Chapter 28: Where the Money Comes From: Fannie Mae, Freddie Mac, Ginnie Mae, and the Secondary Market

"You are not selling a rate. You are assembling a file that somebody you will never meet has already agreed, in writing, to buy — at a price they published before you picked up the phone." — constructed; the working premise of Part VI

Overview

In the first section of Chapter 1 you were asked to write one sentence in the workbook: if this loan closes, who will own the debt in a year, and who will the borrowers actually pay? You were told you would not be able to answer it, and that being unable to answer it was the point.

This is the chapter where you answer it.

Twenty-seven chapters ago the money for 4412 Linden Street was an abstraction — "the investor," a box at the top of a diagram, a party the borrowers would never meet. Since then you have priced the file, documented it, structured it, survived an appraisal, cleared eleven conditions, absorbed a \$611 furniture payment on day 44, and funded \$365,750.00 at 6.625% on day 51. The note is signed. The security instrument is recorded at the county. And the money that hit the closing table was not your employer's money in any lasting sense — it was borrowed on a warehouse line and it has to be paid back, quickly, by selling the loan.

So: sold to whom? For how much? Into what? And what happens to the borrowers' \$3,033.72 when it leaves their checking account on December 1?

Every one of those questions has a specific, traceable answer, and by the end of this chapter you will be able to give it — the pool, the security, the guarantor, the servicer, the payment date, and the reason each one takes the cut it takes. This is not background. It is the mechanism that wrote every guideline in Part III, every disclosure requirement in Part IV, and every price adjustment in Chapter 29. Loan officers who never learn it spend their careers experiencing the rulebook as arbitrary harassment. Loan officers who do learn it experience the same rulebook as a price list — which is what it actually is, and which is a much more useful thing to be holding when a borrower asks you why.

Chapter 2 gave you the history: how securitization was invented, why the government-sponsored enterprises exist, what a pass-through security was designed to solve, and how the private-label market grew and collapsed. This chapter gives you the plumbing. Same story, different altitude.

In this chapter, you will learn to:

• Follow one funded loan from the closing room into a security held by an institutional investor
• State what Fannie Mae and Freddie Mac do, and the four things they do not do
• Explain why Ginnie Mae is structurally different from both, and why the exam tests it constantly
• Walk the securitization process step by step, naming each party and each document
• Explain the To-Be-Announced market and why it is the reason you can lock a rate at all
• Split a note rate into coupon, guarantee fee, and servicing fee, in dollars
• Compare execution options and value a servicing right, including its inverse relationship to rates
• Explain to any borrower, agent, or new hire why the guidelines are what they are

Learning Paths

🎓 Exam — §28.2 and §28.3 are the highest-yield pages in Part VI. The Fannie/Freddie/Ginnie distinction is tested repeatedly and from several directions. Memorize the comparison table in §28.3 and be able to reproduce it cold. 🏠 New LO — §28.1 and §28.10. Read §28.1 as narrative and §28.10 as your answer to every "why do they need that?" question you will be asked for the rest of your career. 🤝 Partner — §28.1 and §28.8. Agents ask "why did my client's loan get sold?" constantly. Thirty seconds of accurate answer buys enormous credibility. 📊 Operations — §28.4, §28.5, and §28.7. Delivery deadlines, pool certification, and execution choice are where post-closing and capital markets actually meet.

28.1 Following one loan from the closing table to a pension fund

It is Friday, day 51 — October 24. The signing took fifty minutes. The borrowers signed a note promising \$365,750.00 at 6.625% in 360 payments of \$2,341.94 beginning December 1, and a security instrument pledging 4412 Linden Street against that promise. The closing agent disbursed. The seller got paid. The keys changed hands in the parking lot.

Everyone in that room believes the transaction is over. From the money's point of view, it has barely started.

Here is the whole journey, in order. Read it once for the shape; the rest of the chapter takes each step apart.

ONE LOAN, END TO END — the Linden Street file        [constructed teaching example;
                                                      timing after day 51 is illustrative]

  DAY 51  ┌─────────────────────────────────────────────────────────────────┐
  Fri     │ CLOSING · FUNDING · RECORDING                                   │
  Oct 24  │ The closing agent disburses $365,750 of loan proceeds. The       │
          │ money came from the lender's WAREHOUSE LINE — borrowed, at       │
          │ interest, and due back in days. The security instrument is       │
          │ recorded in the county land records. The note is endorsed.       │
          └────────────────────────────┬────────────────────────────────────┘
                                       ↓  the loan now exists as an asset
  +2 to    ┌─────────────────────────────────────────────────────────────────┐
  +7 days  │ POST-CLOSING REVIEW · SALABILITY AUDIT                          │
          │ Post-closing QC confirms the file is what the investor was told  │
          │ it would be: signatures, dates, disclosure timing, the final     │
          │ 1003, the AUS findings, the appraisal, the MI certificate.       │
          │ A missing signature here is a loan nobody will buy.              │
          └────────────────────────────┬────────────────────────────────────┘
                                       ↓
  +1 to    ┌─────────────────────────────────────────────────────────────────┐
  +3 wks   │ COLLATERAL FILE → DOCUMENT CUSTODIAN                            │
          │ The ORIGINAL note (endorsed), a certified copy of the recorded   │
          │ security instrument, the title policy, and the assignment go to  │
          │ a third-party custodian, who certifies the pool. The custodian   │
          │ is not the lender and not the agency. Its whole job is to say    │
          │ "the paper exists and matches the data."                         │
          └────────────────────────────┬────────────────────────────────────┘
                                       ↓
  Nov      ┌─────────────────────────────────────────────────────────────────┐
          │ DELIVERY · POOLING · GUARANTEE                                   │
          │ The CAPITAL MARKETS DESK delivers the loan to Fannie Mae, which  │
          │ groups it with 141 other 30-year fixed loans into a POOL with a  │
          │ $49,700,000 original face. Fannie GUARANTEES timely payment of   │
          │ principal and interest to whoever holds the security, and takes  │
          │ a GUARANTEE FEE for doing it. The pool gets a pool number and a  │
          │ CUSIP. A MORTGAGE-BACKED SECURITY now exists.                    │
          └────────────────────────────┬────────────────────────────────────┘
                                       ↓
  Nov      ┌─────────────────────────────────────────────────────────────────┐
          │ SETTLEMENT — the security is delivered against a trade the desk  │
          │ SOLD ON DAY 12, six weeks before the loan closed (§28.5). The    │
          │ buyer is a broker-dealer, which allocates it onward.             │
          └────────────────────────────┬────────────────────────────────────┘
                                       ↓
          ┌─────────────────────────────────────────────────────────────────┐
          │ THE INVESTOR — a pension fund's core fixed-income allocation.    │
          │ It wanted agency-guaranteed monthly cash flow at a 6.000% coupon.│
          │ It has no idea 4412 Linden Street exists.                        │
          └─────────────────────────────────────────────────────────────────┘

  Dec 1   Borrowers pay $3,033.72 to the SERVICER.
  Dec 25  Certificateholders receive their share of that payment.

