The Meritocracy Trap

by Daniel Markovits

  The data set, which is based on survey responses, collected 340,000 observations in 1950 and has expanded, after 1970, to include between two and eleven million valid observations per year. See IPUMS-USA Table 2, “Valid Income Observations.” For each person, the data set reports both income and, for persons who worked in the previous year, the average number of hours usually worked per week. (For 1940, 1950, 1980, and 1990, the data reports “Hours worked last week”—HRSWORK1—and beginning in 1980 the data reports “Usual hours worked per week”—UHRSWORK. The data also include an income variable, INCWAGE, which includes wages, salaries, commissions, cash bonuses, tips, and other money income received from an employer but does not include payments in kind or reimbursements for business expenses. See IPUMS USA, “INCWAGE: Wage and Salary Income: Description,” https://usa.ipums.org/usa-action/variables/INCWAGE#description_section. The income variable is top-coded, but in a way that makes it possible to identify the very top earners, by coding U.S. state means above the top-coding threshold until 2002 and coding at the 99.5th percentile for each state from 2003 onward. See IPUMS USA, “INCWAGE: Wage and Salary Income: Codes,” https://usa.ipums.org/usa-action/variables/INCWAGE#codes_section. The top 1 percent in the figure may thus depart from the true top 1 percent, but not by much and in either direction.

  The data make it possible to depict the evolving relationship between the income and usual hours, as in the figure in the text. The patchwork character of the data behind the figure counsels against reading a false precision into the levels that it reports. But the trends are robust enough to remain persuasive even if necessarily imprecise. Moreover, the figure’s basic lesson—of a rising time divide between elite and non-elite work hours—is confirmed by any number of other studies, which gather data on work hours using a wide variety of methods, ranging from surveys like those that underlie the figure to time diaries to buzzers that are worn by subjects and solicit contemporaneous responses stating whether or not the subject is working at random times throughout the day.

  Income Poverty, Consumption Poverty, and the Income Share of the Top 1 Percent: The data behind the figure may be found in the following sources: The World Wealth and Income Database, Top 1% Income Share—Including Capital Gains; U.S. Census Bruse, “Historical Poverty Tables: People and Families—1959 to 2017,” Current Population Survey, Table 2, last modified August 28, 2018, www.census.gov/data/tables/time-series/demo/income-poverty/historical-poverty-people.html; Bruce Meyer and James X. Sullivan, “Winning the War: Poverty from the Great Society to the Great Recession,” Brookings Papers on Economic Activity (Fall 2012): 133–200, Table 1.

  It would be instructive to construct a series for the top 1 percent’s consumption share, but existing data do not allow this. The Consumer Expenditure Survey tracks expenditure shares by quintiles of pretax income, and (beginning more recently) by deciles of pretax income. See, e.g., Bureau of Labor Statistics, Consumer Expenditure Survey (2015), Table 1101, www.bls.gov/cex/2015/combined/quintile.pdf, and Bureau of Labor Statistics, Consumer Expenditure Survey (2015), Table 1110, www.bls.gov/cex/2015/combined/decile.pdf. The survey also currently tracks consumption by income buckets that range from “less than $15,000” to “$200,000 or more” (which represents roughly the top 5 percent in 2015). See Bureau of Labor Statistics, Consumer Expenditure Survey (2015), Table 1203, www.bls.gov/cex/2015/combined/income.pdf. But decile tracking began only recently and the income buckets used by the survey have changed over time, so that no good time series for top/bottom ratios can be constructed using these categories. Moreover, the survey still does not track consumption in still narrower economic elites. Time trends in consumption by quintile are summarized over 1984–2010 by Kevin A. Hassett and Aparna Mathur, “A New Measure of Consumption Inequality,” American Enterprise Institute Economic Studies Series, June 25, 2012, 5 and Figure 1, www.aei.org/publication/a-new-measure-of-consumption-inequality/. Hassett and Mathur find only a modest increase in the top/bottom quintile consumption ratios over the period of their study.

