What Makes MBS Different: Prepayment Risk, Negative Convexity, and the Limits of Static Hedging

Introduction

MBS may look like any other fixed-income instrument, but the risk structure underneath is fundamentally different. At the center of that difference is the prepayment option — the right of a mortgage borrower to refinance at any time, at no cost to them. That option is held by the borrower and granted, without compensation, by the investor.

Prepayment exists across structured products, but the mechanism differs. In ABS, it is idiosyncratic and averages out across large pools. In CLOs, a revolving structure allows the manager to reinvest, shielding investors from direct cash flow shortening. In MBS, when rates fall, millions of borrowers gain refinancing incentives simultaneously — the prepayment is systemic, rate-driven, and fully absorbed by the investor.

This generates four distinct risk characteristics: the prepayment S-curve, negative convexity, duration drift, and OAS. Each is examined below.

Prepayment S-curve

CPR measures the annualized rate of prepayment across a mortgage pool. In practice, it is derived from SMM (Single Monthly Mortality), the monthly prepayment rate, and annualized as follows:

SMM = Prepaid Balance / (Beginning Balance — Scheduled Principal)

CPR = 1 — (1 — SMM)¹²

The primary driver of CPR is the refinancing incentive — the spread between a borrower’s current coupon and prevailing market rates. As market rates fall below the borrower’s coupon, the incentive to refinance increases, and CPR rises. Plotting this relationship against rate levels produces the characteristic S-curve: CPR remains low when rates are high, accelerates sharply as rates approach the refinancing threshold, and flattens again once the refinanceable population is largely exhausted.

The chart plots CPR against the refinancing incentive. Three zones define the curve.

Where market rates sit above the borrower’s coupon, there is no refinancing incentive and CPR remains low. As rates approach the coupon level, CPR accelerates sharply. This is the most sensitive region and where prepayment models are most vulnerable to error. Once the incentive turns sufficiently positive, the pool has largely refinanced and CPR flattens. This is pool burnout.

The threshold level and steepness of the curve are not fixed. They shift with the macro environment, housing prices, and borrower credit quality. A model calibrated to one rate cycle may significantly misprice prepayment risk in the next. This instability is the root cause of the risk characteristics examined in the sections that follow.

To illustrate, consider a hypothetical $500 million pool of 30-year fixed-rate mortgages originated in early 2020 with a 3.75% coupon. The three points below map how CPR shifted as market rates moved across the refinancing threshold between 2020 and 2022.

The same $500 million pool produces vastly different cash flow profiles depending on where rates sit relative to the coupon.

• Point A (rates 333 bps above coupon): monthly prepayments of $418,483. Minimal refinancing activity.
• Point B (rates at coupon): monthly prepayments jump to $8,192,800. The pool sits at the most sensitive region of the curve.
• Point C (rates 110 bps below coupon): monthly prepayments reach $14,734,400, a 35x increase from point A. The investor receives principal back at a rate they almost certainly did not model at origination, in an environment where reinvestment options are limited.

Negative Convexity

A standard bond exhibits positive convexity: as rates fall, prices rise at an accelerating rate. MBS behaves differently.

When rates fall, prepayments accelerate and investors receive principal back in a low-rate environment, capping price appreciation. When rates rise, prepayments slow, extending duration and amplifying price declines. In both directions, MBS underperforms a comparable straight bond. This is negative convexity.

The implication for risk management is that the hedge ratio cannot be fixed. A standard bond’s price response to rate changes is predictable. In MBS, the magnitude and direction of that response shifts depending on where rates move. This makes static hedging insufficient and requires continuous recalibration as the rate environment changes.

The chart shows three lines. The straight black line is the duration approximation, a linear estimate of how price responds to yield changes. The orange curve is the actual price-yield relationship of a standard bond, which bows above the duration line, reflecting positive convexity. The red dashed line is the MBS price-yield relationship, which bends below the duration line as yields fall.

This divergence is negative convexity. When rates decline, MBS price appreciation is capped relative to a standard bond. When rates rise, the price decline is amplified. In both directions, the MBS investor is on the wrong side of the curve.

The following uses hypothetical figures to illustrate the mechanism. Actual convexity values vary by rate environment, coupon, and vintage.

When market rates fall, a standard bond benefits fully from the lower discount rate: future cash flows are worth more, and price rises. In MBS, when rates fall below the coupon, prepayments accelerate. Borrowers refinance into cheaper loans, returning principal to investors in a low-rate environment where reinvestment options are limited. The price appreciation that a lower discount rate should generate is offset by the loss of future cash flows. This is the mechanism behind negative convexity.

The price change for a given rate shock is approximated as:

ΔP = (-D × Δy × P) + (0.5 × C × Δy² × P)

Where D = 7, P = $100M, C = +80 (Standard Bond), C = -40 (MBS).

Consider a $100M MBS position with duration of 7 years and a coupon of 3.75%. The table below shows how the convexity gap grows as the rate shock increases.

• At -100bps, the gap is $600,000
• At -200bps, the gap widens to $2,400,000
• At -300bps, it reaches $5,400,000, nine times larger

The gap compounds because convexity is a second-order effect: it scales with the square of the rate move. This is why negative convexity becomes increasingly costly in large rate moves, precisely the environment where it matters most.

The practical implication is that hedging an MBS position requires continuous recalibration. Unlike a standard bond where duration is relatively stable, the duration of an MBS shifts as rates move, making static hedges unreliable. This is the problem examined in the next section.

Duration Drift

Duration measures a bond’s price sensitivity to changes in interest rates. For a standard bond, duration is relatively stable. For MBS, it is not.

