Mortgage-backed securities: pass-through structure, prepayment risk, and how MBS behave
Mortgage-backed securities (MBS) pass through to investors the principal and interest payments from a pool of underlying mortgages. They are a distinct fixed income asset class with the largest market capitalisation of any non-Treasury debt sector in the United States, and their defining feature is negative convexity arising from the option borrowers hold to prepay their mortgages.
What MBS are
An MBS is created when a pool of mortgages—typically several thousand individual residential or commercial loans with similar characteristics—is securitised. The aggregated cash flows are passed through to securities holders monthly, who receive a pro-rata share of principal and interest collected from borrowers, net of servicing fees. Most of the US market is structured this way, in pass-through form, although tranched structures (CMOs, CMBS) carve the cash flows into different risk profiles.
Agency MBS—issued or guaranteed by Fannie Mae, Freddie Mac, or Ginnie Mae—carry no meaningful credit risk for an investor in the United States, because the agencies guarantee timely payment. Non-agency MBS (also called private-label MBS) lack this guarantee and carry credit risk on top of the prepayment dynamics. The 2007-2008 credit crisis was concentrated in the non-agency segment.
How they work
The defining feature of an MBS is negative convexity. When interest rates fall, mortgage borrowers refinance into lower rates, prepaying the existing mortgage in the process. The MBS holder receives an early principal return precisely when rates have fallen—and must reinvest at the new lower rate. When rates rise, prepayments slow as borrowers cling to their below-market loans, and the MBS holder is locked into a longer duration just as that duration is most painful.
This asymmetric duration response—extending in rising rates, shortening in falling rates—is the opposite of what a portfolio holder typically wants. The result is that MBS investors earn an excess yield over comparable-duration Treasuries to compensate for the negative convexity. The size of that yield premium varies with market expectations about prepayment behaviour, which depend on the level of rates, the rate path, and refinancing friction.
What the evidence shows
Historical agency MBS returns have averaged 50-100 basis points above comparable-duration Treasuries (Fabozzi, 2020), although the realised excess return varies considerably with the rate environment. The extreme rate moves of 2022-2023 produced large prepayment-related losses and gains across the MBS universe.
The 2007-2008 housing crisis demonstrated that non-agency MBS—particularly subprime structures—could produce far larger losses than their ratings implied. The crisis led to substantial reforms in MBS structuring, ratings methodology, and risk retention rules, but the basic dynamics of prepayment-driven negative convexity remain central to the asset class.
Limitations and trade-offs
The duration metric reported on an MBS is necessarily a model output rather than a direct calculation. Different prepayment models can produce materially different effective durations for the same security, and these durations can change rapidly in response to rate moves. Investors evaluating MBS exposure should be aware that the headline duration is a more uncertain quantity than the equivalent number for a Treasury or corporate bond.
For a self-directed investor, the practical implication is that MBS exposure should be sized as a separate fixed income allocation rather than a substitute for similar-duration Treasuries. The yield premium is real but is compensation for a specific structural risk, not free money.
Mortgage-backed securities in pfolio
Mortgage-backed security ETFs are available within pfolio's fixed income universe. The negative convexity that distinguishes MBS from straight bonds is reflected in the price series; the duration metric reported in pfolio Insights captures effective duration where the underlying ETF data supports it.
Related articles
Disclaimer
Get started now
