Agency vs. Non-Agency RMBS Differences

Equicurious TeamintermediatePublished: 2025-11-02Updated: 2026-02-19
Illustration for: Agency vs. Non-Agency RMBS Differences. I now have all the research and style context I need. Let me compose all four ar...

Agency vs. Non-Agency RMBS: The Credit-Risk Trade That Defines Your Fixed-Income Allocation (Why It Matters)

If you own a diversified bond fund, roughly 25-30% of your fixed-income exposure sits in residential mortgage-backed securities. Whether those bonds carry a government guarantee or not changes everything: the risk you bear, the yield you earn, and how your portfolio behaves when markets crack. The agency MBS market alone stands at approximately $9.2 trillion in outstanding securities (SIFMA, 2024), making it the second-largest fixed-income market after U.S. Treasuries. The non-agency segment adds another $1.7 trillion (Janus Henderson, 2024). Understanding the line between these two markets is not optional for anyone running a serious bond portfolio.

The point is: agency and non-agency RMBS look superficially similar (both are pools of home loans), but they expose you to entirely different risk factors. Agency MBS = prepayment and interest-rate risk. Non-agency RMBS = credit risk and structural complexity. Getting this distinction wrong means you are either giving up yield you should be earning or taking credit risk you do not understand.

The Agency Side: Government-Backed, Prepayment-Exposed (Core Mechanics)

Agency MBS are issued or guaranteed by three entities: Ginnie Mae (explicitly backed by the U.S. government), Fannie Mae, and Freddie Mac (government-sponsored enterprises with an implied—and, since 2008, effectively explicit—guarantee). When a borrower defaults on an agency-backed mortgage, the guarantor absorbs the loss. You, the bondholder, get paid regardless.

This guarantee eliminates credit risk but does not eliminate all risk. What remains is prepayment risk—the uncertainty of when (not if) borrowers will return your principal.

How Prepayment Risk Actually Works

Consider a concrete scenario. You buy a Fannie Mae 30-year pass-through with a 6.0% coupon at par. Six months later, mortgage rates fall to 5.0%. Borrowers refinance aggressively:

  • Your bond's weighted-average life (WAL) drops from ~7 years to ~3 years
  • You receive principal back at par when your bond was priced for a longer duration
  • You must reinvest that principal at lower prevailing rates

The reverse is equally damaging. Rates rise to 7.5%, nobody refinances, and your bond's WAL extends from 7 years to 12+ years. You are now stuck holding a below-market coupon for far longer than planned.

Why this matters: agency MBS investors are short a call option to the borrower. Higher rates → extension risk. Lower rates → contraction risk. This "negative convexity" is the price you pay for the government guarantee.

Conforming Loan Requirements

Agency loans must meet specific criteria to receive the guarantee:

Criterion2024 Limit2025 Limit
Baseline conforming loan limit$766,550$806,500
High-cost area limit$1,149,825$1,209,750
Minimum down payment (conventional)3%3%
Maximum DTI ratio45-50%45-50%
Credit score floor (typical)620+620+

Sources: FHFA (2024); Fannie Mae Lender Letter LL-2024-03.

Any loan exceeding these limits (or failing other criteria) cannot be guaranteed by the GSEs. That loan either stays on a bank's balance sheet or enters the non-agency market.

The Non-Agency Side: Credit Risk, Higher Yield, More Complexity (Why Investors Cross Over)

Non-agency RMBS are securitized by private entities—banks, specialty finance companies, mortgage originators—without a government guarantee. If borrowers default and recoveries fall short, bondholders absorb losses according to the deal's waterfall structure.

The durable lesson: you earn a credit spread above agency MBS (and above Treasuries) precisely because you bear default risk. That spread has historically ranged from 75 to 400+ basis points depending on the tranche, collateral quality, and market conditions.

Non-Agency Sub-Sectors

The non-agency market is not monolithic. Four distinct sub-types dominate:

1. Prime Jumbo — High-balance loans to creditworthy borrowers exceeding GSE limits. Borrower profiles resemble agency loans (high FICO, low LTV), but the loan amount disqualifies them. Default rates historically track close to agency levels.

2. Non-QM (Non-Qualified Mortgage) — Loans to borrowers who fail one or more Qualified Mortgage criteria: bank-statement income verification (self-employed borrowers), interest-only payment structures, DTI ratios above 43%. These are not subprime by definition, but they carry incrementally higher risk.

3. Re-performing / Non-performing (RPL/NPL) — Previously delinquent loans that have been modified or cured. Purchased at a discount, these pools offer attractive yields but require intensive collateral analysis. The seasoning of the loans (how long since modification) matters enormously.

4. Credit Risk Transfer (CRT) — A hybrid. Fannie Mae and Freddie Mac issue CRT securities (since 2013) to shift credit risk on otherwise conforming loans to private investors. Over $2.1 trillion in CRT deals have been issued to date (Alliance Bernstein, 2024). You get agency-quality collateral with non-agency-style credit exposure. The structure issues only mezzanine and subordinate bonds (M1, M2, B1, B2)—no senior tranche.

Structural Protections in Non-Agency Deals

Because there is no government guarantee, non-agency deals use credit enhancement to protect senior bondholders:

  • Subordination: Junior tranches absorb losses before senior tranches. A typical post-crisis deal might have 20-30% subordination at the AAA level.
  • Overcollateralization: The loan pool's face value exceeds the bond issuance (e.g., $105 million in loans backing $100 million in bonds).
  • Excess spread: The difference between the weighted-average coupon on the loan pool and the weighted-average coupon paid to bondholders creates a cash cushion.
  • Reserve funds: Cash set aside at closing to cover shortfalls.

