Home › Guides › How Your Credit Score Affects Your Mortgage Rate
Your credit score is one of the biggest things that decides what mortgage rate a lender offers you.
Two people can walk into the same bank on the same morning and walk out with very different mortgage offers. One has a 780 credit score and gets a quote of 7.00%. The other has a 640 credit score and gets a quote around 7.60%. The bank isn't being arbitrary - most of that gap comes from a published pricing rule that links credit score (and a few other factors) to the rate the borrower is offered; lender overlays and mortgage insurance account for the rest of what weaker-credit borrowers pay.
Your credit score is one of the biggest things that decides what mortgage rate a lender offers you.
Two people can walk into the same bank on the same morning and walk out with very different mortgage offers. One has a 780 credit score and gets a quote of 7.00%. The other has a 640 credit score and gets a quote around 7.60%. The bank isn't being arbitrary - most of that gap comes from a published pricing rule that links credit score (and a few other factors) to the rate the borrower is offered; lender overlays and mortgage insurance account for the rest of what weaker-credit borrowers pay.
The rule has a name: the Loan-Level Price Adjustment matrix, usually called LLPA. Fannie Mae and Freddie Mac - the two government-sponsored enterprises that buy most U.S. mortgages from lenders - publish the LLPA matrix on their websites. It's a grid: rows are credit-score bands (780 and above, 760–779, 740–759, on down through under-640), columns are loan-to-value ratios (the size of your mortgage as a percentage of the home's value). Each cell in the grid is a price adjustment in "points," where one point equals 1% of the loan amount paid upfront.
The practical effect: at a typical 80% loan-to-value (20% down payment), even the top-tier borrower (780+ credit) pays 0.375 points; a 0.00 adjustment only appears at lower loan-to-value ratios (roughly 70% or below for top scores). The bottom-tier borrower (under-640 credit) pays roughly 2.75 points upfront - that's about $11,000 on a $400,000 loan. Most borrowers don't pay that as cash at closing; instead, they accept a higher interest rate that bakes the cost into monthly payments over the life of the loan. A hit that size translates to roughly 0.6–0.75% higher rate.
This is why "shopping your rate" between lenders can only do so much. Lenders compete on margin, processing fees, and small pricing tweaks, but the LLPA matrix is the same across virtually all GSE-eligible loans, because every GSE-bound loan has the same underlying pricing. The biggest single thing you can do to lower your mortgage rate is improve your credit score - and even within a tier, moving from 720 to 740 is worth real money over the life of a 30-year loan.
The GSEs' longstanding 620 minimum credit score was eliminated in November 2025; below that level, eligibility now comes down to the automated underwriting system's risk assessment, and many lenders keep their own minimums. Borrowers who don't clear those hurdles are typically looking at FHA, VA, or portfolio loans instead - each with their own pricing schedules, which usually charge more than the equivalent GSE rate would have charged.
GSEs publish an LLPA matrix that converts FICO band + LTV into a price adjustment in points.
Fannie Mae and Freddie Mac (collectively the GSEs, or Government-Sponsored Enterprises) publish detailed Loan-Level Price Adjustment matrices on their respective Selling Guide and Single-Family pricing pages. The matrices are public - anyone can look them up - and lenders use them as the pricing baseline for any loan they intend to sell to a GSE (which is the vast majority of conforming loans).
The matrix is two-dimensional: rows are FICO credit-score bands (≥780, 760–779, 740–759, 720–739, 700–719, 680–699, 660–679, 640–659, <640), columns are LTV ratios (≤60, >60 to 70, >70 to 75, >75 to 80, >80 to 85, >85 to 90, >90 to 95, >95). Each cell contains an LLPA expressed in points. One point equals 1% of the loan amount paid upfront. A borrower can pay points as cash at closing or, more commonly, accept a higher interest rate that has the points priced in.
