Negative Convexity in Agency CMOs Is Quietly Resurfacing

The Quiet Return of a Familiar Risk

Negative convexity never fully disappeared from the agency mortgage market – it just got overshadowed. During the rate volatility of 2022 and 2023, most of the attention went to duration risk and spread widening, while the structural quirks of collateralized mortgage obligations got comparatively little airtime. Now, with rate expectations shifting again and prepayment models starting to reprice, negative convexity in agency CMOs is making itself felt in ways that portfolio managers are only beginning to acknowledge publicly.

The mechanism is straightforward but the timing catches people off guard every cycle. When rates fall, homeowners refinance – and the bonds holding those mortgage cash flows get paid back early, right when the investor least wants to reinvest at lower yields. When rates rise, those same homeowners hold onto their cheap mortgages longer, extending the bond’s duration exactly when the investor wants shorter exposure. That double-edged dynamic is the textbook definition of negative convexity, and it sits at the heart of every agency CMO structure.

Why CMO Structures Amplify the Problem

Plain-vanilla agency mortgage-backed securities carry negative convexity by design. CMOs were originally created to redistribute that risk across different tranches – planned amortization classes absorb the prepayment volatility so that other tranches can behave more predictably. The problem is that redistributing risk does not eliminate it. It concentrates it. The support tranches that absorb prepayment variability can swing dramatically in both price and duration, and in an environment where rate expectations are unstable, those swings become harder to hedge.

Support tranches – sometimes called companion bonds – are particularly exposed right now because their cash flow timing depends almost entirely on what the rest of the deal does. If prepayments come in faster than modeled, support bondholders get repaid quickly with cash they now have to redeploy at lower rates. If prepayments slow, they’re stuck holding longer exposure in a rising-rate environment. There’s no comfortable middle ground, and the current rate environment – where the market is still debating the Fed’s path and mortgage rates remain volatile – is exactly the kind of backdrop that makes support tranche behavior erratic and difficult to price accurately.

The convexity issue is further complicated by the fact that prepayment models are notoriously backward-looking. They’re calibrated on historical behavior, and borrower behavior has shifted. The so-called “lock-in effect” – where millions of homeowners are sitting on sub-4% mortgages they have no intention of refinancing away from – has flattened the prepayment curve at the top. But that effect doesn’t apply uniformly across all loan pools. Newer vintage loans with higher coupons are more responsive to rate moves, and the CMOs built on those pools carry meaningfully different convexity profiles than older, slower-paying pools. The market hasn’t always been clean about distinguishing between them.

Volatility in implied rates compounds all of this. Negative convexity is not just a function of where rates are – it’s a function of how uncertain the path of rates is. Higher implied volatility means a wider distribution of possible prepayment outcomes, which means the optionality embedded in mortgage bonds is more valuable to the borrower and more costly to the investor. Option-adjusted spreads on certain CMO tranches have been quietly widening as a result, a dynamic directly connected to the broader prepayment uncertainty now being priced into agency MBS spreads.

Where Institutional Money Is Sitting

The holders of agency CMOs are not a monolithic group, but a significant portion of the exposure sits inside insurance company portfolios and bank investment books – two categories of investor that manage to accounting-driven constraints rather than purely to market value. That creates a certain structural inertia. Positions that might be trimmed in a mark-to-market portfolio get held through periods of negative convexity because the accounting treatment insulates the holder from feeling the full price impact. The risk doesn’t disappear; it just doesn’t show up where it’s most visible.

Pension funds and some fixed-income mutual funds have also maintained meaningful CMO allocations because the agency guarantee removes credit risk from the equation, which checks the right boxes for liability-matching strategies. The credit quality is real – these are government-sponsored enterprise obligations – but the interest rate risk profile is anything but simple. A bond that extends when rates rise and shortens when rates fall is exactly the wrong shape for a liability-matching investor, and yet the agency label provides enough comfort that the convexity risk sometimes doesn’t get the scrutiny it deserves.

Hedging Convexity Is Getting More Expensive

The practical challenge for active managers who do want to hedge their CMO convexity exposure is that the instruments available to do so have gotten more expensive. Swaptions and interest rate caps – the most common tools for offsetting the embedded short option in mortgage securities – are priced off implied volatility, and implied volatility has been elevated relative to realized volatility for long enough that the carry cost of running a convexity hedge is a meaningful drag on portfolio returns. Some managers are tolerating more unhedged convexity exposure than they would otherwise prefer, simply because the cost of protection is hard to justify against current yield levels.

The dynamic creates a slow-building tension. Portfolios that have been running light on convexity hedges look fine in a stable rate environment. But if rates move sharply in either direction – a Fed pivot, a fresh inflation surprise, a sudden flight to safety – those positions will reprice quickly and in ways that are not linearly related to the rate move. Duration extensions in the 20-30% range on support tranches are not unusual in a significant rate backup, and that kind of non-linearity is what makes negative convexity dangerous in portfolios that are sized for normal behavior.

There’s also a liquidity dimension that gets underappreciated. Agency CMOs, particularly seasoned or complex tranches, are not the most liquid instruments in fixed income. In a stress scenario where many holders want to reduce exposure simultaneously, bid-ask spreads widen and price discovery becomes unreliable. The agency guarantee ensures that settlement will eventually occur at some price, but the path to that price can be painful enough to force realized losses that wouldn’t have appeared on any risk model run during a quiet market. The positions that look manageable in a portfolio report tend to look very different when the bid side disappears.

What the Current Setup Suggests

The current moment is not a crisis, but it has the texture of a slow accumulation. Rate expectations are shifting enough to keep prepayment models in a state of frequent revision. Implied volatility remains high enough that hedging is costly. And the investor base holding agency CMO exposure is, for structural reasons, not particularly reactive to gradual repricing. The conditions that typically precede a sharper dislocation in mortgage market convexity are present in mild form – enough to be worth watching, not enough to be triggering alarms at most institutions.

The tranche-level detail matters more than the headline. Average-life extensions, PSA multiples, and collateral vintage don’t make for clean news coverage, but they’re exactly the variables that separate a manageable CMO position from one that will behave badly in the next 100-basis-point rate move. Portfolios that were built when rate stability seemed like a reasonable working assumption may now be carrying convexity risk that simply wasn’t there when the positions were sized – and that mismatch, quiet as it is, is what makes the current setup worth paying attention to.