Home › Guides › How Are Mortgage Rates Determined?
The U.S. 30-year mortgage rate comes from two big sources: Freddie Mac's weekly PMMS index and Optimal Blue's daily lock data.
When the news says "mortgage rates hit 7% this week," they're almost always citing one of two specific numbers. The older, more famous one is the Freddie Mac Primary Mortgage Market Survey, usually shortened to PMMS. It's been published every Thursday since 1971, which means it's the longest continuous mortgage-rate dataset in the United States and the source most economists and journalists reach for first.
The U.S. 30-year mortgage rate comes from two big sources: Freddie Mac's weekly PMMS index and Optimal Blue's daily lock data.
When the news says "mortgage rates hit 7% this week," they're almost always citing one of two specific numbers. The older, more famous one is the Freddie Mac Primary Mortgage Market Survey, usually shortened to PMMS. It's been published every Thursday since 1971, which means it's the longest continuous mortgage-rate dataset in the United States and the source most economists and journalists reach for first.
The newer one is the Optimal Blue Mortgage Market Indices, or OBMMI. Optimal Blue is a company whose pricing engine handles roughly one-third of all U.S. mortgage locks. They publish a daily index based on the actual locked mortgages flowing through that software. PMMS is once a week and, despite the name, is no longer a survey: since November 2022 Freddie Mac has computed it from actual loan applications submitted through its Loan Product Advisor underwriting system. OBMMI is every day and is built from what borrowers actually locked.
The two numbers often look noticeably different on any given day. PMMS might say 7.10% while OBMMI says 6.85%. Neither is wrong - they're answering slightly different questions. PMMS reflects rates on loan applications from low-risk borrowers (very good credit, 20% down) for a vanilla 30-year fixed. OBMMI reflects whatever rates actually got locked yesterday, across the full mix of borrowers (including FHA loans, jumbo loans, refinances, and rate-and-term variations).
For a borrower, the practical takeaway is to ignore the headline number when shopping. The rate you actually qualify for depends on your credit score, your down payment, your loan size, the property type, and the lender's individual pricing. Headlines are useful for tracking the general direction (rates rising, rates falling), but the actual quote you get could easily be a half-point higher or lower depending on your specifics.
PMMS (Freddie Mac weekly index since 1971, application-based since 2022) vs OBMMI (daily, ~⅓ of U.S. locks).
PMMS - Freddie Mac's Primary Mortgage Market Survey - kept its name but retired its survey. Since November 17, 2022 it has been computed from actual first-lien conventional conforming purchase-loan applications submitted through Freddie Mac's Loan Product Advisor automated underwriting system, aggregated over the week and published Thursday. The profile is still vanilla: 30-year fixed-rate conforming loans to low-risk borrowers, typically 80% loan-to-value, FICO 740+, owner-occupied single-family primary residence, conforming loan amount. Before November 2022, PMMS really was a survey: roughly 80 lenders polled each week on the rates they were offering. PMMS also publishes a 15-year fixed rate; its 5/1 adjustable-rate mortgage (ARM) series was discontinued in November 2022.
Until November 2022, PMMS was published alongside an average points figure (e.g., a 7.10% rate with 0.7 points); since the methodology change it is rate-only, because application data does not reliably capture points. Discount points are upfront fees paid to buy down the rate; one point is 1% of the loan amount. The all-in cost of a mortgage is the rate plus the points, and rate-vs-points tradeoffs vary by lender and rate environment. As a rough conversion in typical environments, one point reduces the rate by about 10–25 basis points depending on the curve and the specific lender, though this varies widely.
OBMMI - Optimal Blue Mortgage Market Indices - is derived from rate locks flowing through Optimal Blue's Product, Pricing, and Eligibility (PPE) engine, which handles locks covering approximately one-third of all U.S. mortgage transactions. The index publishes daily (the day after the trade date) and is calibrated to actual locked rates. Optimal Blue publishes six daily product indices: conforming 30-year fixed (OBMMIC30YF on FRED), jumbo 30-year fixed (OBMMIJUMBO30YF), FHA 30-year, VA 30-year, USDA 30-year, and conforming 15-year, plus FICO/LTV credit-tier breakouts of the 30-year conforming index. All are fixed-rate; there are no ARM or 15-year jumbo indices.
