Mortgage-Backed Securities: Prepayment, Extension, and Convexity Risk

A mortgage-backed security (MBS) represents claims on cash flows from a pool of mortgage loans. In a basic pass-through, the security holder receives a pro-rata share of principal and interest collected from the underlying borrowers, after the deductions specified in the governing documents. Principal includes both the scheduled amortization in monthly payments and unscheduled principal paid ahead of schedule. As principal leaves the pool, the balance that can generate future interest shrinks.

This guide focuses on U.S. residential agency pass-throughs backed mainly by fixed-rate mortgages. “Agency” describes the issuer or guarantor structure, not a promise that the market price cannot fall. Ginnie Mae securities carry a U.S. government full-faith-and-credit guarantee of timely payment; Fannie Mae and Freddie Mac securities have their own guarantees but are not themselves backed by that same full-faith-and-credit pledge. The terms and scope are security-specific. A private-label MBS can have different credit support, and a collateralized mortgage obligation (CMO) reallocates pool cash flows among tranches, so its tranche-level risks differ from a plain pass-through. See the SEC’s MBS and CMO overview.

A conventional fixed-rate mortgage usually permits the borrower to pay the balance off early, subject to the contract and applicable law. A borrower may refinance, sell the home, make extra principal payments, or pay the loan off for another reason. The investor therefore does not own a bond with a fully known principal schedule: the borrower has an option that can shorten the mortgage cash flows. In a pass-through, each early payoff generally returns principal to investors sooner than the scheduled amortization alone would.

A refinance is more likely when the borrower’s potential savings justify the costs and effort. The gap between the old mortgage rate and a new offer matters, but it is not the only input. Closing costs, points, remaining balance, credit, home equity, loan-to-value, occupancy, underwriting, rate-lock timing, and the borrower’s expected time in the home all affect the decision. Home sales and relocations can generate payoffs even when refinancing is unattractive. Loan age, geography, loan size, servicer processes, and borrower characteristics can make two pools with the same coupon prepay differently.

Keep scheduled amortization distinct from unscheduled prepayment, and keep voluntary payoffs distinct from involuntary events such as foreclosure or certain repurchases. Freddie Mac’s daily report, for example, describes its figures as voluntary prepayments and says they exclude scheduled payments, curtailments, and involuntary prepayments; it also warns that its daily calculation may not predict a future monthly factor for a particular pool. A prepayment measure is only meaningful after checking what its dataset counts.

Single monthly mortality (SMM) measures unscheduled principal prepayment as a fraction of the pool balance remaining after that month’s scheduled principal payment. If beginning balance is $100.0 million, scheduled principal is $0.4 million, and unscheduled prepaid principal is $0.6 million, the post-scheduled balance is $99.6 million. SMM is $0.6 million ÷ $99.6 million = 0.006024, or about 0.6024%. The denominator is not the original $100 million, and the $0.4 million scheduled payment is not counted as a prepayment.

Conditional prepayment rate (CPR) expresses that monthly SMM as an annualized conditional rate: CPR = 1 − (1 − SMM)^12. Substituting 0.006024 gives about 6.99%. This is slightly different from multiplying SMM by 12 because it applies the monthly rate to a declining balance. It also assumes the same conditional monthly speed continues for twelve months. The result is an annualized measure, not a promise that 6.99% of this pool will actually prepay over the next year. FHFA’s Investment Portfolio Management module shows this calculation and explains why prepayment assumptions shape modeled MBS cash flows and yields.

A published prepayment-speed series may not capture every change in a pool’s balance. Freddie Mac’s Daily Prepayment Report defines its SMM and CPR as voluntary prepayments and excludes scheduled amortization, curtailments, and involuntary payoffs. Match that scope before comparing it with another factor series. Daily readings also do not predict the next monthly factor for a particular security; they describe only the component covered by the report’s definition.

When market mortgage rates fall below the rates on existing loans, more borrowers may find refinancing worthwhile. Faster prepayments return principal earlier, shorten modeled average life, and reduce future interest collected on the shrinking pool. This shortening is called contraction risk. An investor may have to reinvest the returned cash at lower rates just when yields on newly issued mortgages and other bonds have declined.

The effect on a premium-priced security can be especially important. If an investor paid more than par, receiving principal faster can return that principal at par before the premium has been earned back through coupon payments; the realized yield can be lower than a slower-prepayment scenario implied. Conversely, faster prepayment can help a discount-priced security recover its discount sooner, but that does not guarantee a higher total return: reinvestment rates, purchase price, coupon, fees, and the full cash-flow path still matter. Freddie Mac’s 2026 UMBS and MBS offering circular explicitly ties yield to purchase price, coupon, and prepayment assumptions.

