Persistent Racial Disparities in U.S. Mortgage Approval: Evidence from 42 Million Applications, 2020-2024
The public-data Black-White mortgage approval gap at national scale: how large, where it lives, and how it responds to institutional boundaries.
The discovery in one figure
Swipe to see the whole figure
The paper in five minutes
Every U.S. mortgage application leaves a public record. Reading all 42 million of them from 2020 to 2024, Black applicants are approved about 15 percentage points less often than White applicants. This paper measures that gap carefully: most of it (68%) cannot be explained by anything visible in the public data, and three quarters of it lives inside individual lenders rather than between them. Two natural experiments, an insurance-driven threshold at 80% loan-to-value and the Federal Reserve’s 2022 tightening, each widen the gap further. The paper never claims to prove intent; it maps the scale and the structure.
The research question
How large is the observable racial approval gap in U.S. mortgage lending, how much sits within lenders, and what do quasi-experimental boundaries reveal about its structure?
How it works
42,323,519 HMDA applications, 2020-2024. DFL reweighting, within-lender fixed effects, regression discontinuity at the 80% LTV / PMI boundary, difference-in-differences around the 2022 tightening with HonestDiD sensitivity, Manski partial-identification bounds, and permutation tests.
- 01Assemble42,323,519 applications, 2020-2024
- 02ReweightDiNardo-Fortin-Lemieux: compare statistically similar applicants
- 03Look within lendersfixed effects: between institutions, or inside them?
- 04Natural experimentsRDD at the 80% LTV boundary · DiD around the 2022 tightening
- 05Bound the unknownpartial identification calibrated to consumer-finance data
Experimental results
Raw gap 14.95 pp; 68% unexplained by public observables; 74.6% of the gap is within-lender, rising from 66.8% in 2020 to 78.3% in 2024. The approval differential rises by about 2.0 pp above the 80% LTV threshold in purchase loans and widens by about 1.5 pp within lenders after the 2022 tightening. Bounds keep at least 44% to 55% unexplained under conservative assumptions; scale is about 126,000 fewer approvals for Black applicants annually.
14.95 pp16.1 pp9.3 pp74.6% of the national gap sits within individual lenders; scale is about 126,000 fewer approvals a year.
- 14.95 ppraw Black-White approval gap
- 74.6%of the gap within lenders
- ≈126,000fewer annual approvals at scale
Stated honestlyHMDA lacks credit scores, assets and reserves, so public-data estimates sit above the 1 to 2 point residual gaps found with confidential data (Bhutta, Hizmo and Ringo, 2025).
What this changes
It documents scale and institutional structure without overclaiming intent, and positions public-data estimates explicitly against the confidential-data literature.
Citation
Manuscript under review. The citation will be posted on acceptance. Reach me at rajveerpall04@gmail.com.