A worked example. Take a $400,000 30-year fixed-rate conforming purchase loan with 20% down (80% LTV). Even at a 780+ FICO, the LLPA cell is 0.375 points ($1,500 on this loan); at 760–779 it is 0.625 points. A 0.00 cell shows up only at lower LTVs - roughly 70% or below for top scores. At a 720–739 FICO, the LLPA is 1.25 points (so $5,000) - that converts to roughly 30 bps higher rate (e.g., 7.30% instead of 7.00%). At a 640–659 FICO, the LLPA is 2.25 points ($9,000), converting to roughly 55 bps higher rate. Below 640 FICO, the LLPA rises to roughly 2.75 points and the borrower may also face lender overlays (additional pricing). The numbers move with each LLPA revision; check the live matrix for current cells.
The 1-point-to-rate conversion isn't fixed - it depends on the secondary-market pricing schedule that day, the loan amount, and lender-specific pricing. As a rough rule of thumb in typical rate environments, 1 point of LLPA hit translates to roughly 25 bps of rate. So a 2-point LLPA hit means roughly 50 bps higher rate. Lenders sometimes price the conversion more aggressively (charging the points-equivalent rate hit but waiving some lender fees, or vice versa) so the apparent gap between borrowers depends on how each lender chooses to deliver the LLPA.
LLPA applies only to conforming loans sold to Fannie Mae or Freddie Mac. FHA (Federal Housing Administration) loans have their own pricing schedule and a higher upfront mortgage insurance premium (UFMIP, currently 1.75%) plus annual MIP. VA (Department of Veterans Affairs) loans have funding fees and no LLPA. USDA (Department of Agriculture) rural-development loans have their own structure. Jumbo loans - above the conforming loan limit, $832,750 baseline for 2026, higher in high-cost areas - are priced by individual lenders or jumbo securitization platforms, not by GSE schedules; jumbo pricing can be either more or less favorable than conforming depending on the lender's portfolio appetite.
Lender overlays sit on top of LLPA. An overlay is a lender's own additional risk adjustment - common overlays include minimum FICO higher than the GSE floor (e.g., a lender that won't go below FICO 660 even though the GSEs dropped their 620 floor in November 2025), maximum DTI ratio, additional reserves required, or extra documentation requirements. Overlays vary by lender; shopping multiple lenders can find better pricing for borrowers near tier boundaries because overlay practices differ.
LLPA at 80% LTV - 780+: 0.375 pts, 740-759: 0.875, ..., 660-679: 1.875, under 640: ~2.75 points (current matrix).
LLPA structure: matrix rows are FICO bands (≥780, 760–779, 740–759, 720–739, 700–719, 680–699, 660–679, 640–659, <640); columns are LTV bands (60 or below, >60–70, >70–75, >75–80, >80–85, >85–90, >90–95, >95). Each cell holds an LLPA in points. The matrix is updated periodically by FHFA direction; the May 2023 revision was the most consequential recent change.
At 80% LTV (a common purchase-loan downpayment) the current purchase matrix reads: ≥780 FICO → 0.375 points; 760–779 → 0.625; 740–759 → 0.875; 720–739 → 1.25; 700–719 → 1.375; 680–699 → ~1.75; 660–679 → 1.875; 640–659 → 2.25; <640 → ~2.75 (verify against the live Fannie Mae matrix; cells vary slightly by GSE and by revision). At the ~25-bps-per-point rule of thumb used throughout this guide, that runs from roughly 9 bps of rate impact at the top band to roughly 70 bps at the bottom - directionally accurate but not lender-quoted prices.
LLPA payment mechanics: a borrower can pay the LLPA as upfront cash at closing (the "points paid" version) or accept a higher interest rate (the "premium pricing" version, where the lender takes the higher rate as additional income and credits the borrower back the LLPA cost). The math is approximately equivalent over a typical loan-life expectation; the choice matters for cash-flow planning and tax treatment (points may be deductible in the year paid for purchase loans; consult a tax professional).
Rate-to-points conversion: 1 point ≈ 25 bps of rate in typical environments, but the exact conversion is set by the mortgage-backed-securities secondary market pricing schedule on the lock day. The conversion is steeper (more bps per point) when the rate environment is volatile or coupons are scarce, shallower when production is heavy. Lenders sometimes call the conversion the "buy-up / buy-down ratio."