  Ratios of Representative High, Middle, and Low Incomes over Time: Data from the World Top Incomes Database, Post-tax national income / equal-split adults / Average / Adults / constant 2015 local currency, https://wid.world/country/usa/.

  key points in the overall income distribution: The figure uses post-tax-and-transfer rather than market incomes in order to avoid repeating the errors made in computing the official poverty statistics. The true conditions of the rich, middle class, and poor in the United States today reflect the circumstances that they each enjoy after the state—with both its taxes and its social welfare programs—has intervened in their lives.

  U.S. Top-End, Bottom-End, and Full Gini Coefficients over Time: Data from the World Top Incomes Database, Post-tax national income / equal-split adults / Average / Adults / constant 2015 local currency, https://wid.world/country/usa/.

  calculated in three ways: The figures again calculate the Gini coefficients using post-tax-and-transfer incomes, in order to capture the true circumstances of the various segments of the economy that the coefficients describe.

  bottom seven-tenths of the U.S. income distribution: Some studies go even further and question whether there has been any steady or even significant rise in economic inequality across the bottom 99 percent of the distribution. For a review, see Robert J. Gordon, “Misperceptions About the Magnitude and Timing of Changes in American Income Inequality,” NBER Working Paper No. 15351 (September 2009), www.nber.org/papers/w15351.

  Ratios of Education Expenditures by Income and Education: The data used to construct the ratios of education expenditures present two complexities. First, the most comprehensive data on education expenditures, produced by the Bureau of Labor Statistics, distinguish incomes by quintiles only rather than by any finer slices. So the rich are thus represented, in the education expenditure series, by the top quintile rather than by the average of the top 1 percent (as they are in Figure 1.3). Second, the BLS data independently measure education expenditures of many households composed of students and allocate these to the bottom quintile by income, which introduces a dramatic and misleading upward distortion into the education expenditure series for the poor. The figure therefore takes the second rather than the bottom quintile as its representative of education expenditures by the poor.

  Both these decisions are validated by the third, shorter data series also represented in the figure. This series reports the ratios of elite/median and median/bottom expenditures on education, measuring eliteness not by income but rather by the highest degree held by the most educated household member. The BLS’s education categories separate out both households without anyone who holds even a high school degree and households with members who hold post-BA degrees. These are truer measures of low and high socioeconomic status than the income categories that the BLS makes available. The series constructed in this way are shorter than the others (the BLS data do not go as far back in time). But they align nicely with the ratios reported using the longer if less precise income series.

  90/50 and 50/10 Income Achievement Gaps for Reading and Math: See Reardon, “The Widening Academic Achievement Gap,” 102, 103, Figures 5.7, 5.8. See also Reardon, “No Rich Child Left Behind.”

  Reardon observes that the timing of divergence in the achievement and income gaps does not quite match up, although efforts to make the association are complicated by the fact that a family’s annual and lifetime incomes do not necessarily move together. Setting these complications aside, it remains true that a dollar of income appears to buy more academic achievement today than it did in past decades (100–104).

  constructed by the sociologist Sean Reardon: Reardon aggregates the results of many achievement tests, administered in overlapping series over many years, and this aggregation requires him to make adjustments that render individual test results comparable and to fit a line over many data points. Studies are rendered comparable in spite of measuring achievement on
different scales by adjusting scores for the reliability of the tests and then expressing test score gaps in terms of standard deviations. This is, as Reardon says, “standard practice when comparing achievement gaps measured with different tests (see, for example, Clotfelter, Ladd, and Vigdor 2006; Fryer and Levitt 2004, 2006; Grissmer, Flanagan, and Williamson 1998; Hedges and Nowell 1999; Neal 2006; Phillips et al. 1998; Reardon and Galindo 2009). So long as the true variance of achievement remains constant over time, this allows valid comparisons in the size of the gaps across different studies using different tests.” See Reardon, “The Widening Academic Achievement Gap,” 94. The basic result that 90/50 achievement gaps have been rising even as 50/10 gaps have held roughly steady and in some cases even fallen reappears across a range of estimation techniques. See Charles A. Nelson and Margaret A. Sheridan, “Online Appendices and Glossary” to “Lessons from Neuroscience Research for Understanding Causal Links Between Family and Neighborhood Characteristics and Educational Outcomes,” in Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances, ed. Richard Murnane and Greg Duncan (New York: Russell Sage Foundation, 2011), section 5.A2, www.russellsage.org/sites/default/files/duncan_murnane_online_appendix.pdf.

