The small-loan convexity hedge in agency MBS is wearing out

For several decades, the most well-established convexity protection in agency mortgage-backed securities has been the small loan: origination costs are largely fixed, so a small borrower keeps less of the benefit of refinancing into a lower rate and feels correspondingly less pull to move or repay once market rates rise. That indifference is the bondholder's hedge, dampening the faster prepayments and shorter durations that arrive when rates fall and doing the converse when they climb.

Home prices have been eroding that hedge since 2020, as accelerated growth changed the size of the average loan and, with it, how originators choose to pool mortgages and how investors choose to buy the securities built from them, according to Insurance AUM Journal's agency MBS market note for August, written by Tyler Patla. The unit that makes the shift measurable is the maximum loan size in a specified pool: before the pandemic, pools carrying a ceiling above $225,000 barely existed, the ceiling then crept upward, and over the past two years it jumped from $300,000 to $400,000, with loans above each threshold held out of the generic, cheapest-to-deliver pools.

The direction of travel outranks any single bucket: securities defined by a higher ceiling are the ones gaining share of dollar issuance, and because the note's table weights by dollars rather than loan counts, the drift lands where a portfolio would feel it. Convexity protection bought at the pool level is only as good as the loans underneath, and those loans are being repriced by the housing market rather than by anything a portfolio manager decides, so the pools a convexity buyer wants are the ones being issued less. That makes this a supply problem before it is a pricing problem, and it means the hedge is turning into a depreciating feature of the asset class rather than a constant of it.

August was an odd month to be paid for patience, as the Treasury curve bear-flattened after the Treasury announced an initiative to lower long-term yields, then moved higher and flatter again when Fed Chair Warsh told Jackson Hole that the economy was strong and resilient and committed unequivocally to the 2% personal consumption expenditures target. The 30-year Treasury yield touched 5.3%, its highest in 19 years, and after it did, Treasury Secretary Bessent announced a doubling of the size of the Treasury's debt buyback operations for longer-term securities, while the implied probability of a September rate hike went from 38% to 65%.

Mortgages absorbed all of that and tightened anyway, with the Bloomberg US MBS Index returning 0.52% on the month, a 0.23% excess return over Treasuries, and sector option-adjusted spread two basis points narrower at +29. Gains were relatively consistent across lower and higher coupons, and they came despite a soft finish, as the note describes a sector that ground tighter for most of August before weakening in its final days. Investment-grade corporates lagged, the Bloomberg US Corporate Index outearning Treasuries by only 0.12%, and corporate credit is the other place general account dollars go, so the balance-sheet side of the AI buildout, which this publication has written about, is a likely competitor for them.

The prior month's installment had observed that agency MBS tends to underperform in a bear steepening environment, and August made it two consecutive months in which the sector broke from historical precedent. Eight weeks is not a regime change, and a sector that entered the month cheap enough to absorb a backup, or a buyer base that has shifted since the last time the curve moved this way, would both fit the evidence, with neither established. The collateral table is the more durable evidence, and it points one way.

The sector's spread is a thin fee for a position that amounts to a short prepayment option, and the note's own figures show the collateral behind that option becoming more rate-sensitive rather than less. Index eligibility, deep liquidity and a Treasury now doubling its purchases of long bonds are real supports, but none of them restores the small-loan hedge general accounts have been collecting for decades. A second tension sits in the month's policy mix: an initiative designed to pull long-term yields lower works against the insurer trying to lock in a 5.3% thirty-year, the richest long-end level in 19 years; the buyback supports prices and holds down new-money book yield at the same time.

Two numbers to watch from here: the first is the September policy decision, which the note's read of the market put at 65% odds of a hike; a hike slows prepayments, extends duration and delivers the negative convexity of an agency book at the moment a fixed income portfolio can least afford it. The second is the pooling ceiling itself, $225,000 before the pandemic and $400,000 now, and general accounts buying specified pools for convexity should be writing maximum-loan-size limits into their purchase guidelines, because the market has stopped supplying that hedge for free.