FHFA’s training example makes the timing effect concrete. For a hypothetical newly originated 30-year pool with a 6.5% mortgage rate, its table shows 59% of the starting balance still outstanding after seven years under a 6% CPR assumption, compared with 23% under 18% CPR. The corresponding modeled WALs are 10.7 and 4.7 years. These are outputs from that example’s assumptions, not current market observations or a forecast for a named MBS.

When new mortgage rates rise above the rates on existing loans, refinancing usually becomes less attractive. Borrowers may keep the below-market loan, so principal arrives more slowly than a faster-speed scenario assumed. The expected life of the MBS can extend just when market yields are higher. This is extension risk: the investor remains exposed to the older, lower-coupon cash flows for longer while the discount rate applied by the market has risen.

Extension changes interest-rate sensitivity as well as the calendar timing of principal. A duration estimate based on one prepayment path can understate how much a security’s price may respond if higher rates slow prepayments. A bond-like illustration with a fixed schedule misses that feedback: the market-rate move changes borrower incentives, borrower behavior changes projected cash flows, and the revised cash flows change the security’s value and duration. The size of the effect depends on the coupon and price relative to current rates, loan characteristics, and the model used.

For an option-free bond with fixed cash flows, price usually rises more as yield falls than it falls for an equal-sized yield increase, after allowing for positive convexity. A mortgage borrower’s prepayment option can bend that response the other way over some rate ranges. Falling rates can accelerate principal return and limit price appreciation; rising rates can slow principal return and prolong lower-coupon cash flows. That rate-dependent shortening and extension is why residential mortgage assets are often described as having negative convexity.

“Negative convexity” does not mean the security has a negative coupon, that its price must fall, or that every MBS behaves alike. The response can be small or large depending on whether borrowers have a meaningful refinance incentive, the pool’s coupon, its purchase price, loan age and composition, and assumptions about rates and volatility. FHFA’s interest-rate-risk material describes the embedded borrower prepayment option and its contraction/extension channel. The bond convexity guide explains positive convexity for option-free bonds and why embedded options require a different model.

Weighted-average life (WAL) is the principal-weighted average time until principal is returned. In a simplified hypothetical schedule, suppose $25 million of principal is returned at year 1 and $75 million at year 3. WAL = ($25 million × 1 + $75 million × 3) ÷ $100 million = 2.5 years. WAL weights principal amounts by their payment time; it does not discount every interest and principal cash flow to present value. A mortgage pool’s WAL therefore depends on the projected prepayment path.

Effective duration estimates price sensitivity to a yield change while allowing the model’s expected cash flows to change as rates move. It is calculated by repricing under upward and downward rate shocks, not by taking a simple average date. CPR is an annualized prepayment-speed measure; SMM is its monthly building block; WAL is an average principal-return time; effective duration is a modeled price sensitivity; contractual final maturity is a legal term in the underlying loan documents. They are related inputs and outputs, not interchangeable labels.

Option-adjusted spread (OAS) is also model-dependent because the valuation must account for rate paths and borrower options. FHFA’s research on prepayment uncertainty and OAS notes that prepayment shocks capture only part of model error and that mortgage-security OAS can also reflect forecasting errors, credit, and liquidity risk. A model’s precise-looking duration, WAL, or OAS should be read with its date, assumptions, and scenario set.

Start with the security type and governing documents: agency or private-label, pass-through or CMO tranche, fixed-rate or adjustable-rate collateral, guarantee terms, payment delay, servicing deductions, and any special pool features. For a pass-through, compare the current principal factor, weighted-average coupon, loan age, remaining term, balance distribution, geography, and borrower or property characteristics where disclosed. A security coupon is not necessarily the same as the mortgages’ weighted-average note rate because servicing and guarantee fees may be deducted.

Then inspect the prepayment assumptions behind quoted yield, WAL, and duration. Ask whether CPR is historical, current, or a modeled scenario; whether the data includes curtailments or involuntary payoffs; and what happens under faster and slower speeds alongside parallel and nonparallel rate moves. Check the price paid, premium or discount, accrued interest, liquidity, and whether a quote is for a specific pool or a broader cohort. Do not compare model outputs that use different cash-flow conventions as though they were observed facts.

Finally, separate borrower behavior from guarantees and credit exposure. A payment guarantee does not remove market-price, liquidity, reinvestment, prepayment, or model risk. Private-label securities and CMO tranches can add credit support waterfalls, subordination, or principal-allocation rules that a plain pass-through does not have. This article covers U.S. residential agency pass-through mechanics; it is not a security recommendation, a forecast, or a description of every country’s mortgage market. For related rate sensitivity, see the bond ETF duration guide.