Now the same journey with the parties named, because "the money goes to investors" is the kind of sentence that sounds like understanding and is not.

The lender. Your employer funded \$365,750.00 at the table using a warehouse line of credit — short-term borrowing, secured by the loan itself, that Chapter 31 takes apart properly. The warehouse bank charges interest daily. Every day the loan sits unsold costs money, which is why post-closing departments are under real pressure and why a missing initial on page four of the security instrument is not a clerical annoyance but a financial event.

The capital markets desk. Somewhere in the building — or, for a small lender, at a correspondent investor's shop — sits the group that decides where each closed loan goes. They price the loan against several possible exits, commit it for delivery, and manage the hedge that has been sitting against this loan since it was locked. This is the capital markets desk, sometimes called secondary marketing. Its function is to convert closed loans into cash at the best available execution and to make sure the price the loan officer quoted six weeks ago is still the price the company actually realizes.

The document custodian. A third party — often a bank's corporate trust division — that physically holds the original note and certifies to the agency that the paper matches the data file. Agencies do not take a lender's word for the existence of a note. The custodian exists because in a system where debt is sold repeatedly and rapidly, somebody neutral has to be able to say the promise is real and we are holding it.

Fannie Mae. Buys the loan. Puts it into a pool with other loans that share the relevant characteristics. Guarantees the security. Charges a fee for the guarantee. Publishes, in advance and for free, the entire rulebook that determined whether this loan was eligible in the first place.

The servicer. Collects the borrowers' \$3,033.72 every month, keeps a slice of the interest as compensation, remits the rest, pays the property taxes and the homeowners insurance out of escrow, pays the mortgage insurance premium to the MI company, and handles every phone call for the next thirty years. On this file the originating lender retained servicing — a decision worth real money, which §28.8 works out.

The investor. A pension fund, an insurance company, a mutual fund, a bank's securities portfolio, a foreign central bank. It bought a security paying a 6.000% coupon with an agency guarantee. It did not underwrite your borrowers. It could not name them. It is relying entirely on the guarantee and on the fact that the pool contains what the disclosure said it contained.

Notice what is true of every party after day 51: not one of them ever spoke to the borrowers. The only party the borrowers will ever interact with again is the servicer, and the servicer is not the one who owns the debt. This is the structure that produces the single most common post-closing phone call in the business.

📞 On the Phone

Borrower, six weeks after closing: "I got a letter saying my loan was sold. Did we do something wrong? Can they change my rate?"

The answer that does not work: "That's normal, don't worry about it." True, useless, and it leaves a frightened person exactly as frightened.

The answer that works: "Nothing went wrong — this was always going to happen, and it happened on schedule. Here's the actual mechanism. We funded your loan with borrowed money and sold it, the way we told you we would at application. Your note is now inside a pool of loans backing a security that Fannie Mae guarantees. Nothing in your note changed and nothing can: your rate is 6.625% for 360 months, your payment is \$2,341.94 of principal and interest, and the only thing anybody is allowed to change is where you mail it. You'll get two letters, one from us and one from the new servicer, and they have to give you a sixty-day grace period where a payment sent to the old place can't be treated as late. If the two letters disagree about anything, call me before you send money."

That last sentence is the one that matters. Servicing-transfer scams are real, and the borrower who has been told to call you is the borrower who does not wire a payment to a fraudster. Chapter 23 covers the transfer notices and the sixty-day protection in full; your job on this call is to make the structure make sense and to leave the door open.

Here is the answer to the Chapter 1 workbook question, written out plainly.

Who owns the debt in a year? Not your employer. The loan sits in a trust; Fannie Mae is the guarantor; investors hold certificates representing shares of the pool's cash flows. The borrowers' promise has been divided among hundreds of holders, none of whom can identify their share of it.

Who will the borrowers actually pay? The servicer — which, on this file, happens to still be the company whose name was on the note, because your employer retained the servicing. If it later sells that servicing, the borrowers will pay somebody else, and nothing about their loan will change except the address on the coupon book.

And where did the money at the closing table come from? From a warehouse bank, temporarily, and before that and after that from an investor who had already decided — before you took the application — what they would pay for a 30-year fixed loan at this coupon. That decision is why the rate was 6.625% and not something else. Chapter 29 shows you the arithmetic of the quote. This chapter shows you whose arithmetic it is.

28.2 The GSEs: what they do and what they do not

Chapter 2 told you where Fannie Mae and Freddie Mac came from and why. Take the history as read. What matters at the desk is what they actually do — and, far more often the source of trouble, what they do not.

Fannie Mae is the Federal National Mortgage Association. Freddie Mac is the Federal Home Loan Mortgage Corporation. Both are chartered by Congress. Both are shareholder-owned corporations rather than government agencies. Both have been in conservatorship since September 2008, with the Federal Housing Finance Agency as conservator. Both remain in conservatorship as of this writing; this is a live policy question and you should verify the current status before you assert anything about it to a borrower or a client.

For everyday purposes the two are close substitutes. They publish parallel rulebooks, run competing automated underwriting systems, buy overlapping populations of loans, and now issue a common security. Where their guidelines diverge, they diverge in details that matter enormously to a specific file and not at all to the shape of the system.

What they do

They buy loans. This is the fact that separates them from Ginnie Mae and the single most important thing to hold on to. A lender approved as a seller delivers closed loans to Fannie Mae or Freddie Mac, receives cash or securities in exchange, and uses the proceeds to pay down its warehouse line and lend again.

They pool and securitize. Purchased loans are grouped into pools; the pools back securities; those securities are sold into the capital markets. Since 2019 both enterprises issue a common security — the Uniform Mortgage-Backed Security — so that a Fannie pool and a Freddie pool with the same coupon and maturity are deliverable against the same trade. Before that, the two agencies' 30- year securities traded separately and at different prices for identical credit risk, which was a persistent inefficiency the Single Security Initiative was designed to fix.

