Testing mortgage denial gaps in Massachusetts HMDA data
Black non-Hispanic applicants are denied at a higher rate than White non-Hispanic applicants for first-lien conventional owner-occupied home-purchase loans.
Confirmed by a two-proportion z-test and the Wald-CI odds ratio, and the gap persists inside the lowest income band (H5). All five primary tests reject at BH-FDR q = 0.05 and Bonferroni. This is a screening signal, not a causal claim. See the per-hypothesis caveat.
disparity ruler
every effect estimate against its null
Each point is the primary effect estimate; the bar is its 95% confidence interval; the tick marks the null (no disparity). A row reads amber when its interval clears the null. Units differ by hypothesis: risk difference, Hedges' g, eta squared. Each row is scaled to itself and the exact values sit alongside.
denial rate by race
all applications
Raw rates without covariate adjustment. See hypothesis cards for tests and CIs.
denial rate by income band and race
non-joint applicants
Income binning is the simplest control HMDA supports. The Black vs White spread persists across all bands (H5). Cells with n < 30 are suppressed; small-n cells produce unstable rates. Hover any bar for the rate and exact n.
result summary
five preregistered hypotheses
| key | title | method | p value | effect | 95% CI | n A / B |
|---|