Mortgage-Backed Securities

The instrument in one breath

A bank originates a thirty-year fixed-rate mortgage at, say, 6 percent. Without securitisation it sits on the bank's balance sheet for the next three decades, tying up roughly 25 percent of regulatory capital per dollar of loan and locking the bank out of further lending until either the loan amortises or the bank raises new equity. Securitisation breaks the loop. The bank gathers several thousand such mortgages into a single pool, transfers them to a bankruptcy-remote special-purpose vehicle (SPV), and sells investors securities that entitle them to the monthly cash flow the pool produces. The bank books an origination fee, returns to the lending business, and the credit and interest-rate risk shifts to the investors who bought the securities.

That is the entirety of the financial engineering — pool, partition, pass through — and the consequences for the American mortgage market have been enormous. Securitisation lets a Florida pension fund effectively own a slice of a Wisconsin homeowner's mortgage payment, expands the pool of capital available to fund US housing far beyond what local deposit-taking institutions could provide, and (until 2008) was widely credited with lowering the 30-year fixed mortgage rate by roughly 50–100 basis points relative to a non-securitised counterfactual.

Lewis Ranieri and the 1977 invention

Ginnie Mae issued the first US mortgage pass-through security in 1970 — but only on government-insured FHA and VA loans. The breakthrough that built the modern market came from Salomon Brothers' mortgage desk under Lewis Ranieri. In 1977 Ranieri and his colleague Bob Dall structured the first private-label pass-through securitising conventional (non-government-guaranteed) mortgages, originating from Bank of America. The deal required new state-by-state legal infrastructure to allow institutional investors to hold the securities across jurisdictional lines, which Ranieri lobbied for over the following decade and which culminated in the Secondary Mortgage Market Enhancement Act of 1984.

Ranieri's contribution was not the mathematics of pooling. It was operational: standardising prospectuses, working with the rating agencies, building a market-making book, training a generation of mortgage traders, and persuading regulators that institutional investors could safely hold the new instrument. By the mid-1980s Salomon's mortgage desk was, by some accounts, the single most profitable trading operation on Wall Street, and Ranieri appeared on the cover of BusinessWeek as "the man who invented the modern mortgage market". Michael Lewis's Liar's Poker (1989), set on that very desk, made the operation famous beyond finance.

Agency vs private-label — the central distinction

Every modern MBS falls into one of two categories, and the distinction shapes every aspect of how it trades.

Type	Issuer / guarantor	Loans pooled	Credit risk	Share of US issuance (2024)
Ginnie Mae MBS	Government National Mortgage Association (US Treasury full-faith-and-credit guarantee)	FHA, VA, USDA-insured loans	None to investor	≈ 30 %
Fannie Mae MBS	Federal National Mortgage Association (GSE, federal conservatorship since 2008)	Conforming conventional loans	Implicit federal backing; explicit since 2008	≈ 40 %
Freddie Mac MBS	Federal Home Loan Mortgage Corporation (GSE, federal conservatorship since 2008)	Conforming conventional loans	Implicit federal backing; explicit since 2008	≈ 25 %
Private-label MBS	Investment banks (Goldman, JPMorgan, etc.)	Subprime, Alt-A, jumbo (non-conforming)	Borne fully by tranche investors	< 5 %

The agency–private divide reduces to a single question: who eats a default? For agency MBS, the guarantor makes the investor whole — agency MBS investors face essentially no credit risk and are pricing only prepayment and interest-rate exposure. For private-label MBS, default losses flow directly through to the security holders, mediated only by the tranching structure inside the deal.

The private-label market peaked at around 2 trillion dollars outstanding in 2006 — overwhelmingly subprime and Alt-A — and collapsed in 2008. Today it is a niche under 200 billion dollars, while the agency market grew to over 12 trillion through both organic growth and Federal Reserve quantitative-easing purchases.

Tranching — how subprime mortgages became AAA bonds

Within a private-label deal, the pool's monthly cash flow is paid through a waterfall. Imagine a pool of 1,000 mortgages with $200 million of total principal outstanding. The deal structure carves the cash-flow rights into tranches, conventionally:

Senior (AAA)      $160 m     80 %  — paid first, last to take losses
Mezzanine (BBB-A) $ 30 m     15 %  — paid after senior, before equity
Equity (unrated)  $ 10 m      5 %  — paid last, first to absorb losses

Interest and scheduled principal cascade down the waterfall: every dollar collected on the pool pays the senior tranche its required interest, then the mezzanine, then the equity. Losses cascade up from the bottom: the equity tranche is wiped out by the first 5 percent of cumulative pool losses, then the mezzanine absorbs the next 15 percent, and only after losses exceed 20 percent of the pool's principal does the senior tranche begin to be impaired.