They guarantee. The enterprise guarantees timely payment of principal and interest to the security holder, whether or not the borrower pays. If the loan goes delinquent, the investor is still made whole on schedule. This transfers borrower credit risk from the investor to the enterprise, which is the entire product. §28.6 prices it.

They publish the rulebook. The Fannie Mae Selling Guide and the Freddie Mac Single-Family Seller/Servicer Guide are public, free, continuously updated, and enormous. They state exactly which loans will be purchased and on what terms. Every eligibility argument you will ever have with an underwriter is ultimately an argument about a sentence in one of those documents.

They operate automated underwriting. Fannie Mae's is Desktop Underwriter (DU); Freddie Mac's is Loan Product Advisor (LPA). Chapter 15 covers what they do and how to read findings.

They set the uniform instruments and the uniform forms. The note and security instrument signed at Linden Street are agency uniform instruments. So is the URLA. So is the appraisal form. The standardization is not bureaucratic tidiness — a security is only sellable at scale if the underlying promises are the same shape.

They transfer credit risk. Since the years after the crisis, both enterprises have sold portions of the credit risk on their guaranteed loans to private investors through structured credit-risk transfer programs. Case Study 1 takes this up, because it is the most interesting thing that happened to the guarantee after 2008 and almost nobody at the origination desk knows it exists.

What they do not do

They do not lend to consumers. Fannie Mae has never made a mortgage loan to a homebuyer. Not once in its history as a secondary-market institution. It buys loans that other people made.

They do not decide on individual borrowers. This is subtler and it trips up experienced people. When DU returns Approve/Eligible on day 6, Fannie Mae did not approve your borrowers. An automated system compared the file as you entered it against published eligibility criteria and returned a recommendation. Your lender's underwriter then decides whether to lend, applying the lender's own overlays on top. "Fannie denied my borrower" is not a thing that happens. What happens is that a file is ineligible for delivery, and your lender declines to make a loan it cannot sell.

They do not set your borrower's rate. They set the terms on which they will buy, which powerfully influences price, but the rate on the note is set by your lender's rate sheet, built from the security market plus the enterprise's fees plus your employer's margin. Chapter 29 rebuilds that quote from base price.

They do not carry the full faith and credit of the United States. Read that sentence twice. Their securities are guaranteed by the enterprises themselves — corporations in conservatorship, supported by Treasury agreements to purchase senior preferred stock, and priced by the market as if government support is reliable. That is not the same legal thing as a federal guarantee, and the difference is the subject of the next section.

They do not insure the borrower's mortgage insurance, and they do not stand behind your underwriting. Mortgage insurance on the Linden Street file comes from a private MI company. And the representations and warranties run toward the enterprise, not away from it: your employer promised Fannie Mae that this file is what it appears to be, and can be required to buy it back if it is not. Chapter 14 is the contractual form of everything in this chapter, and §28.10 closes that loop.

🎓 NMLS Exam Watch

The exam loves the phrase "does not." Expect stems built on these:

• Fannie Mae and Freddie Mac purchase loans from lenders in the secondary market. True.
• Fannie Mae originates loans to qualified borrowers. False, always, in every phrasing.
• The GSEs guarantee timely payment of principal and interest to MBS investors. True.
• GSE securities are backed by the full faith and credit of the United States government. False. That is Ginnie Mae.
• Fannie Mae and Freddie Mac have been in conservatorship since 2008 under the FHFA. True — and note the regulator: FHFA, not HUD, not the CFPB, not Treasury.

One more distinction worth carrying in: the enterprises are shareholder-owned corporations operating under federal charters. They are not federal agencies. Candidates who file them mentally under "the government" get every backing question wrong.

28.3 Ginnie Mae is different

If you learn one thing from this chapter for exam purposes, learn this section. Candidates merge Fannie, Freddie, and Ginnie into a single mental category called "the agencies" and then answer three different questions the same way. They are not the same kind of institution and they do not do the same job.

Ginnie Mae — the Government National Mortgage Association — does not buy loans.

It does not buy loans. It does not sell loans. It does not issue mortgage-backed securities. It does not underwrite, it does not run an automated underwriting system, and it does not publish eligibility guidelines for borrowers, because it never touches a borrower's file.

What Ginnie Mae does is guarantee securities. Specifically: it guarantees the timely payment of principal and interest on mortgage-backed securities that are issued by approved private issuers and backed by pools of loans that are already insured or guaranteed by a federal government program — FHA, VA, USDA Rural Development, and HUD's Section 184 program for Native American lending.

And that guarantee carries the full faith and credit of the United States.

Sit with the structure for a moment, because it is genuinely different in kind.

TWO STRUCTURES, SIDE BY SIDE                          [constructed teaching example]

  CONVENTIONAL — the Linden Street path
  ─────────────────────────────────────────────────────────────────────────
  borrower → lender → FANNIE MAE buys the loan → Fannie pools it →
             Fannie ISSUES and GUARANTEES the security → investor
             Credit protection: private MI (above 80% LTV) + Fannie's guarantee
             Backing of the guarantee: the enterprise, in conservatorship,
             with Treasury support agreements. NOT full faith and credit.

  GOVERNMENT — an FHA or VA path
  ─────────────────────────────────────────────────────────────────────────
  borrower → lender (an APPROVED ISSUER) keeps or acquires the loans →
             the ISSUER pools them and ISSUES the security →
             GINNIE MAE GUARANTEES that security → investor
             Credit protection: FHA insurance or the VA guaranty on each LOAN,
             PLUS the issuer's obligation to advance, PLUS Ginnie's guarantee.
             Backing of Ginnie's guarantee: FULL FAITH AND CREDIT of the U.S.
             Ginnie Mae never owned the loans.

Three consequences follow, and each one is a testable fact.

First, the issuer, not Ginnie Mae, is on the hook first. In a Ginnie Mae structure the approved issuer must advance the scheduled payment to certificateholders even when borrowers do not pay. The issuer's own capital absorbs the timing gap. Ginnie Mae's guarantee stands behind the issuer, not in place of it. This is why Ginnie issuer approval is demanding and why a failing issuer is a serious event: Ginnie can and does take over an issuer's portfolio.

Second, the credit protection is layered. On a conventional loan above 80% loan-to-value, private mortgage insurance covers a portion of a loss and the enterprise absorbs the rest. On a government loan, the underlying FHA insurance or VA guaranty already protects the lender at the loan level before any security-level guarantee applies. Ginnie Mae is insuring the timeliness and certainty of the cash flow, on top of loan-level protection that the government already provided.

