Case Study 2 — The Market With No Guarantee: Private-Label Securitization, Before and After

A note on scope and honesty. Case Study 1 followed a guarantee that held. This one follows a market that had no guarantee at all — where the credit risk had to be handled inside the structure itself, and where the structure failed. That failure is the single largest event in the history of American mortgage finance and the reason most of the rulebook in Parts III, IV, and V exists.

No issuance volumes, market shares, loss rates, or default statistics are asserted here. Those figures are real, they are published, and they are easy to find — and a textbook that prints them from memory gets them wrong. Every trend below is stated directionally and is verifiable at the Securities and Exchange Commission, the Financial Crisis Inquiry Commission's report, and the Federal Reserve's published research. The deal profile in the second half is explicitly labeled a composite.

Background: what a private-label deal is, and what it has to solve

Recall the agency structure from §28.4. A lender delivers a loan to Fannie Mae; Fannie pools it, guarantees timely principal and interest, and the investor buys a security whose cash flow is protected from borrower default by the enterprise. The investor does not underwrite the borrowers. It does not have to. It bought a guarantee.

Now take the guarantee away.

A private-label securitization is the same mechanical structure — loans into a trust, trust issues securities, securities sold to investors — with one enormous difference: no agency stands behind it. If a borrower stops paying, no one makes the investor whole. The loss lands inside the deal.

That creates a problem that has to be solved structurally, and the solution the market adopted is subordination. The deal is cut into tranches ordered by seniority. Losses are absorbed from the bottom up: the most subordinate tranche is written down first, then the next, and the senior tranche is not touched until everything beneath it is gone. In exchange, the subordinate tranches are paid more. It is an elegant piece of engineering and, on its own terms, it works.

There is a catch, and the whole case study is inside it. Subordination protects the senior tranche against a predicted amount of loss. It does not protect against being wrong about how much loss there will be. The thickness of the subordinate tranches is set by a model. If the model is wrong in the same direction across every deal in the market at the same time, the structure does not distribute the loss — it concentrates the surprise.

The issue: what the pre-crisis market actually assumed

Private-label issuance grew rapidly through the mid-2000s, and it grew disproportionately in loan types the agencies would not buy — subprime, Alt-A, limited-documentation, option-payment adjustable rate, and high-combined-loan-to-value structures with silent seconds. That is not a coincidence. The private-label market existed precisely to finance what the agency market would not, and its growth was a direct measure of how far origination had moved outside the published rulebook.

Three assumptions held that market together, and it is worth naming them because their descendants are still around.

Assumption one: home prices do not fall nationally at the same time. Regional declines had happened repeatedly. A simultaneous national decline had not, in the modern data series the models were fitted to. Geographic diversification within a pool was therefore treated as genuine diversification — spreading loans across many states was thought to reduce correlated loss.

Assumption two: a borrower who cannot pay can sell or refinance. In an appreciating market this is very nearly true, and it makes underwriting look better than it is. A borrower who is in trouble sells at a profit, and the loan pays off instead of defaulting. Rising prices do not merely conceal weak underwriting; they convert weak underwriting into good performance data, which is then fed back into the model.

Assumption three: somebody else did the loan-level work. An investor buying a senior tranche did not review the loans. It relied on the rating, which relied on the model, which relied on loan-level data provided by parties with an economic interest in the deal closing. At no point in that chain did anyone whose own money was at risk read the files.

When national prices declined, assumption one failed. Assumption two failed with it, because the exit that had been quietly resolving weak loans closed. And assumption three meant that nobody in the chain had independent knowledge of what was actually in the pools. Ratings were downgraded — in some cases severely and rapidly — on securities that had been sold as extremely safe. Prices collapsed. Issuance of new private-label deals effectively stopped.

And notice what did not stop. Agency securities kept paying, because there was a guarantor, and Ginnie Mae securities kept paying, because there was a sovereign guarantor. The market that had no guarantee is the market that seized.

What it shows

The guarantee is not a formality; it is the entire product. An investor who buys an agency security is buying a promise from a named guarantor. An investor who buys a private-label senior tranche is buying an opinion about a model. Those feel similar in a calm market and are not remotely similar in a bad one.

When nobody with money at risk reads the files, documentation quality degrades — always, and without anyone deciding to let it. No conspiracy is required. Each participant optimizes locally: the originator is paid on volume, the aggregator is paid on deals, the rating is paid for by the issuer, and the investor relies on the rating. There is no point in that chain where reading a borrower's actual paystub is anyone's job. Chapter 14's representations and warranties exist to place that obligation somewhere specific and enforceable, and the reason they are enforced so seriously in the agency market is that everyone watched what happened where they were not.

"Diversified" is not the same as "uncorrelated." A pool with loans in forty states looks diversified and is diversified against a local factory closing. It is not diversified at all against a national decline in home prices, a national rise in unemployment, or a national tightening of credit — because every loan in it is exposed to the same three things. This is the most transferable idea in the case study and it applies well beyond mortgages.

Outcome: the partial return, and what changed

Private-label securitization did not disappear. It came back — smaller, slower, and structurally different. Directionally, and verifiable at the source, three things changed.

Risk retention. The Dodd-Frank Act required securitizers to retain an economic interest in the credit risk of the assets they securitize — the "skin in the game" requirement, implemented by interagency rule, with an exemption for qualified residential mortgages. The design intent is exactly the failure above: make somebody in the chain lose money when the loans do.

