Of all the economic debates with broad political implications, none competes with the puzzling rise in American income inequality since the late 1970s. Both economists and politicians disagree about the forest and the trees – the overall interpretation of rising inequality and the importance of individual causes. People argue about differences between data sources, about the causes of sinking relative incomes in the middle and bottom percentiles, and have especially contentious disagreements about the interpretation of the leap of relative incomes at the top end. Goldin and Katz (2007) and Piketty and Saez (2003) describe changes in the income distribution over time. In this column we focus on possible explanations for the observed changes at both the bottom and top of the income distribution.
Our initial work on inequality (Dew-Becker and Gordon, 2005) started from an attempt to understand the differential between the growth of mean and median labour income. We documented an important and simple fact: over the period 1966–2001 only the top 10 percent of the income distribution had real compensation growth equal to or above productivity growth. Accordingly here we refer to the lower 90 percent of the distribution as “the bottom” and the top 10 percent as “the top.”
If real compensation growth is roughly equal to productivity growth, then labour’s share of national income will be constant. Figure 1 shows that in fact, over the full period 1950–2006 labour’s share has risen, not fallen. The dotted line adds in the labour portion of proprietors’ income, and shows that labour’s share has been almost exactly flat for more than 50 years. This implies that the growth of mean labour income has been roughly equal to the growth in productivity. But our finding that the bottom 90 percent did not enjoy real income gains equal to productivity growth implies that the growth rate of median income has lagged significantly behind growth in the mean.
Gross measures of the income distribution, however, mask important changes for different groups of workers. Some of the most striking disparities are between men and women. In figure 2, we focus on ratios of percentiles: the 90-10, 90-50, and 50-10 income ratios for both men and women.
Starting with the 90-10 ratio, we can see that the increase for women was fully twice what occurred for men. While the 90-50 ratio for both men and women increased slowly and steadily from 1979 to 2005, the 50-10 ratio showed a sharp jump in 1979–86 that was twice as large for women as for men. Then the 50-10 ratio remained on a high plateau for women about 20 percent above its 1979 value, while for men the 50-10 ratio gradually slipped back to its 1979 value. Over the full sample, changes in the 90-10 ratio for men are driven entirely by changes in the 90-50 ratio. For women, though, the 90-50 and 50-10 affect the 90-10 ratio roughly equally.
Explanations for changes in inequality
The sharp concentration of the increase in the 50-10 ratio for both men and women on the 1979–86 interval provides strong circumstantial evidence for declining unionisation as a cause for men and the declining real minimum wage as a cause for women. However, our examination of the evidence finds only a small role for the decline in unionisation and only for men.
The timing of the subsequent post-1986 evolution of the real minimum wage is also consistent with the 50-10 ratio rising for women relative to men. Women are roughly twice as likely to be paid the minimum wage as men (see Bureau of Labour Statistics, 2006), so we would expect to see the decline of the real minimum wage affect them much more strongly.
The standard explanation for the rise in income inequality is “Skill-Biased Technical Change,” or SBTC. In the simplest version of the SBTC hypothesis, changes in wages are driven by differences in demand for different types of workers. Because workers with some skills are imperfect substitutes for workers with other skills, when demand for one skill increases, relative incomes will rise for people with that skill. At first glance, this story doesn’t seem to fit with the data. For example, wage increases of skilled occupations like engineers and computer programmers have been remarkably slow compared to the rapid income gains of managers. Moreover, it is difficult to think of what particular skill set would account for the meteoric rise of incomes in the top 1 and 0.1 percentiles (see Piketty and Saez, 2003).
Autor, Katz, and Kearney (2008) adopt a three-way distinction between a high-income group doing non-routine cognitive work (including doctors, lawyers, investment bankers, etc.), a middle-income group doing routine repetitive work (accountants, engineers, computer programmers), and a low-income group doing manual but interactive work (truck drivers, nurses, waiters). This distinction emphasises that work at the top and bottom is inherently interactive and is less prone to outsourcing than the non-interactive middle jobs. SBTC has increased the demand for people in the top group.
This enrichment of the concept of SBTC helps to answer an objection we posed in our 2005 paper, where we cited evidence showing that there was no relative increase in the starting salaries of engineering and science BAs in the 1980s relative to humanities BAs, and in fact the reverse was true. Further, there were no above-average wage increases for the occupational groups most directly involved with the development and use of computers, namely, “engineers” and “math/computer”. During 1979–97 fully half of the growth in the college wage premium can be attributed to the increased relative wage of the group called “managers,” and only 17 percent to the computer-related occupational groups. The Autor et al. three-way distinction would place computer programmers and many types of engineers in the middle, rather than high category, as jobs subject to outsourcing and not benefiting from a rapid growth of demand relative to supply.
