Incredible Shrinking Income Inequality

Its rise is an illusion created by the Census Bureau’s failure to account for taxes and welfare.

By Phil Gramm and John Early

March 23, 2021 12:34 pm ET

The refrain is all too familiar: Widening income inequality is a fatal flaw in capitalism and an “existential” threat to democracy. From 1967 to 2017, income inequality in the U.S. spiked 21.4%, and everyone from U.S. senators to the pope says it’s an urgent problem. Yet the data upon which claims about income inequality are based are profoundly flawed.

We have shown on these pages that Census Bureau income data fail to count two-thirds of all government transfer payments—including Medicare, Medicaid, food stamps and some 100 other government transfer payments—as income to the recipients. Furthermore, census data fail to count taxes paid as income lost to the taxpayer. When official government data are used to correct these deficiencies—when income is defined the way people actually define it—“income inequality” is reduced dramatically.

We can now show that if you count all government transfers (minus administrative costs) as income to the recipient household, reduce household income by taxes paid, and correct for two major discontinuities in the time-series data on income inequality that were caused solely by changes in Census Bureau data-collection methods, the claim that income inequality is growing on a secular basis collapses. Not only is income inequality in America not growing, it is lower today than it was 50 years ago.

While the disparity in earned income has become more pronounced in the past 50 years, the actual inflation-adjusted income received by the bottom quintile, counting the value of all transfer payments received net of taxes paid, has risen by 300%. The top quintile has seen its after-tax income rise by only 213%. As government transfer payments to low-income households exploded, their labor-force participation collapsed and the percentage of income in the bottom quintile coming from government payments rose above 90%.

In 2017, federal, state and local governments redistributed $2.8 trillion, or 22% of the nation’s earned household income. More than two-thirds of those transfer payments went to households in the bottom two income quintiles. Remarkably the Census Bureau chooses to count only $900 billion of that $2.8 trillion as income for the recipients. Excluded from the measurement of household income is some $1.9 trillion of government transfers. These include the earned-income tax credit, whose beneficiaries get a check from the Treasury; food stamps, which let beneficiaries buy food with government issued debit cards; and numerous other programs in which government pays for the benefits directly.

Americans pay $4.4 trillion a year in federal, state and local taxes. Households in the top two earned-income quintiles pay 82% of the tax bill, although they never see most of this money because it is deducted directly from their paychecks. When measuring income inequality, however, the Census Bureau doesn’t reduce household income by the amount paid in taxes. Had it done so and counted all transfer payments as income, inequality from 1967 to 2017 would have increased by only 2.3% instead of the reported 21.4%. That’s a difference of almost 90%—a rather large error.

Twice over the past 50 years, the Census Bureau has significantly changed how it collects and records income statistics. In 1993 and 2013 the Census Bureau changed its methods in an effort to collect better information from high-income households. These changes created two major discontinuities and distorted the time-series so that the change in measured income inequality in those years was as much as 15 times the average annual change found for the entire 50-year period. At the time, the Census Bureau explained in detail what it had done. It also explained the limitations the changes imposed on the use of its income-inequality measure to look at changes over extended periods. In subsequent use of the data by the Census Bureau and others, however, those warnings have been neglected.

The simple solution would have been to isolate the distortions caused solely by the changes in data-collection techniques and adjusted the previous years’ measures to reflect the effect of the changes. We made these adjustments and they are shown in the nearby figure. The blue line is the actual reported Census Bureau measurement of income inequality. The yellow line eliminates the effects of the 1993 and 2013 discontinuities caused solely by changes in measurement technique. The black line shows income inequality when the value of all transfer payments received is counted as income, income is reduced by taxes paid, and the two technical corrections are made.

Lo and behold—income inequality is lower than it was 50 years ago.

The raging debate over income inequality in America calls to mind the old Will Rogers adage: “It ain’t what you don’t know that gets you into trouble. It is what you do know that ain’t so.” We are debating the alleged injustice of a supposedly growing social problem when—for all the reasons outlined above—that problem isn’t growing, it’s shrinking. Those who want to transform the greatest economic system in the history of the world ought to get their facts straight first.

Mr. Gramm is a former chairman of the Senate Banking Committee and a visiting scholar at the American Enterprise Institute. Mr. Early served twice as assistant commissioner at the Bureau of Labor Statistics.