Case Study 31.2 — When the Line Goes Away: Warehouse Funding, the Wholesale Channel, and the Fragility of Borrowed Money

Type: real public history (Parts 1, 2, and 4) plus a clearly labeled composite (Part 3) Complements Case Study 31.1, which explained where origination moved after 2008. This one explains how fast a firm in that structure can stop existing — and what happened to the broker channel along the way. No statistics are quoted. Origination volumes, channel shares, and firm-level figures are public and change constantly; sources are named at the end.

Background: the sentence that runs this case

From §31.4: a correspondent funds loans at closing with borrowed money and repays it when the loan is sold.

Read that sentence as a lender rather than as a loan officer and it says something uncomfortable. The firm's ability to open for business tomorrow depends on a credit facility it does not control, is secured by loans somebody else has to agree to buy, and is typically terminable on short notice. A non-bank lender is never more than two counterparty decisions away from being unable to fund a closing — the warehouse bank's decision to keep advancing, and the investor's decision to keep purchasing. Both decisions get made in the same conditions and tend to move together.

Depositories are not immune to bad markets, but they do not fail this way. They fund with insured deposits and hold capital against the risk. That is the whole of §31.6 in operational terms.

Part 1 — 2007: the mechanism, in public

The 2007 failures of non-bank mortgage lenders are usually explained as credit losses. For several of the most prominent, that is not what the public record shows happened first.

The sequence that actually ran, repeatedly, in mid-2007:

HOW A LENDER STOPS FUNDING — the 2007 sequence            [structural summary of the
                                                            documented public pattern]

  1. Investors reprice or stop bidding on a loan category.
        The loans on the warehouse line are suddenly worth less than the advances
        against them, or are worth nothing quotable at all.
                     │
                     ▼
  2. The warehouse bank marks the collateral down and issues a MARGIN CALL.
        "Post cash, or we reduce the advance."
                     │
                     ▼
  3. The lender pays -- from the same cash it needs for haircuts on tomorrow's
     closings. Liquidity falls while obligations do not.
                     │
                     ▼
  4. Covenants trip. Net worth or liquidity minimums are breached.
                     │
                     ▼
  5. Lines are reduced, then suspended, then terminated. Often within DAYS.
                     │
                     ▼
  6. The lender cannot fund closings scheduled for this week.
        Borrowers sitting at settlement tables do not close.
                     │
                     ▼
  7. Locked pipelines are abandoned. Employees are released. The firm files.

One large Alt-A lender announced in August 2007 that it was unable to fund loans it had already committed to, ceased lending, and filed for bankruptcy protection within days — a matter of public record, and the cleanest available illustration of step 6. Note what it was not: it was not primarily a subprime lender, and it did not fail because a large share of its loans had defaulted. It failed because the funding stopped and the loans in its warehouse became unsaleable at the price the advances assumed.

The lesson is in the timing. A credit problem takes years to show up in a loan portfolio. A funding problem takes a week. The borrowers who did not close that August were not denied. Their loans were approved, locked, cleared, and scheduled. There was simply no money at the table.

Part 2 — The wholesale channel contracts, then returns

The broker channel was hit twice, by two different forces.

Force one: the buyers left. Through roughly 2008–2012, several of the largest depository lenders announced exits from wholesale lending. The public reasoning was consistent — representation and warranty exposure on loans they had not originated and could not supervise, plus reputational and legal risk from third-party origination practices. Case Study 31.1's first force, applied to the channel where it bites hardest. When a bank buys a broker's file, it warrants work done by a firm it does not employ, and after 2008 that trade stopped penciling.

The result was straightforward supply-side contraction: fewer wholesale lenders, so fewer approvals, so less choice — and choice, per §31.3, is the broker channel's entire advantage. A brokerage approved with three lenders instead of twelve is a much weaker version of itself.

Force two: the compensation model changed. The loan originator compensation rule (Chapter 26 owns it) ended the practice by which a mortgage broker's compensation varied with the interest rate delivered. Whatever one thinks of the old arrangement, it had been a substantial part of broker-channel economics, and the transition to fixed, pre-set compensation plans was disruptive. A number of brokerages closed or converted to correspondent or retail structures during this period.

