themortgagellm™Use Case · Origination
Forecast or assess origination-stage signals at the application or loan level: pull-through, denial probability, higher-priced loan likelihood, pricing-tier benchmarking, and <strong>loan-level price adjustments (LLPAs / Credit Fees in Price)</strong> on Fannie and Freddie acquisitions across 2017-present vintages. Combine probabilistic scoring with aggregated benchmarks for the same applicant profile.
Sample prompts
Higher-Priced Mortgage Loan probability for a 660-FICO FHA application
“What's the probability this application becomes a Higher-Priced Mortgage Loan (HPML)? FICO 660, DTI 42, LTV 95, FHA loan in Texas.”
For an FHA application with the profile above (FICO 660, DTI 42, LTV 95, owner-occupied first-lien purchase), the Higher-Priced model returns a calibrated probability of roughly 20.3% with a P90 risk-band classification — elevated against the 2018-2023 industry baseline rate of ~5.8%…
Conventional purchase denial rates by race, 2023
“What are typical denial rates by race for conventional purchase loans in 2023?”
Across 2023 HMDA, conventional first-lien purchase denial rates range from 9.5% for Asian applicants to 35.2% for Black applicants on a population of 1.74M (White) / 181K (Black) / 203K (Asian) decisioned applications — a 20.8 pp Black-vs-White gap that remains a focus of fair-lending examination programs…
Pull-through probability for a $350K VA application, FICO 650, income $75K
“Score this application for pull-through probability: VA loan, $350,000 loan amount, applicant FICO 650, annual income $75,000, owner-occupied primary residence purchase.”
The pull-through model returns a calibrated probability of roughly 47.5% that this VA application closes as an originated loan — below the 61.7% industry baseline, driven mainly by the 640-679 FICO band which is the single largest negative driver in the model…
Appraisal waiver probability for a 65 LTV / 760 FICO conventional refinance
“What is the appraisal waiver probability for this loan? Conventional rate-and-term refinance, primary residence, 1 unit, $300K loan on $500K value, FICO 760, DTI 32.”
For a low-LTV, strong-credit conventional refinance, the appraisal-waiver model returns a calibrated probability around 56-58% — comfortably above the ~25% population baseline. Refinances at LTV under 70% with FICO 740+ are the historically densest segment of waiver approvals; the model flags the file as a strong pre-DU/LPA candidate for a no-appraisal close…
LLPA on a 720 FICO / 80% LTV Fannie purchase loan, June 2019
“What upfront LLPA did Fannie Mae charge on a $300,000 purchase loan, owner-occupied single-family, FICO 720, LTV 80%, fixed-rate, originated June 2019?”
Under Fannie's 2017 LLPA grid (in effect from July 2017 through March 2022), this loan paid a 0.75% upfront fee — about $2,250 on the $300,000 loan amount. The fee came from the 720-739 FICO band at 75-80% LTV cell. Under today's grid (effective May 2023), the same loan would pay 1.25% — about $3,750, reflecting Fannie's repricing of mid-FICO / mid-LTV cells under FHFA's aligned pricing framework…
How the May 2023 LLPA redesign shifted average fees paid by FICO band
“How did the May 2023 LLPA redesign shift average LLPA paid by FICO band on Fannie purchase loans?”
The May 2023 LLPA redesign was an FHFA-mandated cross-subsidy — lower-FICO borrowers got pricing relief while mid-to-high-FICO borrowers (especially 700-739 at higher LTV) paid more. Comparing loan-weighted averages on Fannie purchases originated April 2023 vs May 2023 onward, the largest reductions were in the 620-679 bands (-15 to -23 bps) while 700-739 borrowers saw +15 to +25 bps increases. The 680 FICO threshold is the crossover point…
Did 2018-2020 Fannie purchase LLPAs exceed realized credit losses?
“Did 2018-2020 Fannie purchase LLPAs exceed realized credit losses on the 740+ FICO / 70-85% LTV cohort?”
Yes — by a wide margin. On the 2018-2020 Fannie purchase 740+ FICO / 70-85% LTV cohort (~3.4M loans), loan-weighted average LLPA was 42 bps, while realized credit losses through the 2025 reporting snapshot were only 2.1 bps. The 40-bps spread reflects an extremely benign credit cycle for high-FICO cohorts in this window — the COVID-era refi wave prepaid out the weaker tail and the lock-in wall kept defaults muted…
LLPA sensitivity test: lowering fees by improving FICO or LTV (loan officer tool)
“My client is buying a $400,000 home with a Fannie conventional purchase loan. FICO 695, putting 12% down (LTV 88%), DTI 45%. What is their current LLPA, and how much could they save by either (a) paying off some debt to push their FICO into the next bucket, or (b) bringing more cash to closing to drop the LTV below the next bucket boundary?”
Current LLPA: 1.50% = $6,000 upfront fee on the $400K loan (680-699 FICO / 85-90% LTV cell). The single biggest lever is FICO improvement: boosting to 720 cuts LLPA by 50 bps ($2,000). LTV reduction is counter-intuitive at this FICO band — dropping to 80% or 75% LTV actually INCREASES LLPA under the May 2023 grid. The borrower needs to drop below 75% LTV (ideally below 70%) to save money. DTI doesn't affect LLPA directly under the current Fannie grid…