
Its certainly a particularly neat approach for looking at Becker's discrimination/nepotism methodology, but we'd need something more to consider the consequences for underpayment. I'd suggest the stochastic frontier approach, but we'd then be confusing it with other labour market inefficiencies
Something else that has bothers me is that minorities suffer higher levels of unemployment, which isn't explained by Becker's model (as it just predicts lower wages). Forgive my ignorance, but this is where we estimate a production frontier and then determine if the firm operates on it, correct? I've read studies of this being applied to labor unions, to determine what model best describes their behavior. Speaking of unions, do you know of any data if they eliminate or exacerbate inefficient difference in wages for minorities. I could see union officials being nepotistic on one hand, but on the other, trying to represent all their workers regardless of race on the other.
Typically used for production or cost frontiers. But it can be used to measure underpayment. There's probably too much noise though to isolate discrimination effects. Offhand, nope. I'd predict ambiguous effects. In orthodox analysis (with, say, a 'right to manage' model of union activity twinned with a 'taste for discrimination' employer rent seeker) a union is likely to reduce inefficient differentials. However, once we move to the likes of 'insider-outsider' analysis into bargaining all bets are off. As you mention in your post, we have differences in unemployment rates. We therefore can have significant effects on hierarchy and wage gaps accentuated through wage norms