  The precise contours in the figure therefore reflect Reardon’s considered judgments about the data and should be read to illuminate trends rather than for actuarial precision. The basic trends that the figure reports are robust.

  GDP Share, Employment Share, and Relative Income and Education for Finance, 1947–2005: The figure is inspired by Philippon and Reshef, “Wages and Human Capital,” 1558, Figure 1, and 1561, Figure 2. Finance includes insurance but excludes real estate. GDP share is computed as the ratio of nominal value added by the finance sector to the nominal GDP of the United States. Data from Annual Industrial Accounts, Bureau of Economic Analysis. Relative education is computed as the share of hours worked by employees with at least a college degree in the financial sector minus the corresponding share of hours in the rest of the private sector. Data from March CPS.

  relative income and education: Relative income is the fraction of annual income per financial employee in excess of annual income per nonfinancial employee. Relative education is the difference between the fraction of financial employees with college degrees and the fraction of non-financial employees with college degrees. The relative education series is plotted as a linear transformation of the underlying values to allow for visual comparison of the series. While the correlation between the two series is unaffected by this transformation, their similar levels are an artifact of scaling.

  their private-sector counterparts: See Philippon and Reshef, “Skill Biased Financial Development,” 8. These percentages are derived by calculating the share of work hours provided by college-educated workers in each sector.

  began gently to decline: See Philippon and Reshef, “Skill Biased Financial Development,” 5. Philippon and Reshef point out that the changes that they document are driven by a rebalancing of the financial sector’s various subsectors, so that traditional banking has declined relative to other aspects of finance and in particular investment (p. 6). For another view, see Thomas I. Palley, “Financialization: What It Is and Why It Matters,” Levy Economics Institute Working Paper no. 525, December 2007 (using data from the Economic Report of the President [from 2007]).

  For finance’s share of employee compensation, see David A. Zalewski and Charles J. Whalen, “Financialization and Economic Inequality,” Journal of Economic Issues 44, no. 3 (2010): 757–77, reporting on Philippon and Reshef, “Skill Biased Financial Development.”

  increasing relative education and relative income: See Philippon and Reshef, “Skill Biased Financial Development,” 8. These percentages are derived by calculating the share of work hours provided by college-educated workers in each sector.

  fewer increasingly elite workers: To be sure, deregulation changed the structure of finance just as elite finance-sector wages began to rise. Indeed, deregulation enabled the creation and adoption of the financial techniques that make super-skilled financial labor so highly paid today. But it is a mistake to leap from these uncontested facts to the conclusion that deregulated elite finance workers’ immense incomes arise from exploitation or rent seeking. If the techniques employed by deregulated finance make super-skilled finance workers especially productive, then no increase in rent seeking is required to explain the increase in finance-sector wages. The figures, which show that relatively fewer, relatively more-skilled workers are taking the same cut from a relatively rising share of GDP explain rising finance-sector wages in this way, without needing to resort to rising rent seeking.

  None of this demonstrates, nor does it even assert, that finance workers extract no rents. They surely do, and the rents may even have increased in recent decades. But the greater part of rising finance-sector incomes neither requires nor in fact involves increased rent seeking.

  Philippon and Reshef, who have made the most careful study of finance sector rents, conclude that nonmeritocratic causes contributed little to rising finance-sector incomes from the 1970s through the early 1990s and that, since the 1990s, between 20 and 30 percent of finance’s risk-adjusted wages stems from sources besides skill. See Philippon and Reshef, “Wages and Human Capital,” 1553, 1603, 1605. The report of a recent and substantial increase in finance-sector rents rightly received attention; but the study is most notable for the converse of this result, namely that even in recent decades, between 70 and 80 percent of the finance sector’s rising wages stems from its workers’ rising skill. Probably this underestimates the true share, as the study measures finance workers’ skill by counting their number of years at school, and this single-minded focus on quantity neglects distinctive increases in the quality and intensiveness of elite finance workers’ educations. Elite finance workers increasingly and now overwhelmingly graduate from the very most august and intensive colleges and universities, and these schools spend much, much more per student per year than their competitors today or than they themselves used to do.

