themortgagellm^™

Written by The Mortgage LLM Team—a group of industry analysts leveraging our proprietary mortgage-domain language models to synthesize and decode housing data.

Among FHA-insured single-family loans originated in 2018 or later — the window where origination-time borrower and loan attributes are reliably reported — the national 60-plus-day delinquency rate stands at 7.9% as of the first quarter of 2026. Our companion analysis mapped where that stress concentrates above the historical norm. This piece runs the question in reverse, and the answer turns out to be more interesting than the one we expected to find.

Eight states are running 130 to 320 basis points below the national FHA delinquency rate: Montana and Alaska (each near 4.7%), Wyoming (5.3%), West Virginia (5.7%), Idaho (5.8%), New Mexico (5.9%), Hawaii (6.2%), and Nevada (6.6%). A gap that wide invites an obvious explanation — these states must lend to stronger borrowers, or write safer loans. So we decomposed each state’s distance from the national rate into three parts: borrower selection (do they originate to higher-credit, lower-DTI borrowers?), loan composition (lower LTVs, smaller balances, a different purchase-versus-refinance mix?), and a regional residual — whatever survives after both controls. The decomposition applies the national FHA delinquency rate within each credit-score, DTI, LTV, and loan-size bucket to the state’s own mix of loans, so any state that simply holds a “better” book of loans gets no credit for it in the residual.

The headline: across the eight states, borrower and loan characteristics together explain barely a quarter of the outperformance. The other three-quarters is the residual — the part our attributes cannot see. These states are not beating the cycle because of who they approve or what they originate. They are beating it because of where they are.

It isn’t who they lend to

Borrower selection is the explanation everyone reaches for first, and it is real but small. The eight outperformers do skew slightly stronger on credit: Hawaii’s mean FICO is 691 and Alaska’s 687, against a national FHA mean of 676. But borrower selection accounts for only about 40 of the 230-basis-point average gap — under a fifth of it, and small next to the residual.

The state that detonates the selection thesis is West Virginia. Its FHA borrowers carry a mean credit score of 666 — ten points below the national average — and yet West Virginia outperforms the national rate by 224 bps. Run the math and its borrower profile is a negative contributor: on credit characteristics alone, West Virginia should be doing slightly worse than average. It beats the cycle anyway. Whatever is protecting West Virginia’s FHA book, it is emphatically not the quality of the borrowers in it.

And it isn’t what they originate

Loan composition fares even worse as an explanation, accounting for under a tenth of the gap — roughly 17 bps on average. There is exactly one clean exception, and it proves the rule. Hawaii is the only one of the eight where the book genuinely does the work: its mean loan-to-value ratio is 86%, nearly seven points below the national 93%. The reason isn’t bigger down payments — FHA caps purchase LTV at 96.5%, and Hawaii’s median loan sits right there, at 96.5%. It’s the loan mix. In Hawaii’s high-cost market (median FHA loan north of $430,000), only 45% of the active book is purchase money against 64% nationally, while equity-heavy cash-out refinances — which FHA caps at 80% LTV — make up 21% of the book versus 10% nationally. That low-LTV mass (a 25th-percentile LTV of just 78%) pulls the average down and mechanically lowers delinquency. With that mix doing the work, selection and composition together explain roughly 80% of Hawaii’s outperformance, leaving a residual of just 33 bps. Hawaii is the closest thing to a portable story — a loan-mix and credit profile another lender could lean toward — though even that mix is partly a product of the local equity environment.

It is also the lone exception. For the other seven states, composition is a sideshow. Alaska actually originates at a higher average LTV than the nation, making composition a slight drag — and Alaska still beats the cycle by 319 bps.

Figure: Each state’s distance from the national FHA 60+ rate (7.9%), split into the share explained by borrower selection, the share explained by loan composition, and the unexplained regional residual. Outperformers extend right; the five contrast states extend left. The residual dominates on both sides. Interactive version — hover any state for its observed rate, the three-way split in bps, and its mean FICO, LTV, DTI, and loan size against the national book.