Third, Ginnie Mae has no rulebook of its own for borrowers. Eligibility for an FHA loan comes from HUD Handbook 4000.1. Eligibility for a VA loan comes from the VA. Ginnie Mae's own guide governs issuers and pools — who may issue, what may go in a pool, how it is certified, how it is reported. When a new loan officer asks "what are Ginnie Mae's credit score requirements," the correct answer is that there are none, because Ginnie Mae is not in the credit-decision business at any point.

Here is the table to memorize.

Two further details, useful at the desk and occasionally on the exam. Ginnie Mae runs two program structures: Ginnie Mae I, single-issuer pools with a narrow note-rate structure, paying certificateholders on the 15th of the month; and Ginnie Mae II, which permits multiple-issuer pools and a range of note rates, paying on the 20th. Program terms are revised periodically — verify current requirements in the Ginnie Mae MBS Guide rather than trusting a textbook. Ginnie Mae charges the issuer a guaranty fee measured in a small number of basis points; again, verify the current figure at the source rather than quoting one.

🎓 NMLS Exam Watch

This is the distinction the SAFE MLO test returns to most often, and it comes at you from several angles. Train yourself on the verbs.

If the stem says…	The answer is…
"purchases loans from lenders"	Fannie or Freddie
"does not purchase loans"	Ginnie Mae
"guarantees securities backed by FHA and VA loans"	Ginnie Mae
"full faith and credit of the U.S. government"	Ginnie Mae
"government-sponsored enterprise"	Fannie and Freddie, not Ginnie
"in conservatorship since 2008"	Fannie and Freddie, not Ginnie
"wholly owned government corporation within HUD"	Ginnie Mae
"issues the security itself"	Fannie or Freddie (Ginnie's issuers do)

The classic trap stem: "Which entity purchases FHA and VA loans and issues securities backed by them?" Every instinct says Ginnie Mae. Ginnie Mae purchases nothing. If the question insists on a purchaser of government loans, the purchaser is a private issuer or aggregator; Ginnie Mae guarantees the resulting security. Read the verb, not the loan type.

28.4 Securitization, step by step

Chapter 2 told you what securitization is and why it was invented. Here is what actually happens to a closed loan, in sequence, with the artifacts.

Step 1 — the loan is funded and becomes an asset. On day 51 the loan exists. It is carried on the lender's books as an asset, financed by the warehouse line. From this moment a clock runs: warehouse interest accrues daily, and most warehouse agreements limit how long a loan may sit on the line before it becomes ineligible collateral.

Step 2 — post-closing review. The file is audited against the delivery requirements. This is the last chance to catch the thing that makes the loan unsalable: a missing signature, a disclosure delivered outside its window, a final 1003 that does not match the approved 1003, an appraisal transmittal that was never sent to the borrower. A loan that fails here is a scratch and dent loan — salable, but at a discount, sometimes a punishing one.

Step 3 — the collateral file goes to a document custodian. The original note, endorsed; a certified copy of the recorded security instrument; the title policy; the assignment. The custodian verifies the physical documents against the data the lender submitted and certifies the pool. Certification is not a formality — an uncertified pool cannot settle.

Step 4 — pool formation. Loans are grouped by the characteristics the security is sold on: product type (30-year fixed), the pass-through rate, and the pooling criteria the agency publishes. Loans that share a security coupon go together. The pool is assigned a pool number and a CUSIP — the standard security identifier that lets it be traded, cleared, and held like any other bond.

Step 5 — the guarantee attaches. Fannie Mae guarantees timely payment of principal and interest to certificateholders and takes its guarantee fee for doing so.

Step 6 — disclosure. The agency publishes pool-level and, for agency securities, loan-level statistics: coupon, weighted average note rate, weighted average maturity, weighted average loan age, average loan size, credit score and loan-to-value distributions, geographic concentration, occupancy and purpose mix. What it does not publish is anybody's name or address.

Step 7 — settlement. The security is delivered against a trade and paid for. Frequently that trade was executed weeks earlier, before the loans existed at all, which is §28.5.

Step 8 — the cash flows begin. The borrower pays the servicer. The servicer keeps its strip, remits the rest, and the agency passes through to certificateholders on the scheduled date — for the common security, the 25th of the month, reflecting the payment the borrower made on the 1st of that month. That gap is the payment delay, and it is a priced feature of the security, not an administrative accident.

📄 Read the File

text FIGURE 28.1 — "What the investor is actually shown" [constructed teaching example] THE DOCUMENT Pool-level disclosure summary for a 30-year fixed agency pass-through security, as published at issuance. One page. Figures constructed. THE CONTEXT The Linden Street loan was delivered into this pool in November, about three weeks after it funded on day 51. WHAT IT SHOWS Issuer / guarantor Fannie Mae Security type Uniform MBS, 30-year fixed, pass-through Pass-through rate 6.000% Original face $49,700,000 Number of loans 142 (average balance $350,000.00) WA note rate 6.702% WA loan age (WALA) 1 month WA maturity (WAM) 359 months WA loan-to-value 88% WA representative score 741 Largest state 14.6% of the pool balance Occupancy 100% primary residence Payment to holders the 25th of each month Pool number / CUSIP assigned at issuance WHAT IT DOESN'T No names. No addresses. No paystubs, no bank statements, no letters of explanation. The eleven conditions that consumed eighteen days of somebody's life do not appear and never will. It does not say which loans will prepay. It does not say which borrower financed $5,200 of furniture on day 41. It cannot tell you whether the WA 741 score is a tight cluster or a barbell of excellent and marginal files — only the distribution table would. THE DECISION The investor's decision is made on the coupon, the guarantee, the payment delay, and the prepayment characteristics implied by the weighted averages. Not on any borrower. THE LESSON This is the document your loan file becomes. The Linden Street loan is $365,750 of a $49,700,000 pool — 0.736% by balance, and one loan in 142 by count. Everything you documented was to make it indistinguishable from the other 141. That is not an insult to the work. It is the point of the work.

Constructed. Real agency disclosures are published by the issuing enterprise and contain substantially more detail; verify the current disclosure format at the source.

That figure is worth arguing with, because it produces the chapter's most uncomfortable idea.

You spent fifty-one days on this file. You had a two-o'clock deadline on day 0, a mechanic's lien on day 19, eleven conditions on day 28, and a genuine crisis on day 44. And the artifact all of that produced is a line in a table that says the pool has 142 loans instead of 141.

The reason to make peace with that is in the last field of the figure. The investor pays a premium for the pool precisely because it cannot tell the loans apart. Homogeneity is the product. The moment an investor has to wonder whether the loans in a pool were underwritten the way the disclosure implies, the price falls — and every borrower in the country pays for that doubt. Your documentation work is what makes the indistinguishability true. It is quality control on a promise that is going to be sold to strangers.

