Skip to main content
Common Bond Mistakes

Mortgage Bonds Without Prepayment Understanding

Pomegra Learn

Mortgage Bonds Without Prepayment Understanding

Mortgage-backed securities (MBS) promise stable cash flows, but they contain a hidden feature: when rates fall, homeowners refinance, capping your upside. When rates rise, refinancings stop, locking you in lower coupon. This negative convexity turns what looks like a safe bond into a portfolio trap.

Key takeaways​

  • Negative convexity in MBS means principal appreciation is capped when rates fall but principal losses are amplified when rates rise; you get asymmetric downside risk.
  • In the 2021–2023 rate-hiking cycle, the Mortgage Backed Securities Index fell over 20% as MBS holders endured both rate duration losses and prepayment option losses.
  • Prepayment risk was painfully visible in 2022: when 30-year mortgage rates jumped from 2.7% to 6.0%, homeowners stopped refinancing, and MBS investors discovered their "duration" was longer than they thought.
  • MBS underperformed Treasury bonds by 8–10 percentage points annually during the 2022–2023 rate cycle, a textbook case of negative convexity in action.
  • Most retail investors should avoid direct MBS exposure; BND and AGG have modest MBS allocations (20–30%) which is defensible, but buying MBS funds for yield is a mistake.

The mortgage refinancing game​

To understand the MBS trap, start with a simple fact: homeowners have embedded optionality in their mortgages. When mortgage rates fall, refinancing becomes attractive. A homeowner with a 5% 30-year mortgage can refinance at 3% and pocket the savings. The bank (or MBS investor) holding the original 5% coupon is out of luck; they get their principal back early, forced to reinvest in a 3% world. This is prepayment risk.

Prepayment risk creates a non-linear payoff for MBS investors. When rates fall 100 basis points, Treasury bonds gain roughly 8–10 percentage points (duration math); MBS gain only 4–5% because half the bond gets prepaid and returned to you at par. You miss half the capital appreciation. This is negative convexity—the investor gets hurt asymmetrically.

Conversely, when rates rise, the MBS duration unexpectedly extends. Homeowners stop refinancing, so the bond stretches from an expected 5-year life to a 7-year or 8-year life. The investor is locked into an underwater coupon and faces a 20%+ capital loss (from rate duration) while Treasuries of equivalent duration face smaller losses.

In 2022, this played out brutally. When the Fed started hiking in March 2022, mortgage rates rose from 2.7% to 6.0% by year-end. MBS investors who thought they were holding 5–6 year duration bonds discovered they were holding 8–10 year duration. The combination of duration losses plus the collapse of the prepayment option made MBS one of the worst-performing bond segments. BND (which holds ~20% MBS) fell 16%, but MBS-heavy portfolios fell 20–25%.

The asymmetric payoff structure​

A simple example clarifies the trap. Suppose you buy a $1M pool of 30-year mortgages at 5% coupon. You expect to get your principal back over 30 years with annual coupon payments.

Scenario 1: Rates fall to 3%.

  • Homeowners refinance immediately. You get your $1M back in year 1.
  • You're forced to reinvest at 3%, losing the 5% coupon you expected.
  • Your total return is capped. If rates fall 200 basis points on a standard bond, you'd gain 15%+; on MBS, you gain 5–8% because of prepayment.

Scenario 2: Rates rise to 7%.

  • Homeowners hold their mortgages. You're locked into 5% coupon in a 7% world.
  • The bond stretches to full 30-year maturity; duration doubles.
  • Your capital loss is amplified: a 200-basis-point rate rise causes a 15–20% loss on the stretched MBS.

Scenario 3: Rates stay at 5%.

  • Homeowners slowly refinance or pay down mortgages over 30 years.
  • You collect your 5% coupon, and it's acceptable but not great.

Notice the asymmetry: in scenario 1, prepayment costs you; in scenario 2, lack of prepayment costs you more. The payoff is convex against your interests.

Historical precedent: 2022–2023​

The 2022–2023 hiking cycle created a textbook MBS disaster. Here's what happened:

In January 2022, 30-year mortgage rates were 3.2%; the Fed started raising rates in March 2022, and by December, mortgage rates hit 7.1%. This 390-basis-point move is the fastest mortgage-rate rise since the 1980s.

MBS pools issued in 2021–2022 (when rates were 2.7–3.5%) suddenly found themselves underwater. A 3% mortgage coupon in a 7% rate environment is a disaster. Homeowners can't refinance (no incentive), so they hold the mortgage. From the MBS investor's perspective, the bond duration extended from an expected 6 years to 10+ years, and the principal was locked at a 4% loss (relative to current market rates).

The mortgage-backed securities index (FNMX, or VGSLX in Vanguard's MBS index) fell 22% in 2022. During the same period, the 7–10 year Treasury index fell 15–17%, and the total bond market (BND) fell 14.6%. MBS underperformance was severe because of negative convexity.

