Impute missing student loan debt from other sources

[MaxGhenis]

The 2019 SCF estimates student debt to be $1.1 trillion, while FRBNY estimated $1.54 trillion as of 2020Q2.

Matt Bruenig showed in 2019 that the missing student debt is disproportionately among young people, attributing the discrepancy to the construction of SCF household units and young student debtors living with parents.

A simple way to impute the missing debt would be to multiply each household's student debt by an age-varying factor. But Bruenig suggests that the problem is on the extensive rather than intensive margin. Addressing this would be more complicated and we'd probably need to calibrate to some other source for the total number of people or households holding student debt.

cc @ngpsu22

[rickecon]

@MaxGhenis sent me this Matt Bruenig recent article on U.S. student debt and the differences between the three main sources of data.

1. Federal Reserve G.19 is a monthly report that collects data from institutions that own student debt, primarily the federal government and some private lenders.
2. The New York Fed's Consumer Credit Panel (CCP) collects information from credit reporting agencies.
3. The Survey of Consumer Finances (SCF) at the Federal Reserve Board of Governors is a representative survey (not a panel, and small sample size).

Maybe one could use the other two data sources to correct for biases and missing data in the SCF.

[rickecon]

@MattHJensen mentioned a crosswalk on the NBER website for using the SCF data with the NBER TAXCALC model. I think this page has some of that information, which would be helpful in adding SCF variables to the Tax-Calculator input data. @MaxGhenis

[MattHJensen]

Another page, too: https://users.nber.org/~taxsim/to-taxsim/scf/