Mortgage balances fell $74bn in the second quarter. The total stood at $13.1 trillion at the end of June.
That is a strange thing to happen in a quarter when the 30-year Treasury was reaching a nineteen-year high. Households retiring debt are forgoing other uses of the same cash, and cash has rarely been better paid than it is now. On the face of it, paying down a mortgage into a five-percent world is the wrong trade.
It is not, and the reason is the same lock-in that has frozen the rest of the housing market.
Prepayment is priced against your own loan, not the market
A household deciding whether to overpay compares the guaranteed return of retiring debt — its own mortgage rate — against what it could earn elsewhere.
For a borrower who financed at generational lows, that comparison favours saving. Their loan costs less than a Treasury bill pays. Nobody in that position should be prepaying, and mostly they are not.
The $74bn is therefore not coming from them. It is coming from scheduled amortisation across the whole book, from borrowers who did finance at current rates and for whom prepayment is a five-percent risk-free return, and from balances extinguished on sale.
That composition is the finding. An aggregate that falls while the largest cohort of borrowers has every reason not to prepay is telling you about everybody else.
Lock-in works in both directions
This paper has written about the household that cannot afford to move rather than cannot afford to sell: the constraint is the replacement payment, not the equity.
The same fact produces the balance decline. A mortgage that is never refinanced and never rolled amortises on schedule, quarter after quarter, and nothing resets it. A market with normal turnover replaces retiring principal with new, larger loans and the aggregate grows. Freeze turnover and the aggregate falls automatically, because the only thing still happening is repayment.
So $74bn is less a decision than a consequence. It is what a mortgage market does when it stops originating and keeps amortising.
Why this is a markets story and not a housing one
Because of who owns the other side.
Mortgage principal repaid is cash returned to holders of mortgage-backed securities, and it arrives as prepayment on portfolios that were bought at a premium or a discount depending on the coupon. A slow, scheduled, unusually predictable paydown behaves very differently in a portfolio from the fast refinancing waves those books were modelled on.
Duration on a low-coupon mortgage book extends when rates rise, because the prepayment that would have shortened it does not happen. That is the well-known problem. The less-discussed one is that the paydown still occurring is coming disproportionately from the high-coupon, recently written loans — the ones a holder would most want to keep. The book left behind is longer and cheaper than the one that was bought.
That is the same shape as the corporate side, where debt issued at generational lows is coming due into a market priced very differently. In both cases the cheap paper is the paper that stays.
What to watch
Not the aggregate, which will keep falling for as long as turnover is frozen and will tell you nothing new each quarter.
Watch the share of the decline that comes from scheduled amortisation rather than payoffs at sale. Amortisation is passive and will continue regardless. Payoffs at sale are the housing market restarting, and they are the first thing that would turn the aggregate back up — which means the number going up would be the good news, and it will be reported as households borrowing more.
The $74bn decline in mortgage balances in the second quarter of 2026 and the $13.1 trillion total at the end of June, together with aggregate household delinquency of about 4.7 percent and credit card 90-plus day delinquency of 12.9 percent, are as published by the Federal Reserve Bank of New York in its Household Debt and Credit report for the quarter. The 30-year Treasury yield of about 5.31 percent in mid-August 2026, described as a roughly 19-year high, and the federal funds target range of 3.5 to 3.75 percent are as reported by CNBC and the Federal Reserve during August and September 2026. The account of prepayment incentives and lock-in is standard mortgage economics and the application to this quarter is our own.
Topics marketsmortgageshousinghousehold debtinterest rates