May 2023 LLPA revision: the FHFA-directed update flattened the curve between top and bottom tiers - high-FICO borrowers saw modest LLPA increases at higher LTV columns, and middle-tier FICO borrowers saw modest LLPA reductions. The political controversy: critics argued the revision "subsidized" lower-credit borrowers at the expense of higher-credit ones; defenders argued it offset cross-subsidies that had previously gone the other way and improved overall pricing fairness. The associated DTI-based LLPA (which would have added pricing based on debt-to-income ratio) was rolled back before launch after industry pushback. The Single-Family Pricing Framework page on FHFA's site documents the policy.
LLPA scope: applies only to GSE-eligible conforming loans (Fannie Mae and Freddie Mac, with substantially similar but not identical matrices). Loans above the conforming limit ($832,750 baseline for 2026; up to $1,249,125 in high-cost areas) are jumbo and follow lender-specific pricing. FHA loans use UFMIP (1.75% upfront) plus annual MIP (0.55% for most loans), independent of LLPA. VA loans use VA funding fees (1.25–3.3% depending on down payment and first-vs-subsequent use), with disability waivers. USDA Rural Development loans use upfront and annual guarantee fees.
Lender overlays in practice: common overlays include (1) a lender-set minimum FICO (e.g., 660) even after the GSEs dropped their 620 floor in November 2025; (2) max DTI lower than 50% (e.g., 45% or 43%); (3) additional reserves (months of PITI in cash); (4) overlays on property types (condos, manufactured homes, multi-unit); (5) tighter rules on cash-out refinance LTV. Overlay practices vary widely by lender - bank lenders tend to be more conservative; independent mortgage banks tend to apply GSE matrix more cleanly.
Mortgage insurance (PMI) layers on top of LLPA for conventional conforming loans above 80% LTV. PMI is paid monthly as part of the mortgage payment; under the Homeowners Protection Act it can be cancelled on request at 80% LTV and terminates automatically at 78%. FHA loans carry FHA mortgage insurance premiums instead of PMI - upfront UFMIP plus annual MIP - which apply regardless of LTV and last 11 years when the original LTV was 90% or below, otherwise for the life of the loan. PMI rates depend on FICO and LTV in their own matrix, so high-LTV low-FICO borrowers face a double credit-related penalty: higher LLPA + higher PMI rate.
For analytical or research use, Freddie Mac's Single-Family Loan-Level Dataset and Fannie Mae's Single-Family Loan Performance Data provide actual origination rates by FICO/LTV bucket for the GSE universe. The CoreLogic, Black Knight, and ICE Mortgage Technology proprietary databases cover broader scope (including non-conforming) at higher granularity. The data on this site is illustrative - derived from FRED MORTGAGE30US (the PMMS 30-year fixed conforming rate) overlaid with tier spreads computed from the LLPA matrix at 80% LTV. Real lender quotes will differ.
Common analytical mistakes: (1) treating LLPA tier spreads as fixed across rate environments - the points-to-rate conversion shifts with secondary-market conditions; (2) ignoring lender overlays, which can be larger than LLPA differences for borrowers near tier boundaries; (3) treating mortgage insurance cost as separate from the rate - borrowers experience PMI as part of their effective monthly payment burden; (4) treating jumbo and conforming as identically-priced when they're often quite different; (5) ignoring closing-cost differences across lenders, which can swamp small rate differences.
GSE pricing schedule; May 2023 LLPA revision flattened the curve and rolled back DTI adjustment.
LLPA matrices are published by both Fannie Mae ("Loan-Level Price Adjustment Matrix") and Freddie Mac ("Credit Fees"). The two matrices are substantively similar but not identical - small cell-level differences exist, and lenders selling to one GSE vs the other will see corresponding price differences. Most lenders that sell to both effectively use the worse of the two for borrower-facing pricing.
May 2023 revision detail: FHFA Director Sandra Thompson announced the updated pricing framework in early 2023, effective May 1, 2023. The headline change was a re-flattening of LLPAs that had been added in 2008 (Adverse Market Delivery Charges and risk-based LLPAs introduced post-GFC). The revision lowered LLPAs for borrowers in the 60–80% LTV / 680–759 FICO band (middle-tier borrowers) and raised them for borrowers in the very high-LTV / very-high-FICO corner. Public commentary characterized this as redistribution; FHFA's framing was that it corrected accumulated structural drift in the framework.
