Hobson's Choice

free script provided by

Income Inequality-3: Trends

January 27, 2005

[ 1 | 2 | 3 | 4 | 5 | 6 ]

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

Now we turn to the matter of evolving income inequality, as treated in the Eckstein & Nagypál report. Overshadowing every research program in the social sciences is a theory that purports to explain what is happening, and why it is happening. Usually debates on the latter hinge on a fine-grained examination of the former; so, for example, as the variance in incomes grows, economists or policy makers have suggested this is the result of increased preference on the part of firms for higher-educated workers. This theory is known as the Skills-Biased Technical Change (SBTC) theory, and suggests that the world economy is evolving to one in which there is ample demand for a large number of low-paid, semi-skilled workers, and for workers with extremely demanding, specialized skills; but not much in between.

This can be interpreted either in terms of skill level (measured, perhaps, by years of education) or by occupation (say, professional workers or blue-collar workers). As it happens, the disparity in earnings has grown at about the same rate for either breakdown. The authors also note that income inequality for women has increased at a lower rate than for men, while mean incomes have converged overall.¹

The wage gap between men and women declined as women’s educational attainment grew, and their workforce participation increased dramatically. Women made up a significantly smaller share of the workforce, yet accounted for 55 percent of the increase in the number of workers with at least some college education. Despite this fact, women experienced less of an increase in inequality than men did, and, in fact, it was in the most educated groups that women succeeded least in closing the wage gap. ...We show that inequality started to increase for men in 1974 and for women in 1981, and for both genders inequality continued to increase throughout 2002. This is a robust fact regardless of what specific statistic is used to measure inequality.
[p.12]

Note that Eckstein & Nagypál use the Current Population Survey (CPS) to measure earnings of workers, rather than incomes of households. The last point—that the wage gap persists more strongly in the upper percentiles of women, or non-whites, has been very frequently observed. And, as I remarked in Part 2, at the top-most brackets, there are weird anomalies that reflect a very clear segmentation of the managerial labor market.

Here is a chart comparing the increasing spread in earnings by education.
It illustrates the shift in labor demand away from less-educated workers (who were, in turn, declining in numbers) to postgraduate workers; the income spread for women was far smaller, reflecting a massive increase in the supply of higher-educated women; yet, as indicated by their wages failing to reach parity with comparably-educated men, it appears that the labor market remained segmented by gender; even for professions, it seems jobs tend to be "coded" by gender. (Part 4)

NOTE: 1 The reason for this and other phenomena outlined in Eckstein & Nagypál is somewhat mysterious when analyzed simply by comparing quintiles; however, when one analyzes income distribution as I outlined in part one, some of these issues clear up. Income in the USA, while not conforming perfectly to the normal-log distribution, has a very close correspondence for the great majority.

Now, suppose we have a predictive function for income thatconsists of multiple cohorts—women and men, different ethnic groups, different states. Intuitively, one would probably expect it would make difference in the Gini Coefficient C_G if the medians for these groups were themselves widely distributed; yet this is not true. Only the standard deviation σ for each group influences the Gini Coefficient. This has the bizarre corollary that one way in which inequality over an entire society can increase is if a glass ceiling is removed!

Here is an illustration of what I mean: in the chart above, the red curve represents an underpaid half of the population; the blue curve represents the other half. Both have the same internal Gini Coefficient, and the total population (green) has the same Gini Coefficient as the two constituent populations. Now, suppose the median income stays the same but the σ of log-income within the underclass shrinks by 33%.

The underclass is poorer than it was before, on average (although everyone below the median is now richer); yet the Gini Coefficient is actually lower for the overall society than it was before (income distribution is "more equal"). What do these two charts represent, in plain English? The second chart represents a "traditional society" in which women's roles are more limited; there are fewer women in the labor force and they are confined to a narrower range of roles. In the later period (1st chart), women now occupy a broader range of roles, but there is still discrimination.

If median income for women increases (making it closer to that of men), while the σ increases too, then both effects will tend to increase average incomes for women, while resulting in a higher Gini Coefficient (less equal). This sounds paradoxical, but in reality all it means is that inequality of income becomes less a men-versus-women issue, and more of a class issue.