The Evolution of U.S. Earnings Inequality: 1961–2002 (PDF; Zvi Eckstein & Éva Nagypál, Minneapolis Fed, Dec '04)

So far, I've discussed the standard components of a Mincerian wage equation: the probable earnings of a worker based on experience, occupation, and education; we've observed that gender is a distinct component, that statistically women tend to get a smaller return to income from experience, but a higher return to occupation. Returns to education and occupation have increased dramatically over the last 40 years, while both professionals and the more highly-educated cohorts have increased in relative size. Labor force participation has declined slightly for men, but increased substantially for women. None of these are surprises.

Racially, African American men and women have improved their relative position to whites—slightly; Hispanic Americans of both sexes suffered a marked decline in their relative position, which probably has to do with the dramatic increase in numbers, always at the lowest end of the national wage scale (and often in a sector with a lower minimum wage). None of these are surprises, either.

Education (by bracket), occupation (by bracket) and experience can be weighted for their distribution in the general population and multiplied by the predictive variables to get a predictive function of income distribution much like this one here (HC). Notice the preoccupation with standard deviation of income σ; the Gini Coefficient—the most widely-used measure of income distribution—has a unique one-to-one correspondence with σ. However, an advantage of using σ is that it can be decomposed into explained and unexplained components. Another advantage is, of course, that while a Gini Coefficient tells you nothing about how the income is unequally distributed (if, for example, the log of income is linear or normal), the σ does.

Eckstein & Nagypál's paper includes a figure at the beginning showing the progress of wage distribution over the years. Here is the figure:
Here is an unsurprising path of σ since 1961, from .57 to .61 for women (IOW, a ratio of 3.13 to 3.39 for the 16th to the 83rd percentiles); and an increase of σ from 0.52 to .69 (from 2.83 to 3.97). Under a normal distribution, slightly more than 2/3 of the population are within one σ of the median. For women, the ratio between the richest and poorest members of this group has remained somewhat more than three-to-one, while for men it has increased from three- to four-to-one. These are not very large changes, although they only cover the disparity in wage income, and they do reflect a disturbing overall trend.

Even within the minor component explained by Mincerian variables, SBCT has some logical problems:

In fact, the continuous rise of postgraduate wages also raises some questions regarding the SBTC hypothesis as it is used in the literature... The continued rise in the postgraduate wage premium is inconsistent with the view that the economy discretely moved from one stochastic steady state, prior to 1980, to another stochastic steady state, since the mid ’90's
[p.27b & fn 23]

Alternative leading hypotheses to explain the rising wage inequality include the decline of unions and the decline in the real minimum wage, or, more generally, changes in institutions. These explanations focus mainly on the reduction in earnings of the less educated during the ’80s and early ’90s. The role of globalization and the increase in trade and out-sourcing emphasizes the decline in the return to less-skilled workers in the United States.

It seems that it is possible that these factors were important between 1979 and 1993, when wages of the less-skilled went down. But the spectacular rise in wages of postgraduate workers that started in the mid-’60s, continued throughout the four decades, and increased in the second half of the ’90s cannot be explained by these alternative hypotheses.

I turn now to my own opinion as to the most plausible explanation for the growing disparities in income and in earnings. Since it has been expressed well by a commentor, and the hour draws late, I'll just use his words:

John C Halasz, comments: In general, I meant to express skepticism that the changing distribution of income results simply from the rational play of market factors, such as returns accruing to education [...] and the supposed increased marginal product of educated or highly "skilled", vs. semi-skilled labor. Rather I would think the focus should be on mechanisms that effect the distribution of economic surpluses between wages and profits, such as the decline of/assault on labor unions, the increased playing of working populations against each other, the uncompensated effects of increased "free trade", and the decreased relative cost of capital equipment as a substitute for labor, partly as a result of the development of IT