Loan-level disclosure. SEC rules substantially expanded the asset-level data that must be disclosed for registered asset-backed offerings, so that an investor can analyze the actual loans rather than only summary statistics and a rating.

Third-party due diligence. Rated deals now generally involve independent review of a sample — frequently a very large sample — of the loans in the pool, with the findings disclosed. Somebody reads the files, and their report is part of the deal.

Verify the current form of each of these at the SEC and the relevant agencies; the rules have been amended and the market practice around them has continued to evolve.

The loans in the returned market are also different. Post-crisis private-label issuance has concentrated in prime jumbo (large, well-documented loans to strong borrowers, simply too big for the agencies), expanded-credit and non-QM (bank statement income, asset depletion, investor property, recent credit events — Chapter 34's territory), and investor-property and single-family rental structures underwritten on the property's cash flow. What has not returned at anything like its former scale is high-loan-to-value, limited-documentation lending to credit-impaired owner occupants.

Composite deal profile — a modern prime jumbo securitization [composite — constructed from documented industry patterns; not a real transaction, and every figure is illustrative]

text STRUCTURE senior / subordinate, sequential pay COLLATERAL fixed-rate first-lien jumbo loans, primary and second homes LOAN COUNT about 400 loans AVERAGE BALANCE roughly $900,000 WA REPRESENTATIVE about 770 CREDIT SCORE WA LOAN-TO-VALUE about 68% DOCUMENTATION full documentation; self-employed borrowers on tax returns RESERVES typically 12+ months post-closing DUE DILIGENCE independent third-party review of a large sample of the loans, with findings graded and disclosed RISK RETENTION retained interest held by the sponsor per the applicable rule GUARANTEE NONE. No Fannie Mae, no Freddie Mac, no Ginnie Mae. Losses are absorbed by the subordinate tranches, bottom up.

Compare this line by line to the agency pool in Figure 28.1 of the chapter. Same country, same decade, same mechanical structure, radically different loans and radically different protection. The differences are not stylistic; every one of them is somebody's answer to a specific thing that went wrong.

The lesson for a loan officer

Your jumbo guidelines belong to whoever is going to buy the loan, and they change when that buyer's market changes. There is no Selling Guide for jumbo. There is an investor matrix, and behind the matrix is a securitization market or a bank balance sheet with its own appetite. When that appetite tightens, jumbo guidelines tighten — reserves go up, maximum loan-to-value comes down, documentation standards harden — and you will find out because your rate sheet changed on a Tuesday, not because anyone sent you a memo.

A file that is fine today at 80% loan-to-value may not be fine next quarter at 80%. Non-agency guidelines are not published in advance and are not stable. Do not quote from a matrix you printed last month, and do not tell a borrower near the conforming limit that their terms are settled until they are locked and the investor is identified.

The reason the agency rulebook feels rigid is that this is what the alternative looked like. Every verification requirement you find tedious — the written verification of employment, the 4506-C, the sourced deposit, the appraisal review — is a specific answer to a specific thing that happened in a market where nobody had to do it. That is not a moral argument. It is a description of what underwriting rules are: the accumulated record of what has already gone wrong.

And one thing that should make you more comfortable, not less. Nothing in this history is an argument that non-agency lending is bad. Portfolio lending is the oldest form of mortgage finance and a genuine solution for genuinely unusual borrowers. Prime jumbo securitization funds excellent loans to excellent borrowers. Non-QM, properly underwritten to ability-to-repay standards, serves self-employed borrowers whom the agency box handles badly — which is exactly the Fulton Avenue problem. The lesson is not avoid the non-agency market. It is know who is buying, know what they require today, and never assume the rules are published just because they exist.

Discussion questions

1. Subordination allocates losses that occur. It does not protect against being wrong about how much loss will occur. Explain that distinction in your own words, then apply it to a pool spread across forty states in a year when national home prices fall.

2. The chapter argues that in the pre-crisis private-label chain, reading a borrower's actual file was nobody's job. Walk the chain — originator, aggregator, issuer, rating, investor — and at each step say what that party was paid for and what they relied on. Where would you have inserted an obligation, and what would it have cost?

3. Rising home prices convert weak underwriting into good performance data. Explain the feedback loop this creates for a model fitted to recent history, and describe one thing an underwriter can look at that is not contaminated by it.

4. Compare the composite jumbo deal above to Figure 28.1's agency pool on five dimensions of your choosing. For each difference, name the failure it is responding to.

5. A borrower is buying at \$40,000 above the applicable conforming limit and is annoyed that the terms are different. Explain to them, in plain language and under two minutes, why crossing that line changes the rulebook and not merely the loan amount.

6. Non-QM lending is sometimes described in the trade press as "subprime coming back." Argue against that characterization using the ability-to-repay standards covered in Chapter 24 and the structural changes in this case study. Then argue for it as strongly as you honestly can, and say what evidence would settle the question.

Sources to verify

The Financial Crisis Inquiry Commission's final report; the Securities and Exchange Commission's Regulation AB and asset-backed securities disclosure rules; the interagency credit risk retention rule implementing Dodd-Frank's securitization provisions; Federal Reserve and FHFA published research on private-label issuance. For current non-agency guidelines, there is no substitute for the specific investor's current matrix — and for current conforming loan limits, verify at FHFA, because they change annually and by county.