Nobody Refinances at PCE
There's a sentence in almost every 2026 business plan we've read, and in a few of our own. We'll refinance when rates come down.
It wasn't a reckless thing to write. Back in February, the consensus had the funds rate ending this year somewhere near 3.1 percent, all-in commercial borrowing costs drifting from the high fives back toward the low fives, and industrial cap rates compressing thirty or forty basis points on the strength of it. Lenders wrote extensions against that view. Sponsors sized capital calls against it. A lot of 2019 and 2020 paper got another eighteen months on the theory that eighteen months would be enough.
Eight months in, the funds rate is 3.50 to 3.75. It hasn't moved since December. At the July meeting, the committee held 9 to 3, and all three dissents were for a hike.
And the number that was supposed to deliver the cut is being rebuilt. The Fed's preferred inflation gauge is PCE, personal consumption expenditures, which the Bureau of Economic Analysis calculates, and most of us outside the rates desk have never had much reason to think about. Wednesday's report will be the last one prepared under the current formula. The agency is changing how it handles computer software and accessories, since AI demand has done strange things to those prices, along with investment and legal services. BEA revisits its methodology regularly, usually in the fall, so there's nothing improper here. Economists put the effect at a tenth to three tenths off the reported number. PCE sits at 3.7 percent today.
That revision lands about three weeks before the FOMC meets on the 15th and 16th of September, with a new dot plot, in an administration that has been loud about wanting rates lower.
None of it is the number we borrow against.
Commercial real estate doesn't price off the funds rate, and it doesn't price off PCE. It prices off the 10-year and SOFR, plus a spread set by the lender. And while the industry watched the front end, the long end moved in the opposite direction. The thirty-year touched 5.34 percent on August 17th, the highest it's been in nineteen years. The ten-year printed a twenty-month high of 4.75 that same week. Two-thirds of the 392 people Bloomberg surveyed on the 19th expect the ten-year to be above 5 percent by year-end, a level it has barely reached since 2007.
Treasury is now at least doubling its long-maturity buybacks to $4 billion next quarter. That is a lot of institutional effort focused on a single number.
The ten-year is three things stacked on top of each other: what the market thinks short rates will average, what it thinks inflation will average, and the term premium, which is what an investor demands for holding duration through whatever happens next. Term premium is where doubt goes.
Ease the front end while inflation runs near 3.7, and the long end doesn't follow you down. That isn't a forecast. It's a description of the last six weeks. The front end sat still, and the back end sold off.
Whether the revised print reads 3.4 or 3.6 is, for our purposes, a rounding error. The people who set the long end also read the methodology notes.
Run it at the loan level, because that's where this stops being commentary. The Mortgage Bankers Association counted $875 billion of commercial mortgage maturities in 2026, roughly 17 percent of everything outstanding, with another $652 billion behind it in 2027. Most of that is 2019 to 2021 paper. Coupons in the low threes, sized against a ten-year under 2 percent.
A borrower with a 3.25 percent coupon coming due this fall isn't refinancing at 3.25. Today's quotes run from about 5.7 percent on the strongest multifamily credit to the mid-sixes on CMBS, and the older suburban office is wider than that or isn't quoted at all.
The sizing test has changed too. In 2021, the binding constraint was the loan-to-value ratio. In 2026, it's debt yield, net operating income over loan amount, so the lender's question isn't what the building is worth; it's what it earns against the dollars. Both tests get harder when the long end rises.
A quarter-point cut in September would help one group: the floating-rate borrower whose loan is SOFR plus a spread. That's real money, and we shouldn't wave it off. It does nothing for the fixed-rate maturity, which reprices to the ten-year, and the ten-year has spent this summer going the wrong way.
So what changes in how we underwrite?
We stopped modeling exits off the policy rate. Every refinance assumption in every hold gets tested against the ten-year now, and we test it high. Wrong in that direction costs a reserve. Wrong in the other direction costs a capital call.
We're having the amortization conversation with lenders eighteen months out instead of ninety days out. A lender with time to work is a different party than a lender staring down a maturity next quarter.
And on the buy side, the seller whose whole plan was a cheaper 2026 refinance is the seller we want to be talking to. Not because anyone enjoys that. Most of them are capable operators who bought a good building on a rate that didn't hold.
The Bureau of Economic Analysis can change the recipe. The Fed can cut, hold, or hike, and in September it will do one of those in front of a fresh dot plot. None of it moves the maturity date on a loan signed in 2021.
That's the number to put on the wall. Not the inflation print and not the funds rate, but the day the note comes due, and where the ten-year is standing on that day.
We believe we see what the market has already done, and it isn't telling anyone to wait.
— Christopher C. Rising