The two indices often differ by 25–50 basis points on a given day, sometimes more, sometimes less. Direction and trend usually agree even when level differs. Reasons they diverge: PMMS is weekly and built from application rates; OBMMI is daily and built from realized lock rates. PMMS is normalized to a vanilla profile; OBMMI reflects whatever mix locked that day. Both are legitimate; for trend purposes either works; for level purposes, the choice depends on the question.
The two have different historical depth. PMMS goes back to April 1971 - the longest continuous mortgage-rate series in the U.S. It captures the entire rate cycle since the high-inflation 1970s and 1980s, including the 18.63% peak in October 1981. OBMMI history begins in January 2017, so it covers only the recent cycle. For long-run analysis, PMMS is the only choice. For current-week granularity, OBMMI is more informative.
Third-party alternatives exist. The Mortgage Bankers Association (MBA) publishes the Weekly Mortgage Applications Survey each Wednesday, which includes average contract rates and application indices; the data is available from MBA directly (mba.org) and via commercial vendors, not on FRED. Mortgage News Daily quotes a daily rate derived from actual lender rate-sheet pricing; its lender pool is not fully documented publicly. Bankrate, NerdWallet, and lender aggregator sites publish daily numbers that vary in methodology. For institutional analysis, the practitioner-standard sources are PMMS for the long-run anchor and OBMMI or MBS-derived primary rates for daily resolution.
PMMS: weekly, from Loan Product Advisor applications since Nov 2022, 80% LTV / 740+ FICO profile; OBMMI: daily PPE lock data.
PMMS operational detail: since November 17, 2022, computed from actual first-lien conventional conforming purchase applications submitted through Freddie Mac's Loan Product Advisor (LPA) automated underwriting system, aggregated over the week from thousands of lenders nationwide and published Thursday. Before that date it was a Monday-through-Wednesday survey of roughly 80 lenders spanning depository institutions (banks, credit unions, thrifts) and independent mortgage banks. The vanilla profile is conforming 30-year fixed, 80% LTV, FICO 740+, single-family owner-occupied primary residence, conforming loan limit ($832,750 baseline for 2026). Reported as a single rate; the companion average-points figure was discontinued with the November 2022 methodology change. PMMS revisions are rare; it's a relatively stable historical series.
OBMMI operational detail: derived from PPE (Product, Pricing, and Eligibility) engine lock submissions across Optimal Blue's lender client base. PPE engines compute lender-specific eligibility and price all available investor programs in real time, then present a borrower-facing rate and points option. The locked rate (the rate the borrower accepts and the lender commits to honor for some lock period, typically 30–60 days) is the OBMMI input. Volume-weighted across the daily lock flow. Daily indices broken out by product and by FICO/LTV credit tier, with FRED tickers prefixed OBMMI*.
Why they diverge structurally: (1) PMMS measures application rates normalized to a vanilla conforming profile; OBMMI measures locked rates that may include or exclude points depending on borrower choice; (2) PMMS draws on applications submitted through Freddie Mac's LPA system; OBMMI draws on locks from lenders using Optimal Blue's PPE software, a client base tilted toward independent mortgage banks, which have historically priced more aggressively at the same risk profile; (3) OBMMI's daily cadence captures intraweek moves PMMS misses entirely - a Monday rate move shows up in OBMMI on Tuesday but in PMMS only the following Thursday; (4) the underlying mix in OBMMI (FHA, VA, refinance, jumbo) shifts week to week, producing apparent rate moves that are really composition-driven.