Then it came back. Through the 2010s, non-bank wholesale lenders rebuilt the channel — investing in broker portals, pricing engines, and turn times, and competing hard for broker approvals. Brokerage counts and broker-channel share recovered substantially from the trough, though not to pre-crisis levels. Two things are worth noticing about the recovery:

  • The wholesale lenders are now mostly non-banks. So the channel's fragility and the funding fragility of Part 1 are now the same fragility. A broker's chosen lender is a firm that funds with borrowed money.
  • The competition got sharp. In 2021 a major wholesale lender required its broker partners to choose between it and two named competitors, on penalty of losing access. The requirement was widely reported and produced litigation. Set aside the merits, which are not this book's business: the episode is structurally instructive, because it is only possible in a channel where the originator's access to product is a contract with a counterparty rather than an employment relationship. A retail loan officer's product menu cannot be revoked by another company's business decision. A broker's can.

That is the mature form of §31.3's trade. Choice is real, and choice is granted by counterparties who have their own interests.

Part 3 — COMPOSITE: how a line actually goes away

⚠️ This section is a COMPOSITE. No such firm exists. Every figure is constructed to illustrate the arithmetic of a warehouse squeeze, and the mechanism is assembled from documented public patterns across 2007, 2020, and 2022. Advance rates, covenants, and pricing are negotiated and confidential — never quote these as typical.

A mid-size non-bank correspondent, going into a rate shock:

THE FIRM, BEFORE                                           [COMPOSITE - constructed]

  Tangible net worth                                      $  8,000,000
  Warehouse facilities, three banks                       $ 96,000,000
  Leverage covenant                                            12 : 1
  Average loan amount                                     $    340,000
  Advance rate                                                   97.5%
     advance per loan   $340,000 x 0.975  =              $    331,500
     haircut per loan   $340,000 x 0.025  =              $      8,500
  Loans outstanding capacity  $96,000,000 / $331,500  =        289 loans
  Average dwell time                                          20 days
     turns per year    365 / 20  =                              18.25
     annual capacity   289 x 18.25 = 5,274 loans  =    $1,793,160,000

Now three blows, in the order they actually arrive.

Blow 1 — the covenant. Rates move against the pipeline. Hedge losses, a collapse in pull-through, and two months of fixed costs against falling revenue produce an operating loss of \$2,200,000. Tangible net worth falls to \$5,800,000. The 12:1 covenant now permits:

$$\$5{,}800{,}000 \times 12 = \$69{,}600{,}000 \text{ of line}$$

Line capacity falls by \$26,400,000**. At \$331,500 per loan, loans outstanding capacity falls from 289 to 209**. Nobody did anything wrong. The covenant did what covenants do.

Blow 2 — the advance rate, which looks like good news. The warehouse banks, watching the same market, cut the advance rate from 97.5% to 95.0%:

at 97.5% at 95.0%
Advance per loan \$331,500 | \$323,000
Haircut per loan (the firm's own cash) \$8,500** | **\$17,000
Loans the \$69,600,000 line permits 209 215

Read that table twice. The line now permits more loans — 215 instead of 209 — because each loan consumes less of it. And the firm can afford far fewer of them, because each one now demands \$17,000** of its own cash instead of \$8,500. To use the line fully, the firm would need $215 \times \$17{,}000 = \$3{,}655{,}000$ tied up in haircuts at all times. A cut in the advance rate is a capital call wearing a different hat.**

Blow 3 — the liquidity covenant closes the trap. The facilities require \$4,000,000 of unrestricted cash. The firm has \$4,600,000. Cash actually available for haircuts:

$$\$4{,}600{,}000 - \$4{,}000{,}000 = \$600{,}000$$

$$\$600{,}000 \div \$17{,}000 = 35.3 \rightarrow \textbf{35 loans outstanding}$$

THE FIRM, AFTER                                            [COMPOSITE - constructed]

  Loans outstanding capacity   BEFORE   289
                               after covenant reduction    209   (line-constrained)
                               after advance-rate cut      215   (line permits)
                               AFTER, cash-constrained      35   <-- the binding limit

  Annual capacity at a 20-day dwell
     before   289 x 18.25 = 5,274 loans  =              $1,793,160,000
     after     35 x 18.25 =   638 loans  =              $  216,920,000
     ─────────────────────────────────────────────────────────────────
     reduction                                          $1,576,240,000

The firm still has \$69,600,000 of committed lines it cannot use. It still has approvals, licenses, technology, and loan officers. What it does not have is the cash to put behind a haircut, and that single constraint has removed roughly \$1.58 billion of annual funding capacity.

