Here’s a fact that breaks most people’s mental model of mortgages: through the summer of 2026 the Federal Reserve did not change its policy rate once. The 30-year fixed mortgage average rose anyway, from 6.43% in early July to a peak of 6.69% in early August, before easing back to 6.65% by 20 August.
If mortgage rates followed the Fed, that couldn’t happen. They don’t, and understanding what they do follow is the difference between reading the news usefully and reading it as noise.
Rates are made in the bond market, not at the Fed
The Federal Reserve sets a very short-term rate, essentially the price of money overnight. A 30-year mortgage is the opposite of overnight.
Fixed mortgage rates track long-term bond yields, particularly the 10-year Treasury, because that’s the market lenders sell mortgage debt into. Those yields reflect what investors expect over years: inflation, growth, how much government debt is coming. When investors expect higher inflation ahead, they demand more yield to lend for a decade, and mortgage rates follow them up.
So a Fed decision matters for mortgages only insofar as it changes those expectations. A hold that markets read as “inflation is stickier than we thought” can push mortgage rates up. That’s close to what happened at the 29 July 2026 meeting: the Fed held its target range at 3.50–3.75%, but three committee members voted to raise instead, a hawkish signal from an unchanged decision. The Committee’s own statement recorded the hold in these words:
“The Committee decided to maintain the target range for the federal funds rate at 3-1/2 to 3-3/4 percent, in support of the Federal Reserve’s dual mandate.”
— Federal Reserve, FOMC statement, 29 July 2026
The 10-year Treasury yield itself tells the same story in real time. It touched a 20-month high near 4.75% in mid-August 2026, then eased to around 4.64% by 20 August after the Treasury Department announced it would increase buybacks of long-dated securities: a supply-and-demand move in the bond market, with nothing to do with the Fed’s policy rate, that still fed straight through into the mortgage quote you’d get that week.
Why a bond, not a bank, sets your rate
The mechanical link is securitisation. A lender that writes your mortgage doesn’t usually sit on it for thirty years. It bundles your loan with thousands of others into a mortgage-backed security (MBS) and sells that bundle to investors: pension funds, insurers, foreign central banks, anyone wanting a long-dated, relatively safe return. The price investors are willing to pay for that bundle sets the rate the lender can afford to offer you, working backwards.
Investors won’t buy an MBS yielding less than a comparable-risk Treasury bond. Why take on mortgage-specific risk (people refinancing early when rates drop, for instance) for the same return as a risk-free government bond? So mortgage rates trade at a spread above the 10-year Treasury, and that spread itself moves with how much extra risk and uncertainty investors are pricing in at any given moment. This is also the mechanism behind loan limits mattering: loans within the FHFA’s conforming limit ($832,750 for a one-unit property in most of the US for 2026) are eligible to be bundled into Fannie Mae- and Freddie Mac-backed securities with the deepest, most liquid investor base, which is part of why conforming loans consistently price better than jumbo loans of otherwise identical risk.
What the numbers actually did
View the data
| Jun 25 | 6.49% |
|---|---|
| Jul 2 | 6.43% |
| Jul 9 | 6.49% |
| Jul 16 | 6.55% |
| Jul 23 | 6.58% |
| Jul 30 | 6.66% |
| Aug 6 | 6.69% |
| Aug 13 | 6.67% |
| Aug 20 | 6.65% |
Source: Freddie Mac Primary Mortgage Market Survey via FRED (MORTGAGE30US) and freddiemac.com/pmms, accessed .
A 26 basis point climb across six weeks, then two small steps back. On a $400,000 loan, moving from 6.43% to 6.69% is roughly $70 a month, about $25,000 over thirty years, for a change most headlines didn’t bother reporting.
The longer view is more useful than the weekly one
Weekly moves are noise. The band matters.
| Year | Low | High | Late-season reading |
|---|---|---|---|
| 2021 | 2.77% | 3.14% | 2.77% |
| 2022 | 3.22% | 7.08% | 4.99% |
| 2023 | 6.09% | 7.79% | 6.90% |
| 2024 | 6.08% | 7.22% | 6.73% |
| 2025 | 6.15% | 7.04% | 6.63% |
| 2026 (to 20 Aug) | 5.98% | 6.69% | 6.65% |
Two things stand out. 2022 was the break — the average more than doubled inside a single year, from 3.22% to 7.08%. And since then the market has been remarkably rangebound: four consecutive years sitting roughly between 6% and 7.8%, with 2026 so far the narrowest of them.
The useful conclusion isn’t a forecast. It’s that “waiting for rates to drop” has been a losing strategy for four years running, and that the 2021 sub-3% window looks increasingly like an artefact of its moment rather than a level the market returns to.
Will the trend actually reverse?
Nobody can answer that with certainty, including anyone confidently telling you otherwise — but the shape of the last four years is informative even without a forecast. 2022 was a genuine regime change: rates roughly doubled in twelve months as the Fed fought inflation with the fastest hiking cycle in decades. Since then, despite plenty of headline volatility, the 30-year average has not broken meaningfully outside the 6%–7.8% band in any of 2023, 2024, 2025 or 2026 so far.
That’s a different situation from “rates are stuck” — the band itself could still shift, and 2026’s range (5.98% to 6.69%) sits at the lower end of the four-year pattern, which is itself informative. What it isn’t is a signal that a return to 2021’s sub-3% environment is on the table. That period required a combination of near-zero Fed policy and emergency-scale bond buying that hasn’t been repeated since, and there’s no current policy signal pointing back toward it. Treat any specific rate prediction with real skepticism, including implicitly: an old article still quoting a rate from months or years ago as if it’s current. The date stamp on a rate claim matters as much as the number itself.