For research and historical-comparison work, conventional practice is: long-run charts and decade-comparison work use PMMS; weekly trends and intramonth changes use either PMMS or OBMMI; institutional reports use both side-by-side. The 0–50 bp gap between them shouldn't be treated as one being right and the other wrong.
The MBA Weekly Mortgage Applications Survey (published Wednesdays; distributed by MBA directly and via commercial vendors, not on FRED) provides a third reference. The MBA rate is application-volume-weighted, so it skews to whatever borrowers are actually applying that week. In refinance-heavy environments it pulls toward refi rates (typically slightly lower than purchase); in purchase-heavy environments it pulls the other way.
For real economic analysis, mortgage rates are best understood as the 10-year Treasury yield plus an MBS spread plus a primary-secondary spread. The 10Y yield is the base. The MBS spread (current-coupon Fannie Mae 30-year MBS yield minus 10Y Treasury, a nominal spread - not the same thing as OAS) reflects MBS market conditions, prepayment expectations, and convexity hedging dynamics - it has ranged roughly 70–200+ bps historically. The primary-secondary spread (PMMS minus current-coupon FNCL yield) reflects mortgage banker margins, origination capacity utilization, and lender appetite - roughly 30–70 bps pre-2008, roughly 100–150 bps in the post-crisis era, and wider when origination capacity is tight (it peaked around 200+ bps in the 2020 refi boom).
The MBS-spread story explains why mortgages don't move 1-for-1 with the 10Y Treasury. When MBS spreads widen (e.g., March 2020 dash-for-cash, September 2022 fast rate selloff), mortgage rates rise faster than the 10Y. When MBS spreads tighten (e.g., Fed MBS QE), mortgage rates fall faster than the 10Y.
Practitioner trading view: mortgage originators are essentially long MBS production and need to hedge interest-rate risk between origination and securitization. The standard hedge is TBA (To-Be-Announced) MBS sales. Pipeline management is a meaningful operational discipline; the gap between rate-lock day and final sale into the secondary market can be 30–90 days, during which rates move and the originator's economics shift. Bigger margins in the origination business correspond to fatter primary-secondary spreads - which is why aggregate origination volume and the primary-secondary spread are inversely correlated.
PMMS aggregates LPA application rates; OBMMI captures locked rates net of concessions; samples differ.
PMMS methodology lineage: launched April 1971 as a means of tracking the conforming 30-year fixed market for Freddie Mac's internal pricing and external publication. For its first five decades PMMS was a lender survey - pre-2007 skewed to large savings institutions, post-2008 broader. On November 17, 2022 Freddie Mac retired the survey entirely: PMMS is now computed from actual conforming purchase-loan applications submitted through its Loan Product Advisor (LPA) automated underwriting system by thousands of lenders nationwide. The historical continuity is preserved through the transition, and methodology changes are documented on Freddie Mac's PMMS methodology page.
OBMMI launch context: Optimal Blue launched OBMMI itself (announced 2018, with index history back to January 2017) specifically to address the gap in U.S. mortgage-rate data: PMMS's weekly cadence was too coarse for capital-markets uses, and no publicly-available daily series existed at lock-level granularity. Black Knight acquired Optimal Blue in 2020; when ICE acquired Black Knight in September 2023, antitrust remedies required the Optimal Blue business to be sold to Constellation Software, which now owns Optimal Blue and publishes OBMMI. FRED tickers begin with OBMMI* (OBMMIC30YF for conforming, OBMMIJUMBO30YF for jumbo).
Sample composition still differs. PMMS draws on the broad base of applications submitted through Freddie Mac's LPA system, normalized to a vanilla conforming profile; OBMMI draws on locks from lenders using Optimal Blue's PPE, a base tilted toward independent mortgage banks, across the full product mix. In the legacy survey era (pre-November 2022) the contrast was sharper: PMMS panelists skewed to depository institutions - roughly 60% in the modern era - while independent mortgage banks priced more aggressively at the same risk profile, so OBMMI tended to print slightly below PMMS in normal markets, with the gap widening when bank pricing turned conservative (typical late-cycle behavior).
