Nobody sends a press release when mortgage rates bottom out. The low arrives quietly, usually in the middle of a grim news cycle, and by the time it feels obvious the window has half closed. So it helps to know roughly where the floor sits before you need to act on it.
The short answer: in a stable economy the 30-year fixed probably bottoms out near 5%. In a real recession it can reach the low 4s. Getting back to the 3% loans of 2020 and 2021 takes a genuine crisis, and that crisis arrives with an unemployment rate that makes cheap money feel like small consolation.
Here is how the math works, and where the real limits are.
The floor is set by two numbers, and neither belongs to the Fed
A 30-year fixed rate is essentially a Treasury yield plus a markup for the risk of lending to you for three decades. The reference point is the 10-year Treasury note, because the average mortgage gets paid off in roughly seven to ten years through a sale or a refinance.
That markup, called the primary mortgage spread, has averaged about 1.7 percentage points for most of the past fifteen years. Run the arithmetic and the question gets much simpler:
- 10-year at 4.5% plus a 1.7-point spread gives you a 6.2% mortgage
- 10-year at 4.0% plus a 1.7-point spread gives you a 5.7% mortgage
- 10-year at 3.0% plus a 1.7-point spread gives you a 4.7% mortgage
For a rate starting with a 3, the 10-year has to fall below about 2%. That is a bond market pricing in a serious recession or deflation, not a soft landing.
Why the Fed gets the credit and the blame
The federal funds rate is an overnight rate between banks. It influences the 10-year only indirectly, through what it signals about future growth and inflation. When the Fed cut by half a point in September 2024, mortgage rates had already fallen more than a full point in the months beforehand, and within weeks they were climbing again. The pattern repeats because markets price expectations, not announcements, which is why your payment doesn’t always fall after a Fed rate cut.
The spread is the part most forecasts skim past
Between 2010 and 2021 that markup sat reliably near 1.7 points. In late 2023 it blew out past 3 points, the widest gap since the 1980s. Three things pushed it there: prepayment risk (when rates are high almost nobody refinances, so the mortgage bonds investors hold last far longer than they want), the Fed shrinking its balance sheet, and plain day-to-day volatility.
The encouraging part is that spreads compress on their own once volatility calms down. If the 10-year simply holds at 4.0% and the spread normalises to 1.8 points, the 30-year fixed lands around 5.8% with no help from anyone. Leave the spread at 2.8 points and the same bond market hands you 6.8%.
What a recession would realistically do to the floor
Recessions pull mortgage rates down because money runs toward safety. Treasury yields fall and mortgage rates follow, though not one-for-one and not always straight away. There is a well-documented pattern where the first wave of bad economic news pushes rates lower, then a credit-quality scare pushes them partway back up. That is what a recession actually does to mortgage rates, including when it backfires.
The 2020 low was an intervention, not a trend
Rates hit 2.65% in early January 2021. The Fed was buying roughly $40 billion of mortgage-backed securities a month and had cut its policy rate to zero, while unemployment sat at 6.7% and had touched 14.7% the previous spring. Nobody refinancing into a 2.65% loan felt fortunate at the time. Reproducing that low requires a shock of similar scale.
The 2012 to 2016 plateau is a better template
Rates bottomed at 3.31% in November 2012, then spent four years bouncing between roughly 3.3% and 4.5%. No crisis, no aggressive bond buying after 2014, just slow growth and a Fed holding short rates near zero. A soft-landing version of the 2020s looks far more like that stretch than like the pandemic low.
Three scenarios, with actual numbers
1. No recession, inflation stuck near 3%
The 10-year holds between 4.0% and 4.5% and the spread stays near 2 points. That is a 30-year fixed in the 5.75% to 6.25% zone. Boring, and currently the most likely single outcome.
2. A clear slowdown and deeper Fed cuts
The 10-year drifts to 3.5%, the spread tightens to 1.8 points, and rates land near 5.1%, maybe 4.9% on a good week. This is the scenario most published forecasts have in mind when they promise meaningful relief. Anything reaching further out should be read carefully: even the 2027 outlook is better treated as a range than a number.
3. Recession with the Fed buying bonds again
A 10-year at 2.8% and a spread back to 1.6 points gets you roughly 4.4%. A brief dip into the high 3s is possible during a panic week, but those quotes tend to come with tighter credit standards, more appraisal problems and a lot of cancelled purchase contracts.
What would genuinely have to break for a 3% mortgage
- The 10-year Treasury below 2%, which means deflation fears or a global flight to safety
- Large-scale Fed purchases of mortgage-backed securities, not just rate cuts
- Credit tight enough that a meaningful share of buyers no longer qualify
- Or a targeted subsidy: builder rate buydowns and assumable loans already hand a small slice of buyers headline rates near 3%
What actually moves the number in the next few months
Trying to time a move inside a 90-day window is a different exercise from forecasting a year. Weekly pricing is decided by a narrow set of inputs, and the three things that decide mortgage rates next week are almost always inflation data, Fed communication and Treasury supply. Annual predictions skip most of that machinery, which is what a proper rate outlook covers that headline forecasts leave out.
The break-even math on waiting
Assume a $400,000 mortgage. At 6.5% the principal and interest comes to $2,528 a month. At 5.5% it is $2,271. That difference is $257 a month, or about $3,080 a year.
Now price the wait. Six months of renting at $2,200 costs $13,200. If home prices rise 3% while you sit out, that is another $12,000 on a $400,000 house. You have spent $25,000 to save $3,080 annually, so the break-even runs past eight years, and that is before rates move the other way, which they do about as often as not. Rates fell from 7.79% in October 2023 to 6.1% by September 2024, then climbed back above 7% by January 2025. The floor is not a one-way elevator.
For anyone already in a home, the useful threshold is smaller. Closing costs typically run 2% to 3% of the loan, and it generally takes a drop of one to one-and-a-quarter points to make a refinance pay for itself inside a couple of years. Going from 7.5% to 6.2% clears that bar comfortably. Chasing a 6.2% to 5.9% move usually does not.
So buy when the payment works at the rate in front of you, not the rate you are hoping for. If it only pencils out at a number nobody has offered yet, you cannot afford that house today. The floor will arrive when it arrives, and it will almost certainly be lower than the last quote you saw and higher than the figure in your head.
