Two shoppers make change for a purchase in New York.

How much has income inequality risen in the United States? Well, how do you define income? And what time period are you looking at? The most well-known statistics about income inequality—including the famous data from economists Thomas Piketty at the Paris School of Economics and Emmanuel Saez at the University of California, Berkeley—are based on tax data and show a significant increase in inequality since the late 1970s. But are the trends revealed in those data over that time period still holding up? Digging into the data reveals that inequality trends since 2000 actually are different from those during the 1980s and 1990s.

A paper released last week as a National Bureau of Economics Research paper takes a look at how income inequality has changed in the 21st century. The two authors, Fatih Guvenen of the University of Minnesota and Greg Kaplan of the University of Chicago, take a look at two different sources of income data. The first is administrative tax data that cover all forms of income, and the second source is data from the Social Security Administration based on W-2 forms, which only cover labor income. In other words, the tax data capture both labor and capital income, while the SSA data pull in labor income alone.

The two datasets show broadly the same trend in top-end income inequality from the early 1980s to the late 1990s. The one exception is the period of 1986 to 1988, when the tax data show a jump in inequality, which Guvenen and Kaplan point out is probably due to tax reform in 1986. But for the most part, the data agree that inequality rose from the late 1970s to the beginning of this century.

But around the year 2000, the data show something different. The tax data continue to show an increase in income inequality, but the SSA data indicate a negligible or no increase at all. The authors explain that they first examined whether the two data series were showing different trends because they used different units of measurement (tax filers for the tax data and individuals for the SSA data). But that doesn’t seem to be a big contributor. Rather, the difference seems to be the kind of income the data are capturing.

Income inequality since 2000 is not a story of rising inequality of labor income, but an increase due to higher capital income. Specifically, the rise is due to increased business income from S-corporations. The rise of pass-through businesses such as partnerships has been the primary source of rising inequality in the 21st century. Guvenen and Kaplan also show that the increase in income inequality hasn’t been primarily a story of the top 1 percent of income earners, but rather the top 0.01 percent. This makes sense, as ownership of pass-through entities is far more concentrated than the high labor incomes that fueled the increases in inequality in the 1980s and 1990s.

The increasing importance of capital income over labor income has been noticeable in the Piketty and Saez data, as J.W. Mason of John Jay College pointed out two years ago. The timing is also interesting given that 2000 is when the labor share of income in the United States started to decline.

As the authors of this new paper argue, we need to be careful with describing the trends and facts about inequality. The data about the rise during the two decades prior to the turn of the century, which shows a leading role for high labor incomes, led to diagnoses and solutions focused on “supermanagers” and other kinds of “superstar” workers in the top 1 percent. That line of thinking doesn’t make sense when rising inequality is instead a story of rising capital income among business owners in the top one-hundredth of the top 1 percent. To paraphrase a quote often attributed to John Maynard Keynes, when the facts on inequality change, we should change our thinking about it.