Yesterday, Junlei Chen of Penn Wharton released an estimate of how the benefits from the proposed Biden student forgiveness plan would be spread out across the income distribution. The relevant conclusion is summed up in the graph below.
The description of Chen’s methodology is sparse, but she appears to be using the Survey of Consumer Finances (SCF) to produce this estimate. As far as I know, the SCF is the only public data source that allows one to produce an estimate like this and the SCF is mentioned in the report’s appendix.
But there is a problem with the SCF that makes it uniquely weak at the thing Chen is using it for. As I noted three years ago, the SCF uses a unique sampling unit called the Primary Economic Unit (PEU), which “consists of an economically dominant single individual or couple (married or living as partners) in a household and all other individuals in the household who are financially interdependent with that individual or couple.”
Student debtors that live with their parents or in some other arrangement where they are not the economically dominant individual or couple in the household wind up outside of any PEU, meaning they are not counted at all in the survey, or they wind up grouped in with another PEU, meaning that their debt is effectively assigned to the economically dominant individual or couple in the household. The practical upshot of this is that a lot of student debt especially held by lower income borrowers ends up uncounted or miscounted in the SCF.
Indeed, the Federal Reserve warned back in 2015 and 2014 that “a considerable portion of aggregate student loan debt is held by individuals outside of the economic core of a household” and that “SCF users should exercise caution when drawing inferences regarding balance sheet items that are primarily held by young people.”
When we compare the distribution of student debt in the SCF to the distribution of student debt in administrative data, we can see that the SCF understates the share held by the lowest income quintile by 40 percent. It also understates the share of debt held by the highest income quintile by 23 percent while overstating the share held by each of the middle three quintiles.
Given this tremendous disparity, it is generally unwise to do what Chen has done, especially without caveating it upfront. If we crudely adjust Chen’s numbers using the disparity between the SCF and administrative data, as I do in the graph below, we can see it makes a big difference.
But even these adjustments are probably not all that accurate. For one thing, Biden’s student debt forgiveness proposal provides no benefit at all to the richest 5 percent of households. If the student debt that is missing from the SCF’s top quintile is disproportionately held by the top 5 percent, which seems plausible, then the share of the benefit flowing to the top quintile is lower than suggested above and the share flowing to the other four quintiles is higher.
Chen is probably unaware of the unique problems with the SCF’s sampling unit as applied to this question. It’s an unusual approach not found in other government surveys like the Current Population Survey and the American Community Survey. But this lack of awareness has resulted in the publication of an estimate that is very clearly inaccurate.