Guide

Mortgage-backed securities explained

Harbor Capital's core bond sleeve held $180M of intermediate Treasuries and investment-grade corporates at a 4.6-year effective duration. When the yield curve bull-steepened — long rates fell 65 bp while front-end fed funds held — the portfolio gained on duration but lagged the Bloomberg U.S. MBS Index by 210 bp over twelve months. Treasury asked why. The answer was not credit risk: agency mortgage-backed securities (MBS) carry minimal default exposure when backed by Fannie Mae, Freddie Mac, or Ginnie Mae, yet they embed a different risk — prepayment. Homeowners refinance when rates fall, shortening the cash flows investors thought they owned. MBS prices rise less than duration models predict on rallies and fall harder on selloffs. That negative convexity is the defining feature of the asset class.

An MBS is a bond backed by a pool of residential mortgages. Investors receive pro-rata shares of principal and interest as homeowners pay their loans — or prepay early. The $9+ trillion U.S. agency MBS market is the largest segment of the domestic bond universe and the primary channel through which mortgage rates transmit from capital markets to 30-year fixed quotes. This guide covers securitization mechanics, agency vs non-agency structures, pass-through coupons and weighted averages, prepayment modeling (CPR and PSA), TBA trading, the Harbor Capital mortgage sleeve refactor, a technique decision table against Treasuries and REITs, pitfalls, and a production checklist.

From mortgage to tradeable bond

A bank or non-bank originator lends to a homebuyer, then often sells the loan into a securitization pool. A government-sponsored enterprise (GSE) or Ginnie Mae wraps the pool into a security, guarantees timely payment of principal and interest (agency MBS), and issues certificates to investors. The pipeline looks like this:

Origination — borrower gets a 30-year fixed mortgage at a note rate (e.g. 6.50%).
Pooling — thousands of loans with similar characteristics are grouped; servicers collect monthly payments.
Securitization — the trust issues pass-through certificates; investors own a fractional claim on all cash flows.
Trading — specified pools or TBA (to-be-announced) contracts trade in a deep secondary market, setting mortgage-rate spreads.

Unlike a corporate bond with a fixed maturity, MBS cash flows are uncertain. Every month the pool pays interest on outstanding principal plus scheduled principal amortization plus any prepayments. The investor's principal balance shrinks over time — sometimes faster than expected.

Agency vs non-agency MBS

Type	Guarantor	Credit risk	Typical investor
Agency pass-through	FNMA, FHLMC, GNMA	Minimal (full faith for GNMA; GSE guarantee for FNMA/FHLMC)	Mutual funds, banks, Fed SOMA, insurers
Non-agency (legacy)	None — private-label pre-2008	High — subprime tranches largely written off	Distressed specialists (legacy paper)
Non-QM / jumbo PLS	None — post-crisis private deals	Moderate — overcollateralization and subordination	Specialty credit funds

Most allocators mean agency MBS when they say “MBS.” The spread over Treasuries compensates for prepayment and liquidity complexity, not corporate default. Non-agency structures re-emerged for loans that do not fit GSE boxes (jumbo, investor properties, non-QM credit) but remain a fraction of agency volume.

Pass-through mechanics: coupon, WAC, and WAM

Coupon vs pass-through rate

A pool's mortgages might carry a weighted average coupon (WAC) of 6.75%, but investors receive a lower pass-through coupon — say 6.00% — because servicing fees and GSE guarantee fees are stripped out. MBS are quoted by pass-through coupon and maturity cohort (e.g. “FNMA 6.0% 30-year”).

Weighted average maturity (WAM)

WAM measures the remaining term of loans in the pool, typically near 360 months at issuance. As time passes and prepayments occur, WAM falls. A pool trading at 48 WAM has very different prepayment sensitivity than one at 120 WAM.

Factor and current face

Each month the trustee publishes a factor — the fraction of original face still outstanding (e.g. 0.842). If you bought $1M original face, current face is $842k. Accrued interest and principal payments apply to current face, not the original notional.

Prepayment risk and negative convexity

Homeowners prepay when they sell, refinance into lower rates, or make extra principal payments. For the MBS investor, prepayment is a call option held by the borrower: when rates fall, more loans refinance, returning principal exactly when reinvestment yields are lowest.

CPR and PSA

Constant prepayment rate (CPR) annualizes the percentage of outstanding principal that prepays in a month, compounded. A 6% CPR means roughly half the pool prepays over ten years under stable assumptions. The PSA model (Public Securities Association) ramps CPR from 0.2% in month one to 6% CPR by month 30 and holds flat thereafter — a standardized benchmark, not a forecast.

Negative convexity

A plain Treasury's price rises faster as yields fall (positive convexity). An MBS at a premium coupon (trading above par) loses extension risk on rate rises and call risk on rate cuts. Net result: negative convexity — prices underperform duration-implied gains on rallies and overshoot losses on selloffs. This is why MBS often cheapen vs Treasuries when volatility spikes: investors demand compensation for uncertain cash-flow timing.

Option-adjusted spread (OAS)

Quoting MBS by nominal spread to Treasuries is misleading without a prepayment model. OAS subtracts the value of embedded prepayment options using a Monte Carlo or binomial model, leaving a spread comparable across coupons and vintages. Harbor Capital moved from Z-spread on corporates to OAS on MBS for relative-value screens.