28.5 The TBA market and why it exists

Here is a thing you do every week that is impossible, unless you understand this section.

A borrower calls on day 12. You lock 6.625% with a half point for thirty days. The loan will not close for another five or six weeks. You have just promised a rate on a loan that does not exist, secured by a house that has not been appraised, to borrowers who have not been approved, out of money that has not been raised.

Somebody has to be on the other side of that promise. The To-Be-Announced market is who.

What a TBA trade actually is

In the TBA market, a buyer and a seller agree to trade agency mortgage-backed securities on a future date without specifying which securities. They agree on exactly six things:

1. The agency or program — Fannie/Freddie common security, or Ginnie Mae
2. The maturity — 30-year, 15-year
3. The coupon — 6.000%, 6.500%
4. The price
5. The par amount
6. The settlement date

That is the entire contract. Which pools will be delivered is to be announced — literally, later, under a notification rule that requires the seller to identify the specific pools by a stated deadline before settlement.

That is a strange-sounding contract until you notice what it makes possible. A lender can sell a security before the loans that will fill it have been originated. The lender is not selling this loan. It is selling "\$3,000,000 of 30-year 6.000% pass-throughs, for November settlement" — a commodity, not an asset. And a commodity can be sold forward.

📄 Read the File

text FIGURE 28.2 — "The trade behind the lock" [constructed teaching example] THE DOCUMENT A capital markets desk's forward sale ticket. One screen. THE CONTEXT Day 12, Monday, September 15 — the same morning the Linden Street borrowers locked 6.625% with a half point for thirty days. The desk is not hedging their loan specifically; it is hedging the day's locked pipeline, of which their loan is one part. WHAT IT SHOWS Trade date Monday, day 12 (September 15) Side SELL Security Uniform MBS, 30-year, 6.000% coupon Settlement November, 30-year conventional settlement class Par amount $3,000,000.00 Price 100-24 (= 100.750) Est. proceeds $3,022,500.00 ($3,000,000 x 1.00750), plus accrued Counterparty a broker-dealer WHAT IT DOESN'T It does not name a single loan. No addresses. No borrowers. No note dates. No pool number, no CUSIP — none of them exist on day 12. The Linden Street loan will not be funded for another thirty-nine days and could still die; on day 44 it nearly does. The ticket does not care. It also does not say what happens if the pipeline does not close as modeled, which is the single largest risk on this desk. THE DECISION Sell forward against the pipeline today, so that today's locked rates are backed by today's security prices. If the desk waits, the lender is long a pipeline of promised rates with nothing behind them. THE LESSON The rate you quoted at 9 a.m. is only real because somebody sold a security at 9:05. Linden Street is $365,750 of this $3,000,000 — 12.19% of one ticket — and it is the reason your borrower's rate survived six weeks of a moving market.

Constructed. Prices, sizes, and settlement conventions vary continuously; the SIFMA Uniform Practices govern good delivery and the settlement calendar. Verify current conventions at the source.

Why this is the mechanism behind your rate lock

Follow the logic once and it will never leave you.

A rate lock is a promise about a future price. When you lock 6.625% on day 12, your employer has committed to fund at that rate on a day that may be forty days away, in a market that will move every one of those days. If rates rise before closing, the loan is worth less than the lender promised — and the lender eats the difference. That is not a small exposure. Across a pipeline of hundreds of locked loans it is company-threatening.

The desk neutralizes it by selling forward. If rates rise, the value of the pipeline falls and the value of the short TBA position rises, roughly offsetting. If rates fall, the reverse. What makes this possible is precisely that TBA contracts do not reference specific loans: the desk can sell against loans that have not closed, and can adjust the position daily as the pipeline changes.

No TBA market, no forward rate lock. In a world where a lender could only sell loans that already existed, a borrower could learn their rate at closing and not before, or would pay a fat premium for a lender willing to warehouse the risk. The thirty-day lock at a quotable price — the single most customer-friendly feature of American mortgage lending — is a downstream consequence of a commoditized forward market in securities.

Two complications you should know about, because they show up as costs on your file.

Pipeline behavior is not certain. Not every locked loan closes. Some borrowers walk, some files die in underwriting, and — this is the important one — when rates fall, borrowers abandon locks to re-lock lower somewhere else. The desk models a pull-through rate and hedges a fraction of the pipeline accordingly. When actual pull-through diverges from the model, the hedge is wrong, and somebody pays.

Files that slip cost the desk money. The Linden Street file was hedged for a closing around day 45 and funded on day 51. When a file misses its window, the desk must either pair off the trade or roll it to a later settlement, and both have a price that moves with the market. The lock extension on day 42 cost 0.250 point — \$914.38 — and the lender absorbed it. That figure is the visible part. The invisible part is the hedge adjustment behind it.

This is theme five of this book stated in capital-markets terms: every day costs money, and the money is real even when nobody sends the borrower a bill for it. Chapter 30 takes locks, extensions, renegotiations, and float-downs apart in full. What you should carry out of this section is the mechanism: your lock is backed by a forward sale of a security that does not yet contain your borrower's loan.

28.6 Guarantee fees

The borrowers pay 6.625%. The security pays a 6.000% coupon. Where does the rest go?

Two places, and the split is the economic heart of the whole system.

THE COUPON STACK — one note rate, three claims       [constructed teaching example;
                                                     g-fee and servicing illustrative]

   BORROWER'S NOTE RATE                                      6.625%
   ──────────────────────────────────────────────────────────────────
   − servicing fee, retained by the servicer                 0.250%
   − guarantee fee, retained by Fannie Mae                   0.375%
   ══════════════════════════════════════════════════════════════════
   = PASS-THROUGH RATE paid to certificateholders            6.000%

   Principal is NOT stacked. Every dollar of principal the borrower
   repays passes through in full. The fees are carved out of INTEREST
   only, and they are charged against the outstanding balance, so both
   the g-fee and the servicing fee shrink as the loan amortizes.

The guarantee fee, universally shortened to g-fee, is what the enterprise charges for standing behind the loan. It is the price of the promise that the investor gets paid on time whether or not the borrower does. Structurally it has two components: an ongoing fee expressed in basis points per year on the outstanding balance, and upfront loan-level price adjustments that vary with the risk characteristics of the specific loan — credit score, loan-to-value, occupancy, property type, purpose, and product. Chapter 29 works the upfront adjustments in detail, because those are the ones that show up on your rate sheet as line items and that you have to explain to borrowers.

