Consider the Evidence

Income inequality in the United States is typically measured with data from a survey that asks around 50,000 households what their income was in the previous year. According to these data, inequality has increased sharply since the 1970s (see the second chart here).

But this survey includes different households each year. It therefore misses any mobility — movement of households up and down in the distribution over time — that occurs. If mobility has increased, the conclusion that there is more inequality might be misleading. Even if the gap between the top and bottom increases over time, if households change places with greater frequency, the inequality of their average income — “true” inequality — may have stayed more or less the same. Rising mobility can offset rising single-point-in-time inequality.

The type of mobility at issue here is relative intragenerational income mobility. Has it increased in recent decades?

To find out, we need panel data — data for the same households (or individuals) over a number of years. There are three main sources of such data. Each suggests the same conclusion: relative intragenerational income mobility in the United States has not increased.

A standard way to assess mobility is to divide households into quintiles (five equally-sized groups) based on their income at the starting time point. Then we look at the share of each of these groups that moves up (or down) in the distribution between time 1 and time 2. More movement indicates more mobility.

One source of data is the Panel Study of Income Dynamics (PSID), a panel survey of nearly 8,000 households begun in 1969. The following chart shows the share in each of the bottom four quintiles that moved up over three successive decades beginning in 1969. (There’s no significance to the choice to show movement up; the graph could just as well show the share moving down. The point is whether the shares increase over time.) The shares were calculated by Katharine Bradbury and Jane Katz. (See also this earlier analysis by Peter Gottschalk and Sheldon Danziger.) There is no indication of an increase in mobility from the 1970s to the 1980s to the 1990s.

A second data source is income tax returns, which are analyzed in a U.S. Treasury Department report (see table A-5). The data are from a sample of returns filed by taxpayers age 25 or older in the initial year. Here too the period examined is roughly a decade. In this study there are two periods: 1987-96 and 1996-2005. The next chart shows the shares moving up in each of the two periods. Again the data do not indicate an increase in mobility.

A third data source is Social Security earnings records. These records are available since 1937. Wojciech Kopczuk, Emmanuel Saez, and Jae Song have used them to study changes in earnings mobility. They conclude that “short-term and long-term mobility among all workers has been quite stable since 1951.”

The fact that all three data sources suggest the same conclusion doesn’t necessarily mean it’s correct, but it offers good reason to favor that conclusion. Rising income and earnings inequality in the United States does not appear to have been offset by increased mobility.