Re: my post immediately below on signaling, Bryan thinks I haven’t answered his question about the importance of signalling. But I have, he is just confused because I don’t use the exact same normalization as he does. In any case, I postulate the wage return to signalling as going away within five years, in say a career of forty years, then with the measure adjusted for the presence of capital and resource income. You can express that in terms of totals, variations, percentages, as you wish but the point remains that signaling is only a temporary factor, and overall only somewhat of a marginal factor (5-10%?) in explaining the overall evolution of wages. That is why it has lost ground to human capital approaches, all the more so with increasing inequality. (One can believe all of that and still think, as I do, that we could organize current education more efficiently and at lower cost.) I also stand by my points that insisting on the break down is missing the more important points about indeterminacy and nestedness and those points too can be applied to any normalization of the units. It would be more useful if Bryan would outline where he disagrees with my assessment, as the entire chain of reasoning is laid out pretty explicitly.
By the way, the easiest way to boost the contribution of signaling is to invoke the “you got your first job by signaling and then from that job quickly gained persistent extra human capital” argument, but even then that increment can, under traditional measures, be assigned to human capital. (You don’t want to rule out all human capital influences, on the grounds that signaling helped create them, any more than you wish to classify gains from signaling as human capital, if the human capital helped you get into the position to signal. But if that doesn’t convince you, revisit the earlier point about indeterminacy, as you can see that the marginal products for human capital and signaling will sum to well over one hundred percent.)