Editor’s note: Scott Winship, a fellow in economic studies at the Brookings Institution, sheds light on an eye-catching new report concerning the increase of the purchasing power of wage-earning U.S. adults.
Much of what I write is an attempt to dispel economic doomsaying on the left. It’s not that I think we have solved every problem of fairness, opportunity, or security; rather, certain problems are real while others are overstated.
Sometimes dramatically so. Last month David Cay Johnston, Pulitzer Prize winner and former New York Times reporter, wrote an essay for the (generally very informative) “taxanalysts” website on income inequality. Johnston tried to make the case that the gains of the rich have come at the expense of everyone else with the eye-raising claim that the income received by “the bottom 90 percent of earners” rose by only $59 in terms of today’s purchasing power between 1966 and 2011.
This is one of those litmus-test claims that in a perfect world would disqualify some people from debate over living standards, inequality, and the state of the middle class. To honestly believe that below the top 10 percent of earners there has been essentially no improvement in 45 years is to declare an extreme disconnection from the real world and a commitment to a negative interpretation of the American economy that is beyond parody.
The claim was dutifully repeated this week by the Center for American Progress’s ThinkProgress, the Huffington Post, and Salon. It has been tweeted, liked, or emailed by over 10,000 readers of those sites and in turn read by their own followers and friends. They will be more accepting of the next fearful economic claim as a consequence.
This is an egregiously erroneous “finding”, as I’ll explain in a moment. But first, take the test yourself. The median household income is the income of the household that is right in the middle of the distribution—half of all households are richer than it and half are poorer. How much do you think median household income has risen since 1966? (Hint: I’d take the “over” if we’re starting with $59.)
The Census Bureau’s Current Population Survey is the most widely-used source of income data, though it has some flaws. The first year for which household income data is readily available from the CPS is 1967. From 1967 to 1979, median household income grew by $5,500. This understates income growth during this period because it does not incorporate non-cash public transfers like food stamps, Medicaid, and Medicare and does not include fringe benefits or realized capital gains (such as from the sale of a home). From 1979 to 2009, we can use improved estimates from the Congressional Budget Office that combine the CPS data with tax return data to partially fill these gaps. CBO indicates that median household income (before taxes) rose by $14,200. After taxes, median income rose by $17,600.
At a minimum, therefore, median household income rose by $20,000 from 1967 to 2009. The data on which Johnston relies shows a decline of $200.
Despite his claim, the data Johnston uses represents tax returns, not “earners”. A married couple filing jointly is a single tax return. But even if we give Johnston the benefit of the doubt and ask what has happened to the median “earner”, he is way off. The readily available CPS data goes back to 1975—the increase in earnings for the median worker was $6,500 from 1975 to 2011, while the tax return data relied on by Johnston suggests a decline of $1,200.
How did Johnston get it so wrong? The most fundamental answer is that he failed to challenge his priors when the data showed an implausible result. But he is not making his numbers up. The problem is that the data he used cannot be used to examine trends in living standards among the non-rich.
Johnston relies on the tax-return-based data of economists Thomas Piketty and Emmanuel Saez—the data that is the basis for the ubiquitously reported (and true) finding that incomes at the top have sky-rocketed. As I have written before, the degree of income inequality we have in the U.S. is truly mind boggling, though that fact tells us nothing about whether inequality is problematic. Clearly it makes a difference if the bottom 90 percent is only $59 richer over 45 years rather than $20,000 richer. Piketty and Saez do show that the income received by the bottom 90 percent of “tax units” rose by just $59. We know because they admirably make mounds of their data available to the public (in this case, see the “Table_Incomegrowth” tab, column J).
Why are their numbers so different from the Census Bureau and CBO figures? The share of household and personal income that is included in the Piketty/Saez measure has declined markedly since 1966. In that year, the total income in the Piketty/Saez data amounts to 109% of the total income in the CPS. That the share is over 100% largely reflects the fact that the CPS does not capture the very richest households and caps the amount of income reported at artificially low levels for confidentiality purposes. By 1979, Piketty/Saez income was just 93% of CPS income, despite the fact the CPS continued to understate income at the top. The decline in share of income accounted for by Piketty/Saez is due to the fact that taxable public transfers are largely excluded from their measure, as are tax-favored employer benefits deducted from paychecks. Both sources of income have grown over time. Their estimates also exclude realized capital gains among the bottom 90 percent, and those were rising over time too.
The share of CPS income accounted for by Piketty/Saez income was largely unchanged in 2009. But this was the period in which income inequality rose. If the CPS could adequately capture top incomes, the increase in CPS income would have been greater than the data indicate, and so the share of CPS income accounted for by Piketty/Saez income would have fallen again. From 1979 to 2009, we can turn to the CBO income estimates, which incorporate public transfers, employer-provided health insurance, full pay (before deductions), capital gains, and top incomes. The share of CBO income accounted for by Piketty/Saez income fell from 89% in 1979 to 77% in 2009.
It is also the case that “tax units”—essentially tax returns, but with a small number of additional “units” added in to account for non-filers—include, for example, teenagers who have summer jobs and college kids on work study who file tax returns and make very little. Changes over time in the size of these groups could also affect the results—not only by pulling the incomes of the bottom 90% down but by pushing the entry point to the top 10% up (more low-income tax returns means the “top 10%” is a richer group). There may also be under-reporting of self-employment and other income in the tax return data, and if that has increased over time, it would also dampen the rise in income in the bottom 90%.
You might think that these problems call into question the basic Piketty/Saez claims about rising income inequality. This is a possibility I’ve considered (and continue to), but evidence from other sources supports the basic conclusion that the incomes of the very rich have skyrocketed. These sources are not without problems themselves (and most rely on the tax return data in some way), but many of the shortcomings of the Piketty/Saez income measure are absent from them.
The Johnston analysis, in the end, is just an extreme case of liberal negativism about living standards. As I discuss in my new essay for National Affairs, “Overstating the Costs of Inequality,” the median household in the U.S. is twice as rich as it was in 1960, at the peak of the supposed “Golden Age” the American middle class. Too many commentators refuse to see that and are all-too-ready to accept the most outlandish claims about living standards. As I’ve argued before, this sort of thing sows economic insecurity and works against continued improvement in our standard of living. And it detracts from real-world problems.