Case Study 31.1 — The Great Reshuffle: How Origination Moved from Banks to Non-Banks After 2008

Type: real, public, structural Sources: public regulatory record — HMDA data published by the FFIEC and the CFPB, Ginnie Mae's published issuer requirements and issuer lists, FHFA seller/servicer eligibility requirements, HUD and Department of Justice public announcements, and the U.S. implementation of the Basel III capital framework. What is NOT here: market-share percentages, settlement amounts, and named companies' financial figures. Those exist in public data and change every year. Look them up in the sources above rather than trusting a number printed in a textbook — this chapter's changing-numbers rule applies with full force to market-structure statistics, which are among the most frequently misquoted figures in this industry.

Background: what the industry looked like before

For most of the post-war era, and overwhelmingly by the early 2000s, the largest residential mortgage originators in the United States were depository institutions — commercial banks, thrifts, and their mortgage subsidiaries — and the largest servicers were the same firms. A handful of very large banks originated through all three channels at once: enormous retail branch networks, wholesale desks buying files from thousands of brokers, and correspondent desks buying closed loans from smaller lenders. They also held vast portfolios of mortgage servicing rights.

The structure made sense on its own terms. A depository has cheap, stable funding in the form of insured deposits. It has capital. It can hold a loan if it wants to and sell it if it prefers. It has distribution — branches full of customers who already trust it with their money. And servicing, which generates a small fee on an enormous number of loans and requires scale, technology, and patience, was a natural fit for an institution with a long horizon.

Independent mortgage companies existed and originated real volume, but they were smaller, thinner, and structurally dependent on the banks — because the warehouse lines that funded their closings were provided, then as now, mostly by banks.

The issue: four forces that pushed depositories out

After 2008 the calculus changed, and it changed for reasons that are structural rather than sentimental. Four forces are visible in the public record.

1. Representation-and-warranty and repurchase exposure

When a lender sells a loan to Fannie Mae or Freddie Mac, it makes representations and warranties about the loan's compliance with the selling guide. If those representations turn out to be untrue, the lender can be required to repurchase the loan — at par, years later, when the loan is delinquent and the collateral has fallen in value. Chapter 14 explains the mechanism.

Repurchase demands after the crisis arrived at a scale nobody had modeled. For institutions that had originated in the millions of loans, the resulting exposure was measured in years of litigation and negotiation. The rational management response was not merely to underwrite more carefully. It was to stop originating the categories of loan most likely to generate a demand — which meant heavy credit overlays above the agency minimums, especially on lower-score and higher-loan-to-value lending, and it meant scrutinizing loans originated by third parties even harder, because a bank buying a broker's file bears the representation risk on work it did not supervise.

That is the first structural push, and note where it lands: it falls hardest on exactly the borrowers this book spends the most time on. The Harlow Street file — a 641 representative score, one income, ratios at 41.48% front and 51.00% back, approvable only through the automated scorecard with compensating factors — is precisely the file a repurchase-averse institution writes an overlay against.

2. False Claims Act exposure on FHA lending

FHA-approved lenders certify to HUD that the loans they endorse for insurance comply with HUD requirements. After the crisis, the Department of Justice pursued cases under the False Claims Act against lenders whose certifications were alleged to be defective, and a number of large institutions entered public settlements. The details, the allegations, and the amounts are all matters of public record; find them there.

What matters structurally is the response. Several large depositories publicly reduced or exited FHA lending, citing certification risk. The exposure was viewed as disproportionate to the economics of the product: an FHA loan is a modest-margin, high-touch product serving borrowers with thinner credit, and the tail risk attached to a certification error was unbounded in a way ordinary underwriting risk is not. HUD and DOJ later worked to clarify certification language and defect taxonomy specifically because lender withdrawal from FHA lending had become a policy problem — which is as clean an admission as one gets that the withdrawal was real and was driven by legal risk.

The consequence for borrowers: the institutions best positioned to serve low-down-payment, lower-credit buyers largely stopped doing so, and non-banks filled the space.

3. The capital treatment of mortgage servicing rights

A mortgage servicing right is an asset — a contractual right to a fee stream, valued by modeling how long the loans will pay. Chapter 28 explains it. Under the Basel III framework as implemented in the United States, MSRs receive restrictive regulatory capital treatment: their inclusion in common equity tier 1 capital is limited, and amounts above the threshold are deducted, with risk weighting applied to what remains. Verify the current requirements — capital rules are revised — but the direction has been consistent.

The practical effect is that holding a large MSR portfolio consumes bank capital in a way it does not consume a non-bank's, because a non-bank has no risk-based capital regime of that kind. That asymmetry made banks natural sellers of servicing and non-banks natural buyers, and the servicing of government loans in particular migrated. Ginnie Mae's issuer base — the firms that actually issue the securities backed by FHA, VA, and USDA loans — shifted markedly toward non-depositories. Ginnie Mae's own published issuer eligibility requirements were revised over the following years precisely because the composition of its issuer base had changed.

4. The cost of servicing loans that go bad

Servicing a performing loan is a routine, low-margin, scale business. Servicing a delinquent loan is neither. Government servicing carries advance obligations — the servicer must advance principal and interest to security holders even when the borrower has not paid — and elaborate loss-mitigation requirements with real penalties for procedural error. After the crisis, servicing costs per delinquent loan rose dramatically and several large servicers entered public settlements over servicing practices.

Banks, weighing the fee income against the operational and legal cost, sold. Non-banks, whose entire business is mortgages and who could build servicing operations sized to that reality, bought.

What made the other side of the trade possible

None of the above would have produced a shift if there had been nobody to shift to. Four things made the non-bank side viable.

