Is the Census Wrong About Median Incomes Stagnating?

At Real Clear Markets, Brookings' Gary Burtless argues that the median income stagnation picked up by the Census' Current Population Survey might not actually be happening. In support of his argument, Burtless offers the following evidence:

  1. GDP growth, consumption growth, and unemployment decline are all trends that should coincide with middle income growth.
  2. Figures from the Commerce Department and the national accounts indicate that disposable and market income is up by around 2% per capita.
  3. Although (2) is consistent with a world where median incomes are stagnating because all income gains are going elsewhere (e.g. to the rich), average household income for the richest 5% of households actually went down in the Census.
  4. Money income per capita has gone down in the Census figures while going up in the Commerce Department data and national accounts.

These are interesting points worth considering and certainly it could be the case that the CPS is a bad survey producing bad income estimates. However, Burtless argument is not sufficient for reaching that conclusion.

All of Burtless' observations about varying data on this are compatible with the following two-point theory: 1) the CPS does not do a good job of capturing top incomes (which is known), and 2) most of the national income gains have gone to the very top. If this were true, then the CPS per-capita income estimates would fall behind estimates from other sources because the CPS is failing to record all the income gains going to the top. I am not saying this is definitely what is happening, but this is precisely the theory that you would come up with based on the known limitations of the CPS, and Burtless has done nothing to rebut it.

Burtless suggests (and I would tend to agree) that administrative sources of income data are much preferred over income surveys (where they exist and can be accessed). One such administrative data source are the wage statistics produced by the Social Security Administration each year based on the reported earnings subject to the federal income tax. This data tends to confirm the basic stagnation story, especially at the median:

Between 1990 and 2000, median personal earnings went up from $25,071 to $28,346, an increase of $3,275. Over the next 7 years, between 2000 and 2007, median personal earnings stayed basically flat, peaking at a bubble year high of $28,914 in 2007. From there, they dipped in the Great Recession and were still below their 1999 levels in 2013.

For mean personal earnings, the story is somewhat different. They too ran up significantly between 1990 ($34,881) and 2000 ($41,721). But they also continued to grow at a decent clip in the 2000s. In the 2007 bubble year, they were up to $43,546. They then tumbled during the recession. Even with the recession, mean personal earnings in 2013 were at $43,041, a gain of $2,165 since 1999.

Based on the SSA data, it does seem that median earnings have stagnated since around 2000 while mean earnings continued to grow, a finding also consistent with the theory that income gains have been absorbed mainly at the top.