When Will Mortgage Rates Fall?

Forecasters see 6.7-6.8% through 2027; the MBA raised its number two months running. Waiting vs buying now: payment parity needs a rate below about 5.9%.

Written by Jonathan Nyst Updated October 2026 4 min read Rates & timing Every figure re-runnable

Nobody knows — and the people paid to know keep moving their answer in the wrong direction. As of October 2026, the 30-year fixed sits at 7.40% (Freddie Mac’s weekly survey, October 8 — the year’s high, and the highest since November 2023), about six-tenths of a point above what the major forecasters projected for this quarter only weeks ago: Fannie Mae’s September forecast says 6.8% for the fourth quarter of 2026 and 6.7% for every quarter of 2027, and the Mortgage Bankers Association says 6.8% through the first half of 2027 — a number the MBA has raised two months running, from 6.5% in July. Our model can’t predict rates either. What it can do is price the bet you’re actually making by waiting — and at 3% price growth, waiting two years only beats buying now if you land a rate below about 5.9%.

What do the forecasters actually predict?

The two most-cited houses, as of their September 2026 vintages: Fannie Mae (published September 11, with rates as of August 31, so before the Fed’s September 16 hike) projects the 30-year at 6.7% in Q3 2026, 6.8% in Q4, and 6.7% in every quarter of 2027. The MBA (September 16) projects 6.7% in Q3 2026, 6.8% in Q4 2026 and the first half of 2027, then 6.7% through the end of 2027. A month earlier the MBA had 6.7% for that whole stretch; in July it was 6.5% — two upward revisions in two months. Fannie Mae, for its part, trimmed the first half of 2027 by a tenth. Both vintages predate the survey’s move above 7% in late September, and neither house had published an October update as of October 5.

That revision is the real lesson. Rate forecasts are honest guesses that move with every inflation print, and the same institutions have repeatedly pushed “rates come down next year” one year further out. Treat any confident date you read — including a Fed meeting — as a guess. The Fed doesn’t set mortgage rates anyway; mortgage rates and the Fed walks through what actually moves them, and what the remaining 2026 FOMC meetings can and can’t change.

What does waiting for lower rates actually cost?

Waiting is a bet that rates fall faster than prices rise. Price the bet instead of vibing it. At our model’s default 3% appreciation, here’s the same $420,000 home bought now versus bought later at the rates optimists hope for, 20% down each time:

Each row assumes a rate drop larger than any forecaster currently projects, and still saves only $50–150 a month while demanding thousands more down. The break-even is precise: after two years of 3% price growth, the waiting buyer needs a rate below about 5.9% just to match today’s $2,124 payment — against a professional consensus of 6.7–6.8%. After one year, the bar is about 6.2%. And a fixed rate is a one-way bet: buy at 6.5% and a genuine rate collapse is still yours via refinancing, while waiting gives you no claim on today’s price.

Is renting while you wait throwing money away?

No — and this is where our model departs from the pressure you’ll read elsewhere. At October 2026 rates, renting isn’t the price of indecision; it’s the winning column. In the default scenario — $2,100 rent against that $420,000 home at 6.5% — renting and investing the difference ends about $67,000 ahead over ten years. Buying only beats renting at this rate when comparable rent exceeds about $2,468 a month. If your rent is below the tipping point for your numbers, waiting costs you nothing in net-worth terms, however long rates take.

The tipping rent moves with the rate, which is the honest way to think about “when rates drop”: at 6% it’s about $2,350 and renting’s margin narrows to about $46,000; at 5.5% it’s about $2,233 and the margin is about $24,000 — a near coin flip; at 4% the default scenario flips to buying outright. Rather than forecasting when that happens, know your threshold in advance — the full rate ladder has every rung.

So what should you actually do?

Stop asking when rates fall and ask which side of the tipping rent you’re on — that question has an answer today. If your comparable rent is below the tipping rent for your price and rate quote, rent, invest the difference, and let the waiting cost you nothing; re-run the numbers when the survey average genuinely moves. If you’re above it — or an affordable-market ratio puts you there — buying already wins at today’s rates and waiting is the speculative move, since the break-even math rewards time in the house. And if the deal only pencils at a rate nobody projects before 2028, the spreadsheet is telling you the answer: it doesn’t pencil. When buying does win at today’s rate, waiting has a price of its own — should I buy a house now or wait puts a number on each year of it.

Run your own numbers

Pre-compute your decision at three rates instead of forecasting one: run your scenario at our 6.5% default, then half a point lower, then a full point lower, with your own price and rent swapped in. If the verdict matches in all three, rates were never your variable. Every figure here is computed by the open-source engine documented on the methodology page — including the tipping rents, which are the only rate forecast this site will ever publish.

Jonathan Nyst

Jonathan Nyst built RentVsBuyMath as an independent project — for himself first. He spent fifteen years marketing financial products, in banking, fintech and payments at CMO level, which is exactly how he knows what a lead-generation calculator looks like from the inside. This is the calculator without the funnel: he is not a lender, broker or agent, the site takes no referral fees, and nobody is paid more if you decide to buy. The whole model is public — every formula documented on the methodology page, the engine MIT-licensed, covered by automated tests including scenarios computed by hand to check it.

Drafted with AI from the engine's output; reviewed and published by Jonathan Nyst. How this site is made · Corrections