Warehouse capacity. Banks did not leave mortgage lending. They largely stopped originating consumer mortgages at their prior scale and instead financed the firms that did — which is a lower-risk, shorter-duration, better-collateralized way to be in the same business. A bank lending on a warehouse line for eighteen days against a saleable agency-eligible loan holds a very different risk than a bank originating that loan and warranting it for years. This is the single most underappreciated fact about the post-2008 structure: the money is still substantially the banks' money. It just arrives at the closing table through a non-bank.

Licensing infrastructure. The S.A.F.E. Act and the NMLS, built in 2008 in response to the crisis, gave non-bank lending something it had never had: a nationwide registry, a common licensing standard, testing, and a way for states to see each other's records. Ironically, the regime built to police non-bank origination also made non-bank origination credible enough to scale.

Technology and focus. A firm whose only business is mortgages allocates all of its investment to mortgages. Through the 2010s, non-banks were consistently faster to adopt automated underwriting integrations, digital document collection, e-signature, and pricing engines.

Agency and Ginnie Mae approval. The seller/servicer and issuer approval processes, with their net-worth and quality-control requirements, gave non-banks a defined path to direct delivery — which is the correspondent structure of §31.4 operating at institutional scale.

Outcome: what the industry looks like now, structurally

Three durable changes, stated without numbers:

1. A large share of origination — and a much larger share of government-insured origination — is done by non-depository lenders. This is visible in HMDA data and in Ginnie Mae's issuer lists, both public.
2. The regulators built the missing capital regime after the fact. FHFA established minimum financial eligibility requirements for seller/servicers — net worth, capital ratio, and liquidity — and Ginnie Mae did the same for issuers, with subsequent revisions adding risk-based capital concepts. In other words, having watched activity migrate to firms with no prudential regulator, the counterparties themselves became the prudential regulator. Verify the current requirements at the source; they have been revised more than once.
3. The fragility moved with the activity. In spring 2020, when pandemic forbearance under the CARES Act let borrowers stop paying while servicers remained obligated to advance, and when securities-market dislocation produced margin calls on hedges, non-bank servicers faced a liquidity squeeze that a deposit-funded institution would not have faced in the same way. The public responses were structural and quick: Ginnie Mae established a pass-through assistance facility for issuers, and FHFA announced a limit on the period for which servicers of loans in forbearance must advance. Both are on the public record and both are worth reading, because they are the clearest available illustration of §31.6's point about where funding fragility lives.

What it shows

Regulation does not eliminate an activity; it relocates it to wherever it is cheapest to conduct. Nothing after 2008 made low-down-payment lending, FHA lending, or servicing delinquent loans less necessary. The rules changed who could do them at an acceptable cost. Capital requirements, certification liability, and repurchase exposure priced depositories out of specific businesses, and those businesses moved to firms not subject to the same pricing — funded, in large part, by the same depositories, one step removed.

The channel and the institution are separate questions, and the institution matters more. A non-bank correspondent, a non-bank wholesaler, and a non-bank retail lender all share the same funding fragility and the same absence of portfolio capability. A bank running a retail division and a bank running a correspondent desk share the same capital regime and the same examiner. §31.6 is the cut that explains the last fifteen years; §31.1's three channels are the cut that explains a Tuesday.

Overlay differences are risk-posture differences. When a new loan officer notices that the independent mortgage company down the street will do a file their bank will not, the explanation is almost never that the bank's underwriters are unreasonable. It is that the two firms face different liabilities for the same loan.

The lesson for a loan officer

Three, in descending order of how soon they will matter to you.

1. Know which kind of institution you work for, and know what it is structurally good and bad at. If your employer is a depository, your advantage is portfolio capability, stability, and branch referral flow, and your handicap is overlays. If it is a non-bank, your advantage is product breadth, speed, and focus, and your handicap is that the firm's funding can be withdrawn.
2. Ask about warehouse capacity and capital when you interview at a non-bank, and do not accept a vague answer. Everything on the public record says that non-bank capacity contracts in exactly the markets where your income also contracts.
3. Do not read a market-share statistic in a book — including this one — and repeat it. HMDA data is free, public, and current. The number you half-remember from a conference three years ago is the number a borrower or an agent will catch you on.

Discussion questions

1. The chapter argues that after 2008, banks largely stopped originating consumer mortgages at their prior scale and instead financed the non-banks that did. Is that a reduction of risk in the financial system, a transfer of risk, or a transformation of one kind of risk into another? Defend your answer with reference to what a warehouse bank actually holds.

2. FHA lending serves borrowers with lower credit and smaller down payments. Several large depositories reduced or exited that business citing certification liability. Was that outcome an intended consequence of enforcement policy, an unintended one, or an acceptable cost? Who bore the effect?

3. FHFA and Ginnie Mae responded to the migration by imposing net worth, capital, and liquidity requirements on non-bank counterparties. Is a counterparty requirement a satisfactory substitute for prudential supervision? Name one thing it can do and one thing it cannot.

4. If you were advising a first-year loan officer in 2026 who had offers from a regional bank and from an independent mortgage company, what would you tell them about the credential question in §31.6 — and would your answer change if they were in year fifteen instead of year one?

5. The case study insists on giving no market-share numbers. Argue the other side: what is lost by teaching this material without figures, and how would you go about getting the current ones before your next client conversation?

6. The Harlow Street file — 641 score, one income, 51.00% back-end ratio, an approve from the automated scorecard with compensating factors — is named in the text as the archetype of a file an overlay is written against. Using this case study, explain to that borrower why they were declined at one lender and approved at another, without disparaging either institution.