The pass-through rate is what is left: the rate at which interest actually passes through to the holders of the security. It is not the borrower's note rate and it never is.

Do not quote a current g-fee from memory or from a textbook. Ongoing guarantee fees are set by the enterprises under FHFA oversight, have been adjusted repeatedly, and are the subject of continuing policy attention. The figures below are constructed so that the arithmetic is clean and visible. Verify current fees and adjustments at the enterprise and at FHFA.

🧮 Run the Numbers

Where the first month's interest actually goes. [constructed teaching example — g-fee and servicing fee illustrative; verify current figures at the source]

The Linden Street loan: \$365,750.00 at 6.625%. First payment December 1. Of the \$2,341.94 of principal and interest, canon gives us interest of \$2,019.24** and principal of **\$322.70.

Now split the interest three ways. Each strip is its annual rate divided by 12, applied to the \$365,750.00 opening balance:

Claim	Annual rate	Month 1	Share of interest
Certificateholders (the coupon)	6.000%	\$1,828.75	90.57%
Fannie Mae (guarantee fee)	0.375%	\$114.30	5.66%
The servicer (servicing fee)	0.250%	\$76.20	3.77%
Total interest	6.625%	\$2,019.24	100.00%

Check the shares directly against the rates: $6.000 \div 6.625 = 90.57\%$, $0.375 \div 6.625 = 5.66\%$, $0.250 \div 6.625 = 3.77\%$. They sum to 100.00%, which they must, because the three strips are the note rate.

A one-cent honesty note. The three rounded components sum to \$2,019.25, one cent more than the \$2,019.24 of interest actually collected. Nothing is wrong: the unrounded strips are \$1,828.750, \$114.297, and \$76.198, totaling \$2,019.2448. Rounding three numbers separately and then adding them is not the same as adding them and then rounding. Remittance systems carry more decimal places than a textbook table does, and a loan officer who understands this will not panic the first time a reconciliation is off by a penny.

Annualized, at the opening balance: the coupon is \$21,945.00 a year, the g-fee \$1,371.56, and servicing \$914.38. Because the balance amortizes, the actual first twelve months are slightly smaller. Canon gives the balance after twelve payments as \$361,757.88, so principal paid in year one is \$3,992.12 and total interest is $12 \times \$2{,}341.94 - \$3{,}992.12 = \$24{,}111.16$. Apply the same 5.66% share and the true first-year guarantee fee is \$1,364.78.

Over thirty years, if the loan ran to term, the guarantee fee would take that same 5.66% share of the loan's \$477,348.40 of total interest — $\$477{,}348.40 \times 0.375 \div 6.625 =$ \$27,019.72. The borrower will never see that number on any disclosure. It is inside the rate.

Two of these figures collide with numbers you have seen before. \$1,828.75 is also what the borrower's half discount point cost at closing, and \$914.38 is also what the fifteen-day lock extension cost on day 42. The collisions are pure arithmetic — every one of these quantities is a fixed percentage of the same \$365,750, and 0.500% and 0.250% of that principal are \$1,828.75 and \$914.38 no matter what you are measuring. A monthly interest flow, a one-time discount charge, and a lock-extension fee are three different things that happen to share a number. Never reuse a figure because it looks familiar.

Three things worth understanding about g-fees beyond the arithmetic.

The g-fee is a credit-risk price, and it moves with policy as well as with markets. It has been used deliberately as a lever — raised broadly after the crisis to rebuild capital, restructured to shift cost across borrower segments, and adjusted for particular loan characteristics to encourage or discourage certain lending. That is why you cannot learn a number; you can only learn the structure.

The borrower never sees it as a fee. No line on the Loan Estimate says "guarantee fee." It is embedded in the rate. When a borrower asks why the number at the top of an advertisement is not their number, part of the answer is the loan-level adjustments in Chapter 29 — and those adjustments are the upfront form of this fee.

The g-fee is the reason the guideline is enforceable. The enterprise is pricing risk it has agreed to bear. It bears that risk on the assumption that the file is what the seller said it is. When the file is not, the enterprise's remedy is not to raise the fee retroactively — it is to make the seller buy the loan back. Chapter 14 named that mechanism. §28.10 is where it comes home.

28.7 Whole loan sales vs. securitized execution

A closed, salable loan has more than one exit. The choice among them is made by the capital markets desk, it is made in dollars, and it explains a surprising amount of what a loan officer experiences as arbitrary lender behavior.

Whole loan sale. The lender sells the loan itself — the entire asset, note and all — to a buyer, for cash, at a price expressed as a percentage of the unpaid balance. The buyer may be an aggregator, a bank, or one of the enterprises under a cash program. Simple, fast, and final: the lender's exposure ends and the cash arrives. A whole loan sale is usually servicing released, meaning the buyer gets the right to service the loan, though servicing-retained whole loan sales also exist.

Securitized execution. The lender delivers the loan into a pool, receives the security, and sells the security in the capital markets. More steps, more infrastructure, more time, and access to the liquidity of the TBA market. Securitized execution is typically servicing retained: the lender keeps the servicing strip and books a mortgage servicing right as an asset.

The desk does not choose on principle. It chooses on total economic value, computed on the same day against the same loan.

🧮 Run the Numbers

Two exits, same loan, same day. [constructed teaching example — all prices illustrative; secondary market execution changes daily and by loan]

The Linden Street loan, \$365,750.00 at 6.625%, priced for delivery several weeks after the trade ticket in §28.5, at a different market level.

Option A — whole loan sale, servicing released, at a price of 101.500. This is the illustrative execution Chapter 1 used in §1.3, reused here unchanged.

Proceeds: \$365,750.00 × 1.01500 | **\$371,236.25**
Gain over par	\$5,486.25
Servicing	sold with the loan; the lender keeps nothing
Total economic value	\$371,236.25, all of it cash

Option B — securitized, servicing retained. Security price 100-19 (= 100.59375).

Security proceeds: \$365,750.00 × 1.0059375 | \$367,921.64
Gain over par	\$2,171.64
Plus the retained MSR, marked at 112.5 bps of balance (§28.8)	\$4,114.69
Total economic value	\$372,036.33
Gain over par	\$6,286.33

Option B is better by \$800.08.

Prove it a second way, in points, which is how the desk actually thinks. The whole-loan bid of 101.500 already includes the servicing, because the buyer is getting it. Strip the servicing's 1.125 points of value out and the whole-loan bid is worth 100.375 to a lender who intends to keep servicing. The securitized route pays 100.59375. The difference is $100.59375 - 100.375 = 0.21875$ points, and 0.21875% of \$365,750.00 is **\$800.08**. The two methods agree, which is the only reason to trust either.

