ProPublica is running a series of articles based on federal tax returns for 25 wealthy individuals which were illegally obtained. The articles present selective information and dubious metrics to push a narrative that rich people pay little tax. Data from four authoritative sources show that tax payments by these 25 individuals are not representative of high earners generally.

Let’s focus on ProPublica’s click-friendly headline: “You May Be Paying a Higher Tax Rate Than a Billionaire.” The article says that a “typical” worker with $45,000 in wages pays a higher federal tax rate than the average of the 25 wealthy people on the stolen tax returns. Including income and payroll taxes, ProPublica says that the typical taxpayer pays 19 percent while the 25 wealthy people pay just 16 percent.

The claim that people in the middle pay a higher tax rate than people at the top is at odds with data from four authoritative sources: the Congressional Budget Office (CBO), the Tax Policy Center (TPC), the Joint Committee on Taxation (JCT), and the Internal Revenue Service (IRS).

The table shows that the ProPublica tax rates are far different than the average effective tax rates for middle and top households published by the CBO, TPC, JCT, and IRS. The rates from these four sources are similar and consistent, while the ProPublica rates are outliers. The total rates in the table are the sum of income and payroll tax rates.

s

ProPublica’s top group is measured by wealth not income, and some people with top wealth may not have top incomes. However, ProPublica says that the income (AGI) on the 25 stolen tax returns was $86 billion over five years, or an average of $688 million for each return annually. That would put most of these returns in the top 0.001 percent of income since the IRS says that the AGI floor for that top group in 2018 was $69 million.

In the table, the ProPublica rates are averages for 2014 to 2018, the CBO rates are for 2017, and the rates for the other sources are for 2018. For CBO and TPC, the 40–60% group is the middle 20 percent of households. For JCT, the 50–90% group is the 50th to 90th percentiles, thus upper middle-income households. For the IRS, the 50–60% group is the 50th to 60th percentiles, thus households with just above average incomes.

While ProPublica says that the “typical” taxpayer pays a higher total income and payroll tax rate than the top 25 group, the CBO, TPC, and JCT data show that households in the middle pay tax rates less than half the rates of the top groups.

Looking just at income taxes, ProPublica has a higher rate for their typical taxpayer than the other sources show for the middle groups. ProPublica’s “typical” person is a single worker not receiving child tax credits or earned income tax credits. Also, the person is working and paying payroll taxes, not retired or receiving government benefits.

ProPublica shows a tax rate for the top 25 group much lower than the rates on the top groups from the four authoritative data sources. ProPublica says that its 25 wealthy taxpayers paid an average income tax rate of 15.8 percent, but the comparable IRS top 0.001 percent group paid an average 22.9 percent. Thus, the 25 wealthy persons with stolen tax returns appear to have very atypical tax situations compared to broader groups of top earners.

ProPublica says that its analysis “demolishes the cornerstone myth of the American tax system: that everyone pays their fair share and the richest Americans pay the most.” That broad claim is not supported by their narrow analysis. Indeed, it is a silly statement because it is based on only 25 apparently atypical high-end returns and a single hypothetical middle-income taxpayer.

Data note: For its typical taxpayer, ProPublica assumes $45,000 in wages, $6,896 in health benefits, and apparently $3,443 in employer-side payroll taxes, for a total gross income of $55,339. The taxpayer pays $3,800 in income taxes and $6,885 in payroll taxes. The tax rates in the table for this worker are based on the gross income.

A further analysis of tax rates by income is here.