The test: when evaluating a non-agency deal, your first question should be "how much can default rates rise before my tranche takes a loss?" If the answer is less than 2x the base-case assumption, the credit enhancement is thin.

Side-by-Side Comparison: The Decision Matrix

FactorAgency MBSNon-Agency RMBS
Credit riskNone (government guarantee)Yes (bondholder bears losses)
Primary riskPrepayment / extensionCredit default + prepayment
Typical yield spread over Treasuries50-150 bps125-500+ bps
LiquidityExtremely high (TBA market)Moderate to low (specified pools)
CollateralConforming loans onlyJumbo, non-QM, RPL/NPL, CRT
ComplexityLow-moderateHigh (structural analysis required)
Negative convexitySignificantLess (many loans are less refinanceable)
Market size~$9.2 trillion~$1.7 trillion
Active management valueModerateHigh

A Worked Example: Yield Comparison on Equivalent Duration

Suppose you are building a 5-year duration allocation. You compare:

  • Agency MBS (Fannie 5.5% coupon): Priced at 100.5, yield of approximately 5.35%, spread of ~90 bps over the 5-year Treasury.
  • Non-agency AAA (prime jumbo, 2024 vintage): Yield of approximately 5.90%, spread of ~145 bps over the 5-year Treasury.
  • Non-agency BBB (non-QM, 2024 vintage): Yield of approximately 7.25%, spread of ~280 bps over the 5-year Treasury.

The incremental 55 bps from agency to non-agency AAA compensates you for (a) illiquidity, (b) structural complexity, and (c) a small but nonzero credit tail risk. The jump from AAA to BBB adds another 135 bps for meaningfully higher credit exposure.

Why this matters: in late 2024, spread compression brought non-agency AAA spreads to near-historic tights (MSCI, 2024). When spreads are tight, you are being paid less for the incremental risk—and that is exactly when discipline matters most.

The 2024-2025 Market Environment (What Changed)

Several forces reshaped the relative value picture:

Mortgage rate lock-in effect. With roughly 60% of outstanding mortgages carrying rates below 4%, prepayment speeds on agency MBS collapsed. The Mortgage Bankers Association refinance index remained near historic lows through 2024. This reduced the negative convexity problem for agency MBS holders (your bond is less likely to be called away) but also compressed yields.

Non-agency credit performance. Non-QM delinquencies ticked higher in 2024, particularly among bank-statement borrowers with variable income streams (Guggenheim, Q1 2025). CRT delinquencies also increased modestly. But home equity cushions—driven by the 40%+ national home price appreciation from 2020 to 2024—kept actual loss severities low.

Federal Reserve MBS holdings. The Fed held approximately $2.5 trillion in agency MBS as of late 2024 (down from a peak of ~$2.7 trillion) and continues to allow runoff without reinvestment. This gradual withdrawal of demand is a structural headwind for agency MBS spreads.

The point is: the "free lunch" of agency MBS (no credit risk, decent yield) has gotten stingier. Low prepayments → lower yield → lower spread → less compensation for duration risk. Non-agency RMBS, by contrast, offers credit-spread pickup, but requires genuine analytical work.

Practitioner Checklist: Building Your RMBS Allocation

Essential (do these first):

  • Define your credit risk tolerance—are you a "no credit risk" investor (agency only) or willing to underwrite borrower default?
  • Match your duration target to the MBS sector: agency for longer-duration mandates, non-agency for shorter WAL profiles
  • Understand the prepayment model assumptions in any agency MBS you buy (CPR, PSA speed)

High-impact (significant edge):

  • For non-agency, analyze deal-level subordination relative to loss scenarios, not just rating labels
  • Monitor CRT as a middle ground: agency-quality collateral, non-agency yield
  • Track the Fed's MBS runoff pace—it directly affects agency spread levels
  • Evaluate non-QM vintage risk: 2020-2021 vintages have high equity cushions; 2023-2024 vintages less so

Optional (for specialists):

  • Trade the basis between agency TBA and specified pools to capture pay-up value
  • Analyze RPL/NPL pools for re-default probability using loan-level tape data
  • Consider non-agency in tax-advantaged accounts (higher income component vs. agency)

Your Concrete Next Step

Pull up the holdings of your largest bond fund (or ETF). Calculate the split between agency and non-agency mortgage exposure. If you hold zero non-agency RMBS and your mandate allows credit risk, you are likely leaving 50-250 bps of spread on the table across a meaningful portion of your fixed-income allocation. Start with AAA-rated non-agency bonds from 2023-2024 vintages as your entry point—you get credit spread pickup with subordination levels that can withstand a serious housing downturn.

Sources:

Related Articles

Extension Risk in Rising Rate Environments

Extension risk amplifies losses in rising rate environments; structured products require proactive duration management to mitigate cash flow instability.

advanced

Stress Testing Securitized Portfolios

In the Federal Reserve's 2025 severely adverse stress test scenario, house prices decline 33%, commercial real estate drops 30%, and the unemployment rate rises from 4.1% to a peak of 10.0% (Federa...

advanced

How Structured Products Trade in Secondary Markets

If you have ever tried to sell a mezzanine ABS tranche and received a bid 3-5 points below where you thought the bond was marked, you have experienced the central reality of structured product seco...

intermediate