The rolled-back DTI-based LLPA: a separate FHFA proposal would have added price adjustments based on debt-to-income ratio (additional LLPA for DTI >40%). The mortgage industry pushed back strongly, citing DTI volatility during origination, sensitivity to last-minute borrower financial changes, and operational complexity. FHFA delayed the DTI LLPA in May 2023 and ultimately rescinded it. The FHFA Single-Family Pricing Framework page documents the rescission.
Historical LLPA evolution: original adverse-market LLPAs were introduced in 2008–2009 in response to the housing crisis, when GSEs needed to price more conservatively. The Calabria-era FHFA (2019–2021) added the 0.5% adverse market refinance fee in 2020 (repealed July 2021); the upfront fee increases on second homes and high-balance loans were announced in January 2022 under Acting Director Sandra Thompson, effective April 2022. The FHFA under Thompson (2021–2025) recalibrated several of those adjustments, including the 2023 revision; Bill Pulte has led the agency since March 2025.
For researchers, Freddie Mac's Single-Family Loan-Level Dataset and Fannie Mae's Single-Family Loan Performance Data provide loan-level origination data for the Fannie and Freddie acquired universe back to 1999. Variables include FICO, LTV, DTI, origination rate, originating channel, loan purpose, occupancy, property type, MSA. These datasets are the canonical sources for LLPA-impact studies, mortgage-rate-by-FICO research, and prepayment modeling.
Non-GSE pricing schedules: FHA pricing depends on UFMIP (1.75% upfront, financed into the loan) plus annual MIP, which varies by base loan-to-value, base loan amount, and loan term (e.g., 0.55% annual MIP for most 30-year FHA loans). Annual MIP is paid monthly as part of the mortgage payment. VA funding fee depends on down payment percentage, first-time vs subsequent use, and military service category; ranges roughly 1.25–3.3% upfront, financed into the loan; service-connected disabled veterans are exempt. USDA Rural Development guarantee fee: 1.0% upfront + 0.35% annual.
Non-conforming jumbo pricing varies dramatically by lender. Some jumbo programs (typically bank balance-sheet products at large depositories targeting high-net-worth clients) price below conforming because the bank values the broader relationship; others (jumbo securitization-platform loans, non-QM jumbo) price above. The jumbo-vs-conforming spread has flipped sign repeatedly in the last decade; in 2024–25 the spread is generally positive (jumbo more expensive than conforming) by 25–75 bps for prime borrowers.
Non-QM (non-qualified-mortgage) and portfolio loans: for borrowers who don't fit GSE/FHA/VA/USDA boxes (self-employed without W-2, foreign nationals, investor loans on rental properties, high-DTI borrowers, recent credit events), specialty programs price independently. Non-QM rates typically run 50–250 bps above prime conforming, depending on the specific deviation from standard underwriting.
FICO score-vintage matters: the classic mortgage scores are FICO Score 2 (Experian), FICO Score 5 (Equifax), and FICO Score 4 (TransUnion). Mortgage lenders pull a tri-merge credit report and use the middle score (or lowest-of-borrowers if multi-borrower). These classic models are older and somewhat less predictive than current FICO models; since July 2025, FHFA permits lenders to use VantageScore 4.0 (or the classic FICO scores) on GSE loans, and FICO 10T has been validated but is not yet required.
MSR (Mortgage Servicing Rights) economics interact with LLPA pricing because lenders earn a servicing income strip from each loan they originate and retain (or sell with retained servicing). MSR values depend heavily on prepayment expectations; LLPA-paying loans (higher-rate loans) tend to have different prepayment profiles than premium-coupon loans. The relationship feeds back into how lenders price LLPA - sometimes the all-in economics differ from the headline LLPA suggests.
Practitioner caution: PMMS and consumer-facing rate quotes are vanilla-profile-normalized. Real borrower rates depend on the full LLPA matrix, lender overlays, PMI, and lock-day market conditions. A consumer reading PMMS at 7.10% should expect their actual lock quote to be 0–200 bps higher depending on their specifics. For institutional analysis comparing mortgage rates to Treasury yields, the appropriate adjustment is to PMMS-equivalent (vanilla profile) - not actual originated rates.