This is credit enhancement through subordination. The senior tranche carries a much higher effective credit quality than the average loan in the pool because it sits on top of a 20-percent loss cushion. Rating agencies — Moody's, S&P, Fitch — would then assign AAA to that senior tranche based on a model of how cumulative pool losses behave. As long as cumulative losses stayed below 20 percent the AAA was safe; the model gave the AAA tranche a 1-in-10,000 probability of impairment.

The arithmetic worked, given the rating-agency models. The models assumed default correlations across geographies and loan types of roughly 0.05–0.15. With that level of independence, even a pool of subprime loans could support an 80 percent AAA-rated senior tranche. The error — the very large error — was about that correlation assumption, which we will return to.

CDO — the MBS of MBS

The investment-bank engineering went one level deeper. The mezzanine tranches of private-label MBS deals — BBB-rated, paying 200–400 basis points over Treasuries — were attractive on a yield basis but hard to place in size. In 2005–2007, Wall Street's answer was the collateralised debt obligation (CDO): a special-purpose vehicle whose collateral was not mortgage loans but the BBB tranches of dozens of other MBS deals.

That CDO would then be tranched again. A typical mezzanine CDO of MBS converted a pool of all-BBB collateral into roughly 75–80 percent AAA-rated senior CDO tranches plus a smaller stack of subordinated CDO tranches. The supposed magic: pooling BBB tranches from many different MBS deals diversified the credit risk enough that the senior tranche of the CDO became AAA. From a 5 percent equity slice of a subprime mortgage pool, the system manufactured AAA paper twice — first inside the MBS, then again inside the CDO that held the MBS mezzanine.

It was bookkeeping, not alchemy, and it depended entirely on the same low-correlation assumption that supported the underlying MBS tranching. Once correlations rose, the mezzanine tranches of every contributing MBS defaulted together, and the AAA tranche of the CDO collapsed first and fastest. The 2008 ABX subprime indices show subprime-mezzanine-CDO AAA tranches trading at 20 cents on the dollar within eighteen months of issuance — a loss profile no AAA bond in modern financial history had ever experienced.

Prepayment risk and negative convexity

Every US fixed-rate mortgage borrower holds a free option to refinance: pay off the existing loan at par and replace it with a new loan at the prevailing rate, paying only modest closing costs. From the MBS investor's perspective, refinancing means the pool's principal flows back at par exactly when rates have fallen — that is, exactly when the bond would otherwise have risen above par. The borrower's option is the investor's anti-option.

The aggregate behaviour of borrowers in a pool is forecast by prepayment models. The simplest is the constant prepayment rate (CPR) — the annualised fraction of remaining principal that prepays per year, typically 5–10 percent for seasoned conforming pools but accelerating to 40–60 percent when rates fall sharply. A more refined model — the PSA prepayment benchmark from the Public Securities Association — ramps from 0.2 percent CPR at month 1 to 6 percent CPR at month 30 and constant thereafter. Modern Wall Street prepayment models are far more elaborate and condition on borrower credit score, loan-to-value, geographic location, refinancing incentive (rate gap between coupon and current market), seasoning, and burnout (the depletion of refinancing-prone borrowers over time).

The consequence for bond pricing is negative convexity. A standard bond has positive convexity: a 100-basis-point fall in yields produces a larger price gain than the loss from a 100-basis-point rise (the price-yield curve is convex upward). An MBS investor, holding a short prepayment option, faces the opposite. As rates fall, prepayments accelerate, the bond's effective duration shortens, and the price gain is capped near par. As rates rise, prepayments slow, duration extends, and price losses are amplified. The MBS price-yield curve bends the wrong way. Investors demand a higher yield than a comparable-duration Treasury — the option-adjusted spread (OAS) — to compensate for sitting on the bad side of this option.

