loan_type = 2), HECM excluded. Every figure carries a machine-readable claim passport with the source hash.
A working agent told me the big national lenders scared her, and that local is usually the answer. That is a testable claim, so I tested it against the complete 2025 federal HMDA record.
I split FHA lenders by footprint: those appearing as high-volume lenders in 40 or more metros (“national”) versus fewer than 40 (“local / regional”), counting only lenders with at least 100 decisioned applications in that metro. Then, in each of the 151 metros where both types compete, I compared their observed denial rates. Same city, same year, same program — the fairest comparison the public record allows.
Volume shows the same pattern: the ten largest FHA lenders by application count have a median denial rate of 22.1%, versus 12.6% across the rest of the top 100.
Metros with at least two local/regional lenders meeting the volume floor, so a single outlier cannot drive the row.
| Metro | National | Local | Gap (pp) |
|---|---|---|---|
| Warner Robins | 31.1% | 6.3% | +24.8 |
| Reading | 35.7% | 11.2% | +24.6 |
| Winston-Salem | 25.2% | 1.7% | +23.5 |
| Youngstown-Warren-Boardman | 31.8% | 8.8% | +23.0 |
| Montgomery | 25.8% | 3.0% | +22.8 |
| Naples-Immokalee-Marco Islan | 29.9% | 7.3% | +22.6 |
| Shreveport-Bossier City | 28.0% | 6.0% | +21.9 |
| Kennewick-Richland | 21.9% | 1.2% | +20.6 |
| Canton-Massillon | 32.9% | 13.2% | +19.8 |
| Hickory-Lenoir-Morganton | 22.4% | 2.7% | +19.7 |
Local does not mean safe. The softest door in the entire record (1.8%) and the strictest (78.7%) are both narrow-footprint lenders. National names cluster in the middle; local lenders spread across the whole range. So “go local” raises the average and raises the variance at the same time — which is exactly why the specific name matters more than the category. Your metro's table is here.
Method: CFPB HMDA 2025 public record, FHA forward loans, denied ÷ (originated + approved-not-accepted + denied), HECM excluded. Companion research: The Door Effect · Persistent Doors · What Denial Rates Cannot See (SSRN 7423798, doi:10.2139/ssrn.7423798). This question came from a working agent's comment — if you have another claim worth testing, send it.