It’s where

Strip out who and what, and three-quarters of the gap is left standing as geography. The pattern in the residual is not random. The outperformers are low-density, lower-cost Mountain West, Plains, and Pacific-outlier states. The mirror image — the five contrast states we ran alongside them — are higher-cost, coastal-urban, and Deep South markets: New Jersey (8.8%), Georgia (9.7%), Illinois (10.3%), Louisiana (10.7%), and Maryland (10.9%), each running well above the national rate, and each, like the outperformers, driven overwhelmingly by a residual rather than by weak borrowers or bad loan mix.

A residual is, by construction, the part we couldn’t measure — so the honest task is to ask what plausibly lives inside it without overclaiming. The strongest candidate is accumulated home equity. The Mountain West and Plains states absorbed some of the largest cumulative price gains in the country during the 2020–2023 boom, on top of FHFA’s roughly 55% national appreciation since early 2020. For a low-down-payment FHA borrower, equity is the difference between a forced default and an orderly sale: a homeowner who hits trouble but holds equity sells the house rather than surrendering it to foreclosure. That mechanism shows up nowhere in a loan’s origination-time FICO or LTV — it accrues afterward, from the market — which is precisely why it lands in the residual.

The macro data also disciplines the story, and we’ll be candid about where it resists a tidy narrative. The obvious single-variable explanations do not cleanly fit. Unemployment is a muddle: Hawaii (2.2%), Montana, and Wyoming (3.4%) run among the lowest rates in the nation, but Alaska (4.8%) and Nevada (5.2%) run among the highest and still outperform — while Georgia, an underperformer, sits at a healthy 3.6%. Recent price growth cuts the wrong way too: Illinois posted one of the nation’s three fastest home-price gains in 2025 (+6.1%) yet sits near the bottom of our delinquency table. The throughline is the accumulated equity cushion of the boom years, not the most recent quarter’s labor or price print.

On the underperformer side, a measurement effect compounds the economics: our delinquency rate counts loans in foreclosure and bankruptcy as still-active and still-delinquent, and judicial-foreclosure states like Illinois and New Jersey keep distressed loans in that pipeline far longer, inflating a point-in-time reading. Louisiana carries its own distinct burden — a property-insurance crisis that has pushed all-in housing costs sharply higher. None of these is a complete account, and two states stay stubbornly unexplained: West Virginia outperforms despite weaker borrowers and middling unemployment, and Georgia underperforms despite low unemployment and fast non-judicial foreclosure. We flag them rather than force them.

The underwriting takeaway: alpha versus geography

For an underwriter or an MBS investor, the split is the whole point. Composition-driven resilience — Hawaii’s lower LTVs, a tighter credit box — is portable: another lender could adopt the same profile and import the same protection. Residual-driven resilience is not. You cannot underwrite your way into Wyoming’s equity cushion or Montana’s housing economy. Three-quarters of what makes these eight books exceptional is a geography, not a strategy, which means it should be priced as a feature of the collateral’s location rather than as evidence of superior origination skill. The same logic flips for the contrast states: their elevated delinquency is largely regional too, so it should inform geographic credit overlays, not blanket conclusions about borrower quality in those markets.

The forward question is whether the equity cushion holds. The Mountain West has cooled — prices there were flat to slightly negative in 2025 — so the residual that has been protecting these books is no longer being topped up. If the 2022–2023 vintages age into their peak-delinquency window against softening prices, the gap that looks structural today may begin to compress. The next disclosure will start to tell us.

---

Methodology. Performance and attribute data sourced from Ginnie Mae’s loan-level monthly disclosure, consolidated via The Mortgage LLM’s analytics instance, covering FHA-insured single-family loans as of the first quarter of 2026 and restricted to originations from 2018 onward, where borrower and loan attributes are reliably populated. The 60+ delinquency rate counts active loans 60 or more days delinquent, including those in foreclosure, bankruptcy, or REO. The composition decomposition computes each state’s expected delinquency rate by applying national FHA delinquency rates within borrower-and-loan-attribute buckets (credit-score, DTI, LTV, and loan-size bands) to that state’s own distribution across the buckets; borrower selection (credit and DTI) and loan composition (LTV and loan size) are separated using a Shapley attribution, and the residual is the difference between observed and composition-expected performance. Supplementary context on home prices and unemployment is drawn from the FHFA House Price Index and the Bureau of Labor Statistics; those associations are directional and do not establish loan-level causation. Puerto Rico and the US Virgin Islands are excluded as structurally distinct territory markets. Informational, not advice.