Now the part that matters. Option A produces \$371,236.25 of cash this week. Option B produces \$367,921.64 of cash and \$4,114.69 of a booked asset that pays out over years, requires a servicing operation to realize, must be marked to market every quarter, and — as §28.8 shows — can lose a quarter of its value in a rate rally. A well-capitalized lender that wants the customer relationship takes Option B. A thinly capitalized lender that needs cash to fund next week's loans takes Option A and is not being foolish. They are pricing the same loan and buying different things.

Two consequences a loan officer feels directly.

Execution choice is why your employer's pricing on one product can look strange. If your company's capital markets desk has strong securitized execution on 30-year conforming and only a mediocre whole-loan outlet for, say, high-balance or manufactured-home loans, your rate sheet will be sharp on the first and dull on the second. That is not the underwriting department's opinion of the product. It is the exit.

Execution choice is also why servicing-released premiums move. A loan officer who notices that the company suddenly wants every loan delivered a particular way, or that a niche product got fifty basis points worse overnight with no change in the market, is watching an execution decision. Ask. The answer is usually mundane and always instructive.

28.8 Servicing retained, servicing released, and MSRs

Every borrower eventually asks some version of the same question: who am I going to be paying? The honest answer requires understanding that servicing is not an administrative afterthought. It is a separate asset with its own market, its own valuation, and — most usefully for teaching — a value that moves in the opposite direction from most people's intuition.

The two words

Servicing released means the originating lender sells the right to service the loan along with, or shortly after, the loan itself. The buyer pays a servicing released premium for it. The borrower will be making payments to somebody new.

Servicing retained means the originating lender keeps the right to collect the payments and keeps a slice of the interest as compensation. The borrower keeps paying the same company. The lender books an asset.

Neither is better in the abstract. Retaining servicing is a bet on the future — a claim on a stream of small payments stretching out over years — and it requires capital, systems, licensing, compliance infrastructure, and the willingness to answer the phone when a borrower's furnace dies in February. Releasing servicing converts that stream into cash today at somebody else's price.

What the asset is

A mortgage servicing right (MSR) is the contractual right to service a loan and to receive the servicing fee, together with the obligations that come with it. On the Linden Street file the servicing fee is an illustrative 0.250% per year on the outstanding balance — \$914.38 in the first year, falling every month as the loan amortizes, and disappearing entirely the day the loan pays off.

Against that revenue run real costs: the cost to service a current, performing loan (an illustrative \$70.00 per loan per year — verify; it varies by servicer and rises sharply for delinquent loans), plus the obligation to advance payments and to fund escrow shortfalls. Alongside it run ancillary revenues: late fees, and the economic value of holding escrow balances. The market values the whole package, not the fee alone.

MSRs are quoted two ways, and they are the same statement:

• As a multiple of the annual servicing fee — "4.5 times"
• As basis points of the unpaid principal balance — "112.5 basis points"

They convert directly: $0.250\% \times 4.5 = 1.125\% = 112.5$ basis points.

Why the value goes UP when rates go up

This is the counterintuitive part, and it is worth slowing down for because most people get it exactly backwards on the first pass.

An MSR is a claim on payments that only exist while the loan exists. The moment the borrower refinances, sells, or otherwise pays off, the servicing fee stops forever. So the value of an MSR depends overwhelmingly on how long the loan is expected to survive.

Now ask what happens to loan survival when rates move.

Rates fall. Every borrower with a 6.625% note has a reason to refinance. Prepayments accelerate. The expected life of the loan collapses from years to months. The stream of \$914.38-a-year payments you were counting on ends early. The MSR is worth less.

Rates rise. Nobody refinances a 6.625% loan into a 7.75% market. Nobody sells a house they cannot replace. The loan sits there paying, year after year, past every model's expectation. This is the so-called lock-in effect, and it is exactly what makes servicing valuable. The MSR is worth more.

So the asset behaves the opposite way from a bond. A bond loses value when rates rise. An MSR gains value when rates rise. That is precisely why lenders hold them: servicing is a natural hedge against the origination business. When rates rise, origination volume collapses, refinances vanish, margins compress, and everybody's pipeline empties — and the servicing book appreciates. When rates fall, origination booms and the servicing book bleeds value as the loans in it run off. A lender that both originates and services owns two businesses that fail in opposite weather.

🧮 Run the Numbers

What the Linden Street servicing right is worth, and what a 100-basis-point move does to it. [constructed teaching example — multiples and cost-to-service illustrative; MSR marks move continuously and vary by servicer, product, and vintage]

The revenue. Servicing fee 0.250% per year on the balance: \$365,750.00 × 0.00250 = **\$914.38 in year one. Less an illustrative \$70.00 annual cost to service, the net is about \$844.38** — before ancillary income and escrow value, and before any allowance for what a delinquent loan costs.

The mark. At an illustrative 4.5× multiple:

Multiple	4.5×
As basis points of balance	0.250% × 4.5 = 1.125% = 112.5 bps
Dollar value	\$365,750.00 × 0.01125 = **\$4,114.69**

Now move rates 100 basis points, and move the multiple with them.

Scenario	Multiple	bps of balance	MSR value	Change
Rates fall 100 bp — prepayments accelerate	3.25×	81.25 bps	\$2,971.72 | **−\$1,142.97**
Today	4.5×	112.5 bps	\$4,114.69	—
Rates rise 100 bp — the loan sits	5.25×	131.25 bps	\$4,800.47 | **+\$685.78**

Check the middle column: 0.250% × 3.25 = 0.8125%, and 0.8125% of \$365,750.00 is \$2,971.72. 0.250% × 5.25 = 1.3125%, and 1.3125% of \$365,750.00 is \$4,800.47.

Read the asymmetry. A 100-basis-point rally costs \$1,142.97. A 100-basis-point selloff gains \$685.78. The downside is 1.67 times the upside for the same size move, because prepayment speeds can accelerate almost without limit while the lock-in benefit saturates — a borrower who was never going to refinance cannot become more not-going-to-refinance. This is negative convexity, and it is the defining behavior of mortgage assets. It is also, incidentally, why mortgage-backed securities themselves yield more than Treasury securities of similar maturity: the investor is being paid to accept an instrument that shortens exactly when they wish it would not.