Treasury bond (positive convexity)
  100 bp drop in yield  →  price up by 8 %
  100 bp rise in yield  →  price down by 7 %

MBS pool (negative convexity)
  100 bp drop in yield  →  price up by only 2 %  (prepayments accelerate)
  100 bp rise in yield  →  price down by 9 %    (duration extends)

The 2008 crisis, in mechanism

The failures of 2007–2008 were not failures of the MBS structure as such — agency MBS performed exactly as expected throughout. They were failures of the private-label subprime market and, even more sharply, the subprime mezzanine CDO market built on top of it. Three structural assumptions collapsed simultaneously.

1. Default correlation rose to one. When US house prices fell — the Case-Shiller national index dropped 28 percent peak-to-trough — defaults rose simultaneously in every region, on every loan type. The rating agencies' assumed correlation of 0.05–0.15 had baked in geographic and idiosyncratic diversification that, on a synchronised national price decline, simply did not exist.
2. AAA tranches lost mark-to-market value before realising principal losses. Even tranches that would eventually pay full principal saw their market price collapse as buyers vanished. Banks holding the paper had to take mark-to-market write-downs — Citigroup wrote down roughly 60 billion dollars, UBS 50 billion, Merrill Lynch 50 billion — which depleted regulatory capital and triggered fire sales.
3. AIG's CDS book imploded. AIG Financial Products had written tens of billions of dollars of credit-default swaps insuring AAA tranches of subprime MBS and CDOs. Downgrades from AAA to AA triggered collateral-posting requirements that AIG could not meet. The Federal Reserve extended an 85-billion-dollar credit line on September 16, 2008 — the day after Lehman filed — to prevent AIG's default from cascading through its counterparties. AIG eventually received over 180 billion dollars in federal support.

Bear Stearns failed in March 2008 after two of its subprime-MBS hedge funds collapsed. Lehman Brothers filed for bankruptcy on September 15, 2008, owing 600 billion dollars, with a balance sheet heavy in private-label MBS, CDOs, and the warehouse loans funding them. Merrill Lynch was forced into the arms of Bank of America the same weekend. Washington Mutual — the largest US thrift, with a mortgage portfolio of 230 billion dollars — was seized by the FDIC on September 25 in the largest bank failure in US history. The trigger for all four collapses was the same: MBS and MBS-derivative exposure.

Post-crisis reforms — Dodd-Frank, QM, QRM

The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 made two principal changes to MBS issuance that directly address the 2007 failure modes.

• Risk retention (Section 941). Securitisation sponsors must retain a 5 percent economic interest in any deal they issue — they cannot pass through 100 percent of the credit risk to investors. The intent is to remove the "originate and distribute" incentive: if the originator must hold a 5 percent first-loss piece, it has a reason to underwrite carefully. Implementation followed in 2015 (commercial MBS) and 2016 (residential).
• Qualified Mortgage (QM) and Qualified Residential Mortgage (QRM) rules. The Ability-to-Repay regulation, issued by the Consumer Financial Protection Bureau in 2013, requires lenders to verify a borrower's capacity to repay before originating a loan. The QM safe harbour exempts loans with conservative features (no negative amortisation, no interest-only or balloon payments, debt-to-income ratio ≤ 43 percent, full income documentation, points and fees ≤ 3 percent) from rebuttable-presumption ATR litigation. QRM is the securitisation analogue: pools of QRM-eligible loans are exempt from the 5 percent risk-retention requirement.

The Federal Reserve also entered the MBS market as a buyer during the crisis. Three rounds of quantitative easing (2008–2014, 2020–2022) ultimately took the Fed's MBS holdings to a peak of 2.7 trillion dollars in mid-2022 — roughly a quarter of the entire outstanding agency MBS market. Quantitative tightening since 2022 is reducing those holdings at a pace of up to 35 billion dollars per month.