What happens next is the part loan officers see. Volume targets are cut. Products with slower execution are suspended, because dwell time is now rationed. Pricing gets worse — not out of greed, but because the firm must widen margins on a smaller book to cover fixed costs, exactly as Chapter 29 describes. Loan officers whose files no longer fit start leaving, which reduces revenue further. And somewhere in that sequence, a borrower on a fifty-one-day purchase contract discovers that their lender has stopped taking new applications in their product.

None of this required a single loan to default.

Part 4 — Two more real stress tests, briefly

Spring 2020. Pandemic forbearance under the CARES Act allowed borrowers to stop paying while servicers remained obligated to advance principal and interest to security holders. Simultaneously, dislocation in the mortgage securities market produced margin calls on the hedges lenders use against their pipelines. Non-bank servicers and originators — no deposits, no discount window — faced an acute liquidity squeeze. The public responses were structural: Ginnie Mae established a pass-through assistance facility for issuers unable to meet advance obligations, and FHFA announced a limit on how long servicers of loans in forbearance must advance. Both are on the public record.

2022–2023. As rates rose sharply, origination volume collapsed and the industry consolidated fast. Non-bank lenders merged, were acquired, wound down channels, or closed. At least one sizable lender exited the wholesale channel entirely, which — per Part 2 — subtracts choice from every brokerage that had been approved with it. Verify current conditions in Mortgage Bankers Association data and company filings rather than from any textbook, including this one.

What it shows

  1. Funding failure is fast and credit failure is slow. Everything a loan officer worries about — ratios, scores, appraisals, conditions — operates on the slow clock. The fast clock is the one that empties closing tables, and almost nobody on the sales side watches it.
  2. A covenant does not need anyone to behave badly. The composite firm made no bad loans. It lost money in a bad quarter, and the contract did the rest. When you ask a non-bank employer about warehouse capacity, you are asking about covenants and cash, not about the headline size of the lines.
  3. Choice is a counterparty relationship. The broker channel's advantage is real and it is granted, not owned. Approvals can be withdrawn, lenders can exit the channel, and a competitor's business decision can change your product menu. Diversify approvals for the same reason a correspondent diversifies warehouse banks.
  4. The channel and the funding structure interact. A brokerage does not have a warehouse line and cannot fail this way — but every lender it places files with can. "No balance sheet risk" means the risk sits one step away, not that it is absent.

The lesson for a loan officer

  • At a non-bank, ask about lines, covenants, liquidity, and who the warehouse providers are — and ask again after a bad quarter. This is not impertinent. It is the same question your employer's underwriter asks your borrower.
  • As a broker, count your approvals, and treat a shrinking count as an early warning about your own business rather than someone else's.
  • In any channel, a lender that suddenly cannot fund is the one risk your discipline cannot offset. Know the second place you would send a file, before you need it. Case Study 31.1's discipline was to know what your institution is structurally good at; this one's is to know what happens if it stops.

Discussion questions

  1. In the composite, the advance-rate cut increased the number of loans the line permitted while decreasing the number the firm could fund. Explain that to someone who does not work in mortgage lending, in under ninety seconds, without using the word "haircut."

  2. The composite firm made no bad loans. Is a leverage covenant that binds in a bad quarter a well-designed contract or a procyclical one? Argue both sides, and say what a warehouse bank would answer.

  3. Part 2 argues that broker choice is "granted, not owned." Does that make the broker channel's advantage less real? What would a brokerage do to make its choice more durable?

  4. Compare the March 2020 non-bank liquidity episode with the composite in Part 3. What is the same mechanism and what is different? Which is more likely to recur?

  5. You are a loan officer at the composite firm on the morning of Blow 3. You have eleven files in process, four of them locked. What do you do first, second, and third — and what exactly do you tell the agent on the file closing in nine days?

  6. The chapter is careful never to say one channel is better. Having read both case studies, is that restraint honest or is it evasion? If you think one channel is structurally safer for a loan officer's income across a full rate cycle, name it and defend it.

Sources for the real material

  • Public bankruptcy and securities filings from the 2007 non-bank failures, and contemporaneous reporting.
  • CFPB / FFIEC HMDA data — origination counts by lender type and channel, free and downloadable.
  • Ginnie Mae — issuer eligibility requirements and published issuer lists; the 2020 pass-through assistance program announcements.
  • FHFA — seller/servicer minimum financial eligibility requirements; 2020 servicer advance announcements.
  • Mortgage Bankers Association — origination volume and channel-share series.
  • Regulation Z, §1026.36 (loan originator compensation) — for the Part 2 compensation change; Chapter 26 owns the substance.

Every figure in Part 3 is constructed. Verify all current requirements, capacity, and market conditions at the sources above.