What discount points actually buy you
Rate news rarely mentions the one variable a borrower controls directly at closing: points. The CFPB defines them simply:
“Discount points are a one-time fee paid at closing to a lender in exchange for a lower interest rate.”
— Consumer Financial Protection Bureau, Data Spotlight: Trends in discount points amid rising interest rates, accessed 22 August 2026
A discount point costs 1% of your loan amount, paid upfront to the lender, in exchange for a lower rate for the life of the loan. On a $400,000 mortgage, one point is $4,000. The CFPB, which tracks this, has found there’s no fixed formula for what a point buys. One point knocking roughly a quarter of a percentage point off the rate is a common rule of thumb, but the actual trade varies meaningfully by lender, and the CFPB’s own research found the rate differential between borrowers who paid points and those who didn’t was often minor between 2018 and 2023.
| No points | One point (~$4,000 upfront) | |
|---|---|---|
| Illustrative rate | 6.65% | ≈6.40% |
| Monthly payment (P&I) | $2,568 | ≈$2,502 |
| Monthly saving | — | ≈$66 |
| Rough break-even on the $4,000 | — | ≈5 years |
That break-even is the whole decision. The CFPB found borrowers paying points climbed sharply as rates rose: for cash-out refinances, from 61% of borrowers when rates were near 2.6% in January 2021 to 87% by September 2023. Borrowers with lower credit scores were more likely to pay them too, which the Bureau reads as lenders using points to help marginal borrowers qualify for a lower payment rather than as a pure rate-shopping tool. If you won’t keep the loan past the break-even point, because you expect to move or refinance, the points are usually a loss. If you’re planning to stay put for a decade or more, they can be worth it.
15-year vs. 30-year: the actual gap, worked
The 15-year fixed averaged 5.95% against the 30-year’s 6.65% for the week ending 20 August 2026 — a 70 basis point discount, priced in because the lender’s money is exposed for half as long.
| 30-year fixed (6.65%) | 15-year fixed (5.95%) | |
|---|---|---|
| Monthly payment | $2,568 | $3,365 |
| Total paid over the term | $924,429 | $605,634 |
| Total interest paid | $524,429 | $205,634 |
The 15-year saves roughly $319,000 in total interest on paper — more than the loan amount itself. It also costs $797 more every single month, for fifteen years, with no flexibility to pay less in a lean month. Whether that trade suits your cash flow is a budgeting question, not a maths question; the maths always favours the shorter term, and the maths was never the hard part of the decision.
What you can actually control
The market average is weather. Your quote is the average adjusted for you specifically, and several of those adjustments are yours to make.
Your credit score. Lenders price risk in bands, and the difference between bands on a loan this size is measured in tens of thousands over the term. If your application is months out, improving the score first usually beats shopping harder.
Your deposit. More equity means less risk to the lender, and it can remove mortgage insurance entirely. If you’re refinancing rather than buying, running the actual numbers through a refinance estimator turns “rates have moved” into a real dollar answer for your situation.
Whether you pay points. As above — a real lever, and one most rate articles skip entirely.
Shopping around. Quotes from different lenders in the same week genuinely differ. Multiple mortgage inquiries in a short window are typically treated as a single event by scoring models precisely so that shopping isn’t punished. The same bond-market logic that caps what a high-yield savings account pays you is what a lender is pricing against on the other side of this trade — it’s the same market, seen from two different products.
The term. The 15-year vs. 30-year numbers above are the clearest version of this lever there is. Outside the US the same lever gets pulled for a different reason: a British borrower weighing a decade-long fix is buying certainty about the rate rather than a shorter payoff.
When you lock. Once you’ve applied, a lender will let you “lock” a specific rate for a set window, commonly 30, 45 or 60 days, protecting you from the market moving against you before closing. Lock too early and you might miss a drop; lock too late and a jump between application and closing lands on you. Some lenders offer a float-down option, letting you claim a lower rate if the market falls after you’ve locked, usually for an extra fee. None of this changes the market average. It changes which point on the market’s own noisy weekly path becomes your number.
What buying points now costs you if you refinance later
One thing the break-even maths above leaves out: you don’t get the points back if you refinance or sell before the break-even date. That $4,000 is gone regardless of what happens next.
This matters most when rates are elevated and falling is plausible, which is close to the environment this page describes. Paying points now to shave the rate, only to refinance in eighteen months if rates drop meaningfully, means paying twice for a lower rate: once in points, once in refinance closing costs, and never fully recovering either. If you think there’s a real chance you’ll refinance within a few years, that’s an argument for skipping points now and revisiting the decision fresh whenever you actually refinance, rather than paying for a rate reduction you might abandon before it pays for itself.
How to read rate news from here
Three habits that make the coverage useful:
- Watch the 10-year Treasury, not the Fed announcement, if you want a leading indicator of where mortgage rates go next.
- Treat weekly moves as noise. The Freddie Mac survey is a weekly average and it wobbles.
- Check the date on any rate you read. Including this page: the figures here are the week ending 20 August 2026, and the sources are listed so you can pull today’s number yourself. For a page built specifically to be read that way, see today’s mortgage rates by loan type.
- Be specific about which rate is being quoted. “30-year fixed” alone is ambiguous between purchase and refinance, conforming and jumbo, and with or without points — two accurate rates for the same week can differ by half a point once you account for what each is actually measuring.
Rates move for reasons that have nothing to do with your bank. Once that clicks, the rest of the mortgages and refinancing guide reads a lot less like a foreign language.