Worked example: 30-year pool, prepayment, tranche cash flow

Consider a simplified $100 million subprime MBS deal at issuance:

Pool:  $100 m of 30-year subprime mortgages at 8 % coupon
Tranches:
  Senior (AAA)  $ 80 m at 5.5 % coupon
  Mezz   (BBB)  $ 15 m at 7.0 % coupon
  Equity        $  5 m residual

Month 1 cash flow from pool:  interest ≈ $0.67 m, scheduled principal ≈ $0.06 m
Waterfall:
  Senior interest:  80 × 0.055 / 12 = $0.367 m   paid first
  Mezz   interest:  15 × 0.070 / 12 = $0.088 m   paid second
  Equity residual:  $0.67 + $0.06 − $0.367 − $0.088 − senior principal = ~$0.21 m
                                                                       per month, until losses

If cumulative pool losses stay below 5 percent over the deal life, the equity tranche earns a high running yield (residual interest minus principal write-downs) and the senior and mezz pay in full. If losses reach 8 percent, the equity is wiped out and the mezz takes a 3-percentage-point hit; the senior still pays in full. If losses reach 25 percent — as many subprime pools experienced in 2008 — the equity and mezz are both wiped out and the senior takes a 5-percentage-point principal loss. In a subprime mezzanine CDO that holds those mezz tranches, the underlying BBB pieces have all gone to zero, and the CDO's own AAA senior tranche is now exposed directly to subprime pool losses with much less protection than its rating advertised.

Variants of the basic MBS

• Pass-through. The simplest structure — investors share pro rata in every dollar collected from the pool. No tranching. The original 1977 Salomon deal. Most agency MBS today are pass-throughs (Ginnie Mae pools, Fannie/Freddie TBA-deliverable pools).
• Collateralised Mortgage Obligation (CMO). Tranches by prepayment timing rather than credit. PAC (planned amortisation class), TAC (targeted amortisation class), and support tranches re-allocate prepayment risk: PAC investors get a smoothed cash flow within a band of prepayment speeds, support investors absorb the variation. Created in 1983 by Freddie Mac and First Boston.
• Interest-only (IO) and principal-only (PO) strips. Decompose the pool's cash flow into its two components. IO strips benefit from slow prepayments (more interest collected over more months); PO strips benefit from fast prepayments (principal returned faster, time-value gain). Both have explosive convexity in opposite directions; they are favourite hedges for traders managing prepayment exposure.
• Commercial MBS (CMBS). Pools of commercial-property mortgages — offices, shopping centres, hotels, multifamily — rather than residential. Different prepayment dynamics (commercial loans are typically defeasance-protected or yield-maintenance, removing most prepayment optionality) and different credit profile (lumpy, with single loans up to 100+ million dollars).
• Asset-backed security (ABS). Same engineering, different collateral — auto loans, credit-card receivables, student loans, equipment leases. Auto-loan ABS is the largest non-mortgage segment (~250 billion dollars outstanding) and largely shrugged off 2008.
• Covered bond. European alternative: the originating bank issues a bond backed by a pool of mortgages that remain on its own balance sheet. If the bank defaults, bondholders have a senior claim on the pool; if the pool underperforms, they have full recourse to the bank. Combines the strengths of both models — popular in Germany, Denmark, the Nordics — and largely avoided the 2008 problems that hit US MBS.

Common pitfalls and confusions

• Conflating MBS with the 2008 crisis as a whole. Agency MBS performed exactly as designed throughout 2008 — Ginnie Mae paid every cent, Fannie and Freddie were placed in conservatorship to ensure their guarantees continued. The crisis was a private-label subprime and CDO crisis. Conflating "MBS" with "what blew up" obscures both what works and what didn't.
• Assuming AAA equals safe. A AAA rating is a credit-rating-agency model output, not a tautological guarantee. The pre-2008 ratings on subprime mezzanine CDOs were built on a default-correlation assumption that turned out to be wildly wrong. Read the model, not just the letter.
• Treating prepayment as merely a credit feature. Prepayment risk is the first-order risk in agency MBS, not credit. An agency MBS investor's main worry is that everyone refinances at once and they get their principal back at the worst possible moment for reinvestment. The yield premium over Treasuries (OAS) is paying for that option, not for default risk.
• Forgetting the implicit-vs-explicit GSE backing. Before September 2008, Fannie and Freddie carried an implicit federal guarantee — markets believed it would be honoured, but no statute required it. In September 2008 the implicit became explicit when Treasury committed to keep them solvent. The conservatorship is ongoing as of 2025; periodic political proposals to exit it have not led to action.
• Believing securitisation is intrinsically destabilising. Securitisation is the dominant funding model for US housing and has been for forty years; the 30-year fixed-rate mortgage that most Americans take for granted exists at scale only because of the securitisation market. The crisis came from specific structures (subprime mezzanine CDOs, off-balance-sheet SIVs, mismatched short-term funding) layered on top — not from securitisation per se.