Dan Herbst
@danherbst.bsky.social
360 followers 920 following 38 posts
Econ Prof at University of Arizona. Interested in labor, public, & household finance. Also a fan of dogs & motorcycles. www.danjherbst.com
Posts Media Videos Starter Packs
Pinned
danherbst.bsky.social
1/ New working paper with Trevor Bakker, @stefanie-deluca.bsky.social, Eric English, Jamie Fogel, and Nathan Hendren: We use newly linked Census, IRS, and credit bureau data to explore differences in credit access by race, class, and hometown.

djh1202.github.io/website/cre...

🧵...
Reposted by Dan Herbst
micreconinsights.bsky.social
New from Daniel Herbst (Assistant Professor of Economics at the University of Arizona) and Nathaniel Hendren (Harvard University):

'A better way to pay for college?'

There is a better way to fund undergraduate study, according to new research on the US.
Reposted by Dan Herbst
nber.org
NBER @nber.org · Jul 25
Constructing new population-level linked administrative data to study households' access to credit in the US, from Trevor J. Bakker, Stefanie DeLuca, Eric A. English, James S. Fogel, Nathaniel Hendren, and Daniel Herbst https://www.nber.org/papers/w34053
danherbst.bsky.social
In case you missed it, @wsj.com dropped a bombshell this week.

I’m speaking, of course, about the recent feature on our working paper about credit access:

www.wsj.com/economy/cred...
danherbst.bsky.social
14/ In sum, our paper documents large differences in credit access by race, class, and hometown. These differences appear driven not by algorithmic bias or resource constraints but by early-life repayment that's rooted in childhood environments.
danherbst.bsky.social
13/ Why do childhood environments matter so much? Survey evidence points to several mechanisms, including differences in social capital, financial literacy, and informal credit networks. Future research might explore how early credit experiences are shaped through these channels.
danherbst.bsky.social
12/ We also find causal evidence place-based effects on repayment: using a mover's design, we find that childhood exposure to places with better repayment behaviors leads to lower delinquency rates in adulthood, even conditional on adult income.
danherbst.bsky.social
11/ Instead, we find that childhood factors play a role--the credit scores of your parents and childhood neighbors predict your later-life repayment, even conditional on income, wealth, and education.
danherbst.bsky.social
10/ Do differences in financial resources explain repayment gaps by race, class, and hometown? Not by much. Repayment gaps persist among individuals with the same income, wealth, marital status, job stability, and employer.
danherbst.bsky.social
9/ Since credit scores predict future delinquency, the key to removing either type of bias and expanding credit access is to understand why some groups fall delinquent more often than others, especially in early adulthood.
danherbst.bsky.social
8/ But when it comes to *balance* bias, the answer is yes: among those who end up avoiding delinquency, Black individuals and those from low-income backgrounds received lower initial scores than other groups.
danherbst.bsky.social
7/ When it comes to *calibration* bias, the answer is no: Among those with a given credit score, Black individuals or those from low-income backgrounds actually fall delinquent at higher rates than other groups. In other words, the credit-score gap understates the repayment gap.
danherbst.bsky.social
6/ Are credit scores simply biased against Black individuals or those from low-income backgrounds? The answer depends on how you define "bias."
danherbst.bsky.social
5/ These differences emerge at early ages and reflect real constraints on borrowing: groups with lower scores also have smaller credit balances, higher credit card utilization, and rely more on high-cost alternative financial products like payday loans.
danherbst.bsky.social
4/ We also see large differences by hometown. For example, by the time they're adults, low-income White children from Brooklyn, NY have 90-point higher average credit scores than low-income White children from Indianapolis.
danherbst.bsky.social
3/ Black individuals and those born to low-income parents have much lower credit scores than other groups. For example, Black individuals from the 90th pctile of parent income have similar average credit scores as White individuals from the 25th pctile.
danherbst.bsky.social
2/ Whether buying a home, building wealth, or managing sudden expenses, affordable credit can be a powerful tool for upward mobility. But our data show how access to this tool is limited for many groups.

Consider credit scores—a metric lenders use to judge creditworthiness...
danherbst.bsky.social
1/ New working paper with Trevor Bakker, @stefanie-deluca.bsky.social, Eric English, Jamie Fogel, and Nathan Hendren: We use newly linked Census, IRS, and credit bureau data to explore differences in credit access by race, class, and hometown.

djh1202.github.io/website/cre...

🧵...
danherbst.bsky.social
Thanks!

And thanks for the thoughtful feedback! The comments are much appreciated :)
danherbst.bsky.social
Yeah it would add cost and complicate recruitment (which mentions the piece-rate option). At the time, I didn't think an hourly conrol group would be worth the trouble because I thought the highest wage offers would get close to complete take-up. In hindsight it might have been a good idea!
danherbst.bsky.social
You're absolutely right--there are a variety of factors apart from moral hazard and adverse selection that influence payment schemes. I mention monitoring costs, but I should delve into behavioral mechanisms like Falk & Kosfeld as well. There's a lot of great work I'm building on here.
danherbst.bsky.social
I'd love to add other dimensions of variation, but unfortunately cost is a very real constraint right now. I'm definitely looking to things like this in the future if those constraints are lifted!
danherbst.bsky.social
Yeah, I tried to make it as generalizable as possible—participants don't know they're in an experiment, data-entry required in many jobs, etc. But I can't claim my estimates would extrapolate different tasks or labor markets (in fairness, neither can most other studies of labor productivity.)