The SBTC hypothesis is about the demand for skilled workers growing faster than the supply. However, an increase in the college wage premium could reflect either increased growth of demand or slower growth of supply. Autor, Katz, and Kearney (2008) and Goldin and Katz (2007) emphasise a slowdown in the rate of growth of the relative supply of college workers from 3.89 percent per year from 1960 to 1989 to 2.27 percent per year from 1980 to 2005, due to college graduate rates reaching a plateau after decades of substantial growth.
Changes at the top end
While the demand and supply story of SBTC and Goldin and Katz can help explain the 90-10 ratio, increased skewness of incomes above the 90th percentile has been driven by a set of fundamentally different factors. To help understand the evolution of the highest incomes, we divide these workers into three categories. Superstars include the top members of any occupation that provides disproportionate rewards to the first-best as contrasted with the second-best. The superstar phenomenon has at its core the magnification of audiences, the fact that a single performance can be witnessed by an audience of one person or ten million people, depending on the perceived attraction and talent. The second category includes law partnerships, investment bankers, and hedge fund managers, where there is no obvious analogy to audience magnification but where there are steep wage premia for the very best in an occupational niche, and where it is apparent that incomes are highly market-driven.
The most contentious question regards the third category, top executives in public corporations. The core distinction is that CEO compensation is chosen by their peers in a system that gives CEOs and their hand-picked boards of directors, rather than the market, control over top incomes. The idea that managers, rather than stockholders, control directors goes back to Berle and Means (1932). This idea that the principal-agent control of stockholders should be reversed has been applied fruitfully by such authors as Bebchuk and Fried (2004). They argue that managerial power lies behind some of the outsized gains in CEO pay.
In general, we believe that better disclosure and better laws regarding corporate governance can help deal with high CEO pay. Precisely determining what counts as reasonable pay is beyond the government’s abilities. However, there is some evidence, reviewed by Dew-Becker (2008), showing that increases in mandatory disclosure lead to better corporate performance and better designed pay packages.
To summarise, we divide the income distribution into two sections: the bottom 90 percent and the top ten percent. The story of the bottom 90 is one of changes in demand due to SBTC and supply due to changes in the skill set of the workforce. The host of secondary factors influencing inequality, such as the minimum wage, union power, and trade, has generally pushed towards higher inequality, but the data have little ability to discriminate among these factors.
It may be too late to reverse the decline of unions, and reversing the growth of import penetration would require protectionist measures that are opposed by most economists. Raising the real minimum wage reminds us of the contentious literature in labour economics regarding the employment effects of the minimum wage. Better education, especially for the poor, might have a more direct impact on incomes at the bottom.
Within the top 10 percent, SBTC has certainly still been an issue, and there is a role of SBTC in contributing to pay premia of entertainment and sports superstars. In a variety of settings, technology has allowed superstars to distribute their talent to a wider variety of consumers. This has driven their incomes up exponentially. Their earnings are an outcome of market forces, and the only policy measure available to achieve greater after-tax equality is an increase in tax rates at the top balanced by a decrease at the bottom. However, for top corporate executives, there is strong evidence that incomes have been driven by non-market forces. This is where policy can have the most positive impact on inequality; increased disclosure and improved corporate governance laws can not only raise firm value but help distribute economic gains more evenly across society.
Autor, David H., Lawrence F. Katz and Melissa S. Kearney. 2008 “Trends in U. S. Wage Inequality: Re-assessing the Revisionists.” Review of Economics and Statistics, forthcoming. Also NBER Working Paper 11627.
Bebchuk, Lucian Ayre and Jesse M. Fried. 2004. Pay without Performance: The Unfulfilled Promise of Executive Compensation. Cambridge and London: Harvard University Press.
Berle, Adolph A. and Gardiner C. Means, 1932. The Modern Corporation and Private Property. New York and Chicago: Commerce Clearing House.
Bureau of Labour Statistics. 2006. “Characteristics of Minimum Wage Workers: 2005.” Obtained from http://www.bls.gov/cps/minwage2005.pdf.
Dew-Becker, Ian. 2008. “How Much Sunlight Does it Take to Disinfect a Boardroom? A Short History of Executive Compensation Regulation.” Mimeo.
Dew-Becker, Ian, and Robert J. Gordon. 2005. “Where Did the Productivity Growth Go? Inflation Dynamics and the Distribution of Income.” Brookings Papers on Economic Activity 36, no. 2: 67–127.
Goldin, Claudia, and Lawrence F. Katz. 2007. “Long-Run Changes in the Wage Structure: Narrowing, Widening, Polarizing,” Brookings Papers on Economic Activity 38, no. 2: 135-65.
Piketty, Thomas and Emmanuel Saez. 2003. “Income Inequality in the United States, 1913–1998.” Quarterly Journal of Economics 118, February: 1–39.