30-Year Treasury Yield Hits 5.31%, a 19-Year High, While Mortgage Rates Fall
The thirty-year Treasury yield closed at 5.31% on Monday, the highest level since June 2007, after finishing the prior week at 5.26%. The instinctive reading is that the long bond is dragging the housing market back toward the rate shock of late 2023. That reading is wrong, and the way it is wrong describes this cycle better than the headline number does.
Start with what 5.31% meant the last time it printed. In mid-2007 the thirty-year sat within roughly fifteen basis points of the ten-year, the funds rate was above 5%, and the entire curve was clustered around the same handle. The long end was expensive because money was expensive everywhere. Today the ten-year is near 4.69%, the front end is lower still, and the spread between the ten and the thirty has opened to roughly sixty basis points. The level is the same. The shape is not remotely the same. Whatever is happening at 5.31% today is not the market repricing the path of policy. It is the market repricing the cost of holding duration.
Four things are feeding that. Deficit-driven Treasury issuance is the structural one and it does not change on any single data point. Corporate supply is the newer one: AI companies have raised on the order of $1.5 trillion in bonds this year, which is competing directly for the same dollar fixed-income allocation that Treasuries need, and much of it is long-dated because the assets are long-dated. Oil is the third, running through the Hormuz blockade and into every forward inflation series. The fourth is the one that actually explains why the move is concentrated past ten years: Fed Chair Warsh has signalled that a rate hike is not his preferred instrument against higher prices, and University of Michigan year-ahead inflation expectations have now printed above 4% for five consecutive months. A central bank that will tolerate an overshoot rather than fight it with the funds rate is precisely the condition under which a thirty-year lender demands more compensation. The long end is charging for a policy reaction function, not for a policy rate.
Which brings us to housing, where the intuition breaks. Mortgages do not price off the thirty-year bond. They price off the ten-year and, more precisely, off the spread that mortgage-backed securities trade at over it. The thirty-year fixed averaged 6.67% on August 13, down from 6.69% the week before, ending five consecutive weekly increases. It stood at 6.58% a year earlier. In the same stretch that the long bond was setting nineteen-year highs, the mortgage rate went down.
The spread is doing the absorbing. In October 2023, with the ten-year near 5%, the thirty-year fixed reached roughly 7.8%, a gap close to three hundred basis points. Today the same comparison is running near two hundred. Roughly a full percentage point of compression has been handed back to the borrower, and it has been handed back for reasons that survive a rising long end. MBS spreads widen on rate volatility, not on rate level, because the borrower’s prepayment option is worth more when the direction is uncertain. A slow, one-directional drift in the long end against a stable ten-year is close to the least damaging configuration available. A refinance universe sitting deep out of the money makes the negative convexity cheap to carry. Bank and money-manager demand has returned to the sector at yields that finally clear internal hurdles.
So the exposure created by 5.31% falls somewhere other than the household. It falls first on the Treasury itself, where every basis point paid on new long issuance is locked in for three decades and compounds against a debt stock that is still growing. It falls second on the corporate issuers financing the data center build, who are both the cause of the supply pressure and its victim; the marginal AI bond gets priced against a long Treasury curve that the previous AI bonds helped push higher. And it falls, in the opposite direction, as a windfall on anyone carrying long liabilities. Pension funds and life insurers discount at these rates. Funded ratios improve mechanically at 5.3%, and improved funded ratios generate real buying at the long end. That bid is the reason moves of this type have historically found a level rather than running.
The number to watch is therefore not the thirty-year yield. It is the mortgage spread. As long as it keeps compressing into a rising long end, the housing transmission stays disconnected and the observable evidence stays where it currently sits: purchase and refinance applications rising, listing prices modestly below year-ago levels, inventory improving from the scarcity of the last several years, and home price growth running near 1.2% against faster inflation, which is a real decline in the value of the asset. If that spread stops absorbing and begins to widen while the long end continues to rise, housing gets the shock the headline currently implies, and it gets it without the ten-year having to move at all.
For now the long bond is repricing the government’s balance sheet, not the household’s.