• Extension risk: When rates rise, prepayments slow and effective maturity lengthens. Duration increases, exposing the portfolio to more interest rate risk than anticipated.
• Contraction risk: When rates fall, prepayments accelerate and effective maturity shortens. Duration decreases, leaving the portfolio underexposed to the rate rally.

In both cases, the hedge breaks down. When rates rise, duration extends beyond the hedge target and losses exceed expectations. When rates fall, duration contracts and the portfolio captures less of the rally than intended. The direction differs, but the outcome is the same: the MBS investor is always on the wrong side of the move.

This is why duration drift makes static hedging unreliable in MBS. The hedge ratio must be continuously updated as rates move, adding both complexity and cost to risk management.

The chart plots effective duration against rate level for an MBS and a standard bond. The standard bond maintains a stable duration regardless of where rates move. The MBS duration rises and falls with the rate environment.

• When rates rise: MBS duration extends beyond the hedge target, leaving the position more exposed to rate risk than intended. Losses exceed expectations.
• When rates fall: MBS duration contracts below the hedge target, and the portfolio captures less of the rate rally than expected.

The shaded regions show the gap between where duration is and where the hedge assumes it to be. This gap is the cost of duration drift. It cannot be eliminated. It can only be managed through continuous rehedging.

The following uses hypothetical figures to illustrate the mechanism.

The reference metric for hedging is DV01 (Dollar Value of 01) — the change in portfolio value for a 1bp move in rates.

DV01 = Modified Duration × Portfolio Value × 0.0001

At hedge inception, a Treasury position of equal and opposite DV01 is established. The problem is that MBS duration shifts as rates move. When duration changes, DV01 changes, and the hedge set at inception becomes immediately misaligned.

Consider a $100M MBS position with an initial duration of 7 years, using the 3.75% coupon pool from the S-curve example.

• +200bps shock: refinancing incentive disappears, prepayments slow, duration extends to 11 years
• -200bps shock: refinancing incentive deepens, prepayments accelerate, duration compresses to 4 years

At hedge inception: DV01 = 7 × $100,000,000 × 0.0001 = $70,000

After +200bps shock (duration extends to 11 years): New DV01 = 11 × $100,000,000 × 0.0001 = $110,000 Hedge Gap = $110,000 — $70,000 = +$40,000/bp (under-hedged)

After -200bps shock (duration contracts to 4 years): New DV01 = 4 × $100,000,000 × 0.0001 = $40,000 Hedge Gap = $40,000 — $70,000 = -$30,000/bp (over-hedged)

In the first case, the portfolio is under-hedged by $40,000 per basis point. In the second, it is over-hedged by $30,000 per basis point. In both directions, the hedge breaks down. This is the cost of duration drift.

This dynamic feeds directly into how MBS is priced in the market, which is where OAS comes in.

Option-Adjusted Spread (OAS)

Option-Adjusted Spread (OAS) is the spread remaining after stripping out the value of the embedded prepayment option. It is not a simple yield spread. It is the pure credit and liquidity premium an investor demands to hold MBS.

OAS=Yield Spread−Option Cost

The option cost rises with interest rate volatility. As volatility increases, the prepayment option becomes more valuable to the borrower, and investors demand a higher OAS in return. This makes OAS a signal of the rate volatility regime, not just a credit measure.

Data: 30-Year Fixed Rate Mortgage Average (MORTGAGE30US) and 10-Year Treasury Constant Maturity Rate (DGS10), Federal Reserve Bank of St. Louis (FRED)

True OAS requires proprietary modeling data. This analysis uses the mortgage-treasury spread as a proxy. It runs wider than true OAS, but the directional pattern holds: when rate volatility rises, the spread widens.

The bottom panel shows the spread across four rate regimes, against a 2019–2026 mean of 2.23%.

• Pre-COVID: Mean spread of 1.83%, range 1.57% to 2.15%. Low rate volatility, stable market conditions. The option premium demanded by investors was modest.
• COVID / Refi Boom: Mean of 1.86% masks significant dispersion. The spread initially spiked before compressing to 1.28% as the refinancing boom took hold. With the prepayment option deep in-the-money, investors required less additional premium.
• Fed Tightening: Mean spread of 2.62%, peaking at 3.26%. That is 43% wider than the Pre-COVID mean. Rate volatility surged, prepayment uncertainty increased, and investors demanded a substantially higher premium to hold MBS.
• High Rate Plateau: Mean of 2.34%, narrowing but still 51bps above the Pre-COVID mean. The market has not fully repriced back to a low-volatility regime.

OAS is not a static measure. It compresses when volatility is low and widens when volatility spikes. For a risk manager, a widening OAS is not just a valuation signal. It is a sign that the market is repricing the cost of the embedded prepayment option in real time.

Closing Takeaway

The four characteristics examined in this piece are not independent. Prepayment behavior drives negative convexity. Negative convexity causes duration drift. The cost of duration drift is reflected in OAS.

The chain has direct implications for risk management:

• If the prepayment model is wrong, the entire cash flow projection breaks down. A CPR miscalibration of the kind shown in the S-curve example can result in monthly prepayments that are 35x higher than modeled.
• If negative convexity is ignored, the hedge ratio is miscalibrated from the start. A -300bps move produces a $5,400,000 gap that a static hedge cannot absorb.
• If duration drift is not tracked, the hedge becomes unreliable the moment rates move. A +200bps shock can leave a $100M portfolio under-hedged by $40,000 per basis point.
• If OAS is misread, the portfolio misses the signal that the market is already repricing volatility risk. The 43% spread widening seen during Fed tightening was visible in real time.

Static position management is insufficient. Each of these failure modes is not hypothetical. They showed up in the data.

📩 Join Investor’s Handbook Digest — get the best investing, markets, and wealth-building insights each week.