Investors who believed they had duration protection in MBS—a "safer" bond investment—found their worst losses came from that "safe" segment. Many rebalancing portfolios sold MBS to maintain allocations, realizing losses at the worst time.

Prepayment models and their failures​

To manage prepayment risk, investors use prepayment models—mathematical functions that estimate refinancing rates based on current mortgage rates, home prices, economic conditions, and borrower behavior. The standard industry model is the Public Securities Association (PSA) model, which assumes a certain percentage of mortgages prepay each month.

These models are unreliable, especially in extreme rate environments. In early 2020, when the Fed cut rates to zero and mortgage rates fell to 2.7%, prepayment models massively underestimated refinancing waves. Mortgage pools that models assumed would pay off over 8–10 years actually paid off in 3–4 years. MBS investors who relied on models to estimate duration and convexity got blindsided.

Conversely, in 2022–2023, when rates rose rapidly, models underestimated how long homeowners would hold mortgages. Pools that models said would stretch to 10–12 years instead stretched to 15+ years or full 30-year lives. Some borrowers are still holding 3% mortgages in 2026, refusing to refinance despite slightly higher rates, because of lock-in psychology and transaction costs.

The lesson: prepayment models fail in the environments where you need them most (crisis, crisis off-ramps, and extraordinary rate moves). Don't assume models are reliable.

Why MBS don't belong in retail portfolios​

For institutional investors managing billions, MBS have a role: they offer yield slightly above Treasuries (typically 50–100 basis points) and they're liquid. The additional yield compensates for the prepayment headaches if you can manage dozens of pools with sophisticated modeling.

For retail investors, MBS exposure should come through broad bond funds (BND, AGG, VBTLX), not through dedicated MBS funds or direct MBS ownership. Here's why:

  1. Negative convexity is a retail killer. Retail investors are trend-followers; they tend to buy MBS when yields are elevated (after a crisis, when rates have risen and yields are attractive). Then, when the Fed cuts rates back down (historically 6–18 months later), refinancing waves hit, and retail investors get whipsawed. An MBS fund marketing a "5.5% yield" in late 2022 was precisely when the 5.5% was about to vanish due to refinancing.

  2. Pricing complexity is high. MBS prices embed option values that change daily based on rate expectations. A retail investor buying an MBS fund is betting on their prepayment model's assumptions about future rates. Professional MBS traders spend careers refining these models; retail investors shouldn't try.

  3. Better alternatives exist. Treasury bonds, high-quality corporate bonds, and short-duration bond funds offer similar yields with simpler risk profiles and no prepayment option. A ladder of Treasury ETFs (SHV for under 1 year, IEF for 7–10 years, TLT for 20+ years) gives you control and clarity. Mortgage pooling is the opposite.

  4. Illiquidity during stress. While MBS are nominally liquid, bid-ask spreads widen substantially in crises. In March 2020, MBS spreads widened to 10–15 basis points; in 2022, even wider. If you need to exit, you'll pay the cost.

The role of MBS in broad bond funds​

BND, AGG, VBTLX, and similar total-bond-market funds hold 20–25% MBS because MBS are a large part of the tradable US bond market. These funds are appropriate for most investors. The MBS allocation is diversified, passively managed, and rebalanced systematically.

The key difference: you're not buying MBS for their yield. You're getting a small MBS exposure as part of a diversified bond portfolio. The fund manager rebalances regularly, avoiding concentration in underwater coupon pools. The MBS are there because they're part of the market-cap-weighted bond index, not because someone decided "I want 25% MBS."

Dedicated MBS funds (GNMA, VGSLX, VMBS) should be avoided. These funds concentrate the prepayment risk and are vulnerable to the cycles described above.

Process: audit your MBS exposure​

If you hold bond funds, here's how to check your MBS exposure:

  1. List your bond holdings. Include fund names (BND, AGG, VBTLX, LQD, etc.) and ETFs.
  2. Look up the fund fact sheet. Vanguard, iShares, and Schwab publish prospectuses showing asset allocation.
  3. Check the MBS weight. If it's 20–25% (part of a total-bond index), you're fine. If a fund is labeled "MBS" or "mortgage-backed" and is >50%, it's too concentrated.
  4. Review recent performance. If your bond fund underperformed Treasury ETFs by 3–5% in 2022–2023, MBS drag was likely the culprit.
  5. Decide on changes. If you want clearer control, replace MBS-heavy funds with Treasury ladders or corporate-bond funds.

Flowchart: should you hold MBS?​

Next​

You've now seen how prepayment optionality turns mortgage bonds into negative-convexity traps. Next, we'll explore a tax-specific mistake: buying zero-coupon bonds (or STRIPS) without understanding the phantom-income implications in taxable accounts.