The Trouble with the Median
Readers write back
Perhaps surprisingly, the issue of citing the median as an indicator of the racial wealth gap has become a site of mild controversy at The Glenn Show. The issue first cropped up in my conversation with historian David Kaiser—you can find a transcript of that part of our talk here. David and I agreed that using the median to measure the racial wealth gap is misleading. If you looked only at that number, you wouldn’t necessarily know that most of the gap is accounted for by disparities among the wealthiest white and black Americans. While that gap is meaningful, it isn’t something that should worry us in the same way that gaps lower on the wealth distribution do.
Citing only the median in debates about the racial wealth gap, then, suggests that the wealth gap is more pervasive than it appears. This figure can be misused (intentionally or not) in all sorts of ways: to represent the aftereffects of redlining practices, to indicate the racially biased distribution of benefits, to symbolize an amorphous but pervasive “systemic racism,” and so on. (John McWhorter and I discussed some of this in a recent conversation.)
So how should we talk about racial disparities and wealth? How do we measure these disparities more accurately, and what should we do about them? Two readers wrote in with some suggestions. The first, Jeffrey Thompson, is an economist who works for the Boston Fed. He links to some useful recent research in the area and also cautions against ignoring the median entirely. The second is a TGS fan favorite: Clifton Roscoe. He weighs in with his typically rigorous and well-documented analysis of the situation.
If we want to have serious discussions about the shape of racial disparity in this country, we need detailed, nuanced, fact-driven conversations like this one. As always, I’m eager to hear more from you.
I just listened to your recent conversation with John McWhorter—specifically on the racial wealth gap—and wanted to suggest a few things to add to the discussion:
Most of the wealth data people are using and that you referenced exclude defined benefit pensions. These pensions are ignored by some because (in addition to not being included in the data) they are less common than in years past, but they continue to account for 15% of all household wealth. DB pensions are particularly important for Black family wealth, in part due to strong representation in public sector employment where these pensions remain very common. When you account for DB pensions in the data, racial wealth gaps, particularly the white/Black gaps are considerably smaller. I discuss this in a recent paper.
Once we acknowledge DB pension wealth, we see that pension wealth (defined benefit + defined contribution) is the largest component of wealth for families with heads in their fifties and sixties, making the employment-based wealth more important than housing wealth.
Also, Asian families are excluded from most wealth data, but my recent work is able to include them. Asian families are the highest wealth racial group. Why this is relevant, I think, is that the data are supportive of the idea that the path to wealth for typical Asian families is through academic achievement to educational attainment to high-paying jobs to wealth.
Education is an inadequate proxy for skills—there is massive heterogeneity in the earnings potential among college graduates by field of study, for example. STEM degree holders are going to earn much more over their careers than humanities and arts degrees, and we see differences by race in degree type that exacerbate the large differences in degree attainment. In the paper I presented at the early October racial disparities conference at the Boston Fed, I show that it is lifetime earnings and some related human capital variables that can account for most of the observed disparities in wealth between race groups. Inheritances don’t seem to matter that much. That paper is here.
I would also offer a quibble with your comment on the unimportance of the median.
Keep reading with a 7-day free trial