Mortgage News Daily (MND) - the third widely-quoted daily rate - derives its index from actual lender rate-sheet pricing rather than a fixed, documented lender panel. MND tends to track OBMMI more closely than PMMS. MND publishes intraday updates and is the source used by many lender-facing tools and consumer-facing sites that quote "today's mortgage rate." Bankrate, NerdWallet, and Mortgage News Daily are commonly cited consumer-side aggregators; their methodologies and lender pools differ and are imperfectly documented publicly.
MBS spread mechanics: current-coupon FNCL 30 (Fannie Mae 30-year MBS at the coupon currently being produced) is the relevant secondary-market reference for primary-rate decomposition. Option-Adjusted Spread (OAS) is the standard MBS valuation metric, computed as the spread over Treasury after stripping out the optionality value of prepayment. Keep the two spread measures separate: the current-coupon nominal spread to the 10Y Treasury has ranged roughly 70–200+ bps, while OAS is far tighter - roughly 0–80 bps in normal markets, briefly ~100+ bps in the March 2020 panic, and negative during peak Fed MBS QE in 2021. Primary-secondary spread = PMMS rate minus current-coupon FNCL yield; roughly 30–70 bps pre-2008, roughly 100–150 bps post-crisis, peaking around 200+ bps at the height of the 2020 refi boom as origination capacity bottlenecked.
Fed MBS QE/QT impact on the primary rate: Fed MBS purchases compress MBS spreads, which feeds through to lower mortgage rates. QE3 (2012–2014) and pandemic-era MBS purchases (2020–2022) both meaningfully suppressed mortgage rates below where they'd have been from Treasury yields alone. MBS QT (post-2022 runoff) has the opposite effect, contributing to MBS spread widening in 2022–23.
Prepayment-speed assumptions matter for MBS valuation, which feeds back into primary-rate pricing. Standard prepayment models (PSA, CPR variants) project how fast borrowers will refinance or pay down loans, which determines MBS cash-flow timing and average life. Higher expected prepayment speeds compress MBS yields. The 2023–24 "locked-in" homeowner effect (millions of borrowers with 3% mortgages unwilling to refi or sell at 7%) reduced expected prepayment speeds materially, lengthening MBS duration and widening spreads.
For borrower-level rate analysis, the LLPA (Loan-Level Price Adjustment) matrix overlay is essential - see the Mortgage Rates by Credit Score explainer. PMMS and OBMMI headline rates are normalized to vanilla profile (PMMS) or composite (OBMMI); actual borrower rates by FICO/LTV combination can diverge from the headline by roughly 25–100+ basis points, more for non-QM or credit-impaired products.
For academic research, the GSEs' own loan-level files - Freddie Mac's Single-Family Loan-Level Dataset (1999-present) and Fannie Mae's Single-Family Loan Performance Data (2000-present) - are the canonical sources for origination-rate analysis by borrower characteristics; FHFA's National Mortgage Database publishes aggregate statistics. The CoreLogic Loan-Level Market Analytics database covers non-conforming and non-GSE originations. Optimal Blue (now a Constellation Software company) sells PPE-derived datasets at higher granularity than the public OBMMI.
Known data issues: PMMS occasionally has noisy weekly readings during low-volume periods. OBMMI has occasional spikes from data-pipeline quirks in PPE lock submissions. The two series sometimes diverge by 75+ bps for a week or two before reconverging. When using either series for institutional analysis, smoothing (4-week moving average for PMMS, 5-day moving average for OBMMI) reduces noise without significantly distorting trend.
A practical research workflow: use PMMS for long-run charts (1971-present); cross-reference OBMMI for the 2017+ window where finer granularity matters; overlay 10Y Treasury yield to see the spread story; track FNCL 30 OAS for secondary-market regime context; consult GSE loan-level data for borrower-mix analysis.