TBA market and dollar rolls

Most agency MBS volume trades to-be-announced (TBA): buyers and sellers agree on agency, coupon, and term, but the exact pool is delivered later. Standardization creates liquidity — the TBA market sets the current coupon (the coupon trading nearest par) that anchors primary mortgage rate sheets.

A dollar roll is a pair of TBA trades — sell for near-month settlement, buy back for forward-month settlement — financing the position. Roll specialness (when demand to hold a coupon exceeds supply) signals scarcity and affects carry calculations. Portfolio managers use rolls to manage settlement exposure without picking specific pools.

Specified pools trade at a pay-up over TBA when they have desirable characteristics: low loan balances (slower prepay), high LTV, geographic concentration, or clean refinance histories. Pay-up analysis separates generic TBA beta from pool-specific convexity bets.

Harbor Capital mortgage sleeve refactor

The refactor shifted 25% of the investment-grade bond allocation from intermediate corporates into agency MBS:

Target — current-coupon FNMA 30-year pass-throughs plus a barbell of discount coupons (4.0%, 4.5%) for extension protection if rates rise.
Prepay model — vendor OAS engine with borrower turnover, refi incentive (note rate minus primary mortgage rate), and burnout from repeated refi waves.
Hedge overlay — partial receive-fixed interest rate swap on $40M notional to offset negative convexity in a +50 bp parallel shock; Treasury futures considered but rejected due to CTD basis vs MBS.
Risk limits — max OAS duration 4.2 years, CPR stress +300% of base on rally scenarios, 5% sleeve cap in specified pools.

Over the next rate-cut cycle, the sleeve captured 85 bp of carry over Treasuries while the swap overlay trimmed 40% of the worst-case selloff tail. The key lesson: MBS allocation is a package deal — yield, prepay model, and convexity hedge must be sized together.

Technique decision table

Approach	Best when	Trade-off
Agency MBS pass-throughs	Yield pickup over Treasuries with minimal credit risk; rate view moderate	Prepayment uncertainty; negative convexity; model-dependent OAS
Treasuries / TIPS	Duration hedge, liquidity, liability matching	Lower yield; no mortgage spread; TIPS add inflation link
Investment-grade corporates	Credit spread income; positive convexity vs MBS	Default and downgrade risk; less rate-sensitive than MBS at same spread
Agency MBS ETFs (MBB, VMBS)	Retail allocation; daily liquidity	Tracking error vs index; embedded fund prepay assumptions
Equity REITs	Real estate equity beta, dividend growth	Stock volatility; different risk factor than bond-like MBS
CMO tranches (IO/PO)	Targeted prepay exposure (IO benefits from slow prepay)	Complex; tranche-specific risks; less liquid
Receive-fixed IRS overlay	Offset MBS negative convexity in rising-rate scenarios	OTC margin; imperfect hedge ratio; basis vs MBS OAS duration

Common pitfalls

Using stated maturity like a bond — 30-year WAM does not mean 30-year duration; effective duration is often 4–6 years and shifts with rates.
Ignoring the coupon stack — premium coupons prepay fastest in rallies; discount coupons extend on selloffs. Mixing them without intent creates unbalanced convexity.
Stale prepayment assumptions — refi waves depend on primary-secondary spread, credit scores, and burnout; 2020-era CPR is not 2026 baseline.
Spread without OAS — comparing a 6.0% MBS Z-spread to a corporate bond spread double-counts the prepay option.
TBA settlement fails — operational breaks in delivery can incur pair-off fees; rolls need financing line capacity.
Over-hedging with Treasuries — MBS spread widening on vol spikes is a separate factor from parallel rate moves.
GNMA vs FNMA/FHLMC confusion — GNMA is full faith and credit; GSEs have different technical factors and pay-up curves.

Production checklist

Define objective: carry income, duration match, or spread RV vs Treasuries.
Choose agency (GNMA, FNMA, FHLMC), coupon stack, and TBA vs specified pools.
Model base, up-100, and down-100 CPR scenarios; compute OAS and effective duration.
Stress negative convexity: price change vs duration-only estimate on ±50 bp.
Size any swap or futures overlay to target residual convexity, not just duration.
Confirm settlement, financing, and roll calendar with custodian and dealer.
Monitor current-coupon spread, primary mortgage rate, and refi incentive weekly.
Reconcile factor, principal, and interest payments against trustee reports.
Review pay-up on specified pools vs TBA; trim pools that lose their thesis.
Document prepay model version and OAS assumptions for audit and attribution.

Key takeaways

Mortgage-backed securities are bonds backed by pools of home loans — investors receive pass-through principal and interest, but cash flows are uncertain because homeowners prepay.
Agency MBS (FNMA, FHLMC, GNMA) carry minimal credit risk; the main risks are prepayment timing and negative convexity, not corporate default.
CPR and PSA model prepayment speed; OAS is the right spread metric because it prices the embedded refi option.
The TBA market provides liquidity and sets mortgage-rate transmission; dollar rolls and specified-pool pay-ups are core manager tools.
Harbor Capital paired a 25% MBS sleeve with a partial receive-fixed swap overlay to harvest spread while capping convexity tail risk.