  Percent Changes in Employment Shares for Routine and Fluid Skills: The figure comes from Jaimovich and Siu, “Job Polarization and Jobless Recoveries.”

  Earnings Segmentation by Education Level: The data for the figure come from Carnevale, Rose, and Cheah, “The College Payoff.”

  the median worker from the top twentieth: Slightly under half of the U.S. population over twenty-five has no education at all beyond high school (roughly 70 percent do not have a BA), slightly over 10 percent hold a post-BA degree, and a little under 5 percent hold a doctorate or a professional degree. See Camille L. Ryan and Kurt Bauman, Educational Attainment in the United States: 2015, U.S. Census Bureau, Current Population Reports no. P20-578 (March 2016), accessed November 19, 2018, www.census.gov/content/dam/Census/library/publications/2016/demo/p20-578.pdf. See also Carnevale, Rose, and Cheah, “The College Payoff,” 6, and Sandy Baum and Patricia Steele, “Who Goes to Graduate School and Who Succeeds?,” the Urban Institute (January 2017), accessed April 9, 2019, www.urban.org/sites/default/files/publication/86981/who_goes_to_graduate_school_and_who_succeeds_1.pdf.

  Incomes of the Bottom 90 Percent and Per Capita Consumption and Debt over Time: The figure is inspired by Robert Hockett and Daniel Dillon, “Income Inequality and Market Fragility,” Figure 8. Hockett and Dillon in fact find an even steeper increase in household debt. The data behind the figure come from: Federal Reserve Board, Flow of Funds—Households and Nonprofit Organizations, Total Liabilities; Bureau of Economic Analysis, Personal Consumption Expenditures (PCE) and PCE Price Index; The World Top Incomes Database, Bottom 90% Average Income Including Capital Gains; U.S. Census Bureau Population Estimates.

  but rather through debt: The debt invites a hidden form of redistribution that lurks in the background of American policy as a kind of safety net below the social safety net of the conventional welfare state. Personal bankruptcy protection for inso
lvent debtors—the middle-class version of the Greenspan put—amounts to an implicit tax on all lenders and borrowers used to establish a social safety net for those borrowers who cannot sustain their consumption, even through ready credit. And bankruptcies have famously skyrocketed in recent years, effectively increasing the implicit tax rate, although tightening bankruptcy laws increasingly withdraw even the failsafe net and reduced even this highly attenuated form of economic redistribution. For more on bankruptcy and its connections to economic inequality and the welfare state, see, e.g., Karen Dynan, “The Income Rollercoaster: Rising Income Volatility and Its Implications,” Pathways (Spring 2010): 3–6.

  as income fell short: Others who have made similar observations include: Drennan, Income Inequality, 62 (“Money taken out from appreciating housing was not used to pay down debt because indebtedness rose. Rather, it was used to support consumption in the face of stagnant income.”); Edward Wolff, “Recent Trends in Household Wealth in the United States: Rising Debt and the Middle-Class Squeeze—An Update to 2007,” Levy Economics Institute, Working Paper No. 589 (“Where did the borrowing go? . . . Middle class households experiencing stagnating incomes, expanded their debt almost exclusively in order to finance consumption expenditures.”); Hockett and Dillon, “Income Inequality and Market Fragility” (“As the wealthy amass more of the aggregate income, the average household ramps up its borrowing to maintain accustomed living standards.”). See also Atif Mian and Amir Sufi, “The Consequences of Mortgage Credit Expansion: Evidence from the U.S. Mortgage Default Crisis,” Quarterly Journal of Economics 24, no. 4 (November 2009): 1449–96; Atif Mian and Amir Sufi, “House Prices, Home Equity–Based Borrowing, and the United States Household Leverage Crisis,” NBER Working Paper No. 15283 (2009); and Atif Mian and Amir Sufi, “Household Leverage and the Recession of 2007 to 2009,” NBER Working Paper No. 15896 (2010).