The scale. \$4,114.69 on one loan sounds trivial. A servicer with 100,000 loans of this size is holding an asset worth roughly \$411 million — and this table says it gains about 16.7% (\$685.78 ÷ \$4,114.69) in a 100-basis-point selloff and loses about 27.8% (\$1,142.97 ÷ \$4,114.69) in a 100-basis-point rally. Servicing valuation is not a back-office topic.

What to tell the borrower

Borrowers ask whether it costs them anything. It does not, and the answer is worth having ready: Your note doesn't change either way — same rate, same payment, same term, no prepayment penalty. What changes is who you call. If we keep it, you call us. If it's sold, you'll get two letters and a sixty-day window where a payment sent to the wrong place can't be reported late, and the new servicer takes over exactly as-is. One piece of advice either way: whoever ends up servicing it, set the payment up yourself rather than assuming an automatic draft carried over. That is where people actually get burned, and it has nothing to do with who owns the loan.

What you must never do is tell a borrower you will service their loan forever. You do not control that decision, MSRs are sold routinely, and a promise you cannot keep is worse than no promise at all.

28.9 Non-agency: jumbo and private-label

Everything to this point describes the agency market: loans eligible for purchase by Fannie Mae or Freddie Mac, or eligible for a Ginnie Mae pool. It is the largest, most standardized, most liquid part of the system, and it is where most loans you originate will go.

Then there is everything else.

Non-agency simply means a loan that will not be bought by an enterprise or pooled behind a Ginnie Mae guarantee. Loans land there for several unrelated reasons:

• Size. The loan exceeds the applicable conforming loan limit. This is a jumbo loan. Limits are set annually by FHFA, vary by county, and are higher in designated high-cost areas — verify the current limits before you quote anything near one.
• Documentation or product. Bank statement income, asset depletion, interest-only, debt-service coverage on an investment property, or any structure outside the Qualified Mortgage definition. Chapter 34 is the non-QM chapter.
• Credit. Recent credit events, or ratios and layered risk beyond agency tolerance.
• Choice. A bank may simply want the loan on its own balance sheet, often to hold a relationship with a valuable depositor.

Non-agency loans have two possible homes.

Portfolio. A depository keeps the loan on its own balance sheet and holds the credit risk itself. This is the oldest form of mortgage lending and it never went away. Portfolio lenders can write their own rules, make genuine exceptions, and move quickly — which is why the answer to a truly unusual file is sometimes a local bank rather than a better agency argument.

Private-label securitization. A private trust — not an agency — issues securities backed by a pool of loans. There is no agency guarantee. Nobody promises the investor gets paid on time. So the credit risk has to be handled inside the structure itself, by subordination: the deal is cut into tranches, and losses are absorbed from the bottom up, so that the senior tranche is protected by everything underneath it. Investors in the junior tranches are paid more precisely because they take the first losses.

That difference is the whole difference, and it is worth stating as plainly as possible.

WHO ABSORBS A LOSS                                     [constructed teaching example]

  AGENCY MBS                              PRIVATE-LABEL MBS
  ────────────────────────────────        ────────────────────────────────
  Borrower defaults.                      Borrower defaults.
       ↓                                       ↓
  MI pays its coverage (if any).          Loss flows to the trust.
       ↓                                       ↓
  The ENTERPRISE makes the                The most SUBORDINATE tranche
  investor whole, on schedule.            absorbs it. Then the next.
       ↓                                       ↓
  The investor's cash flow is             The senior tranche is protected
  UNAFFECTED by the default.              only until the subs are gone.
       ↓                                       ↓
  The enterprise pursues the seller       Nobody is made whole by anybody.
  under reps and warrants (Ch. 14).       The structure allocated the loss.

The consequence for pricing and guidelines is direct. In the agency market, the rulebook is written once, published free, and applies to every seller. In the non-agency market, whoever is going to buy the loan writes the rules, and there are many of them. Jumbo guidelines are not uniform: one investor wants twelve months of reserves and another wants six; one caps debt-to-income at 43% and another will go higher with compensating factors; one will lend to a self-employed borrower on two years of returns and another wants three. There is no Selling Guide for jumbo. There is a matrix per investor, and it changes.

Jumbo pricing is likewise not predictable from first principles. It is sometimes better than conforming — a bank chasing an affluent depositor relationship may price aggressively — and sometimes meaningfully worse when securitization execution tightens. Both are normal.

The private-label market's history is the subject of Case Study 2. The short version, and it is directional rather than statistical: private-label issuance grew enormously before 2008, collapsed almost entirely in the crisis, and has returned in a form that is smaller, better documented, and concentrated in prime jumbo, expanded-credit/non-QM, and investor-property deals. Post-crisis structural changes — credit risk retention requirements, loan-level disclosure, and third-party due-diligence review of the loans in a deal — were designed to fix specific failures of the earlier market. Whether they suffice is a genuinely open question and not one this book will settle.

⚠️ Where Deals Die

The loan that becomes a jumbo loan on day 30.

Here is the mechanism, and it happens every year in every market with rising prices. You take an application near the conforming limit. Everything prices as conforming. Then something moves the loan amount up: the appraisal comes in high and the borrower wants to keep more cash, the seller concession falls through, the buyer wins a bidding war on an escalation clause, or a repair negotiation increases the price. The loan crosses the limit.

What the borrower thinks happened: the loan got a little bigger.

What actually happened: the rulebook changed. The file left a market with one published guideline set and entered a market where every buyer writes their own. Reserve requirements may jump. The maximum loan-to-value may drop, which can mean more cash to close. Documentation standards may tighten. The pricing may move in either direction and by a lot. The AUS findings you have are not evidence of anything anymore. And the timeline resets, because you are re-underwriting to a different investor's matrix.

What the disciplined loan officer does: knows the applicable limit for the county before the first quote, states out loud at application where the ceiling is, and re-checks the loan amount every single time the purchase price, the appraisal, or the down payment moves. If a file is within \$15,000 of the limit, treat it as a jumbo file for planning purposes and be pleasantly surprised.

Verify current conforming and high-cost-area limits at FHFA; they change annually and by county.

28.10 Why any of this is your problem

Now the argument the whole book has been building toward. It comes in one sentence, and then it takes the rest of the section to earn it.

Every guideline in Part III, every disclosure requirement in Part IV, and every price adjustment in Chapter 29 exists because somebody several steps removed from your borrower is deciding the terms on which their money will show up.

Reread the last twenty-seven chapters through that lens and the rulebook reorganizes itself.

Why must income be stable and likely to continue for three years? Because the investor is buying a thirty-year cash flow and the enterprise has guaranteed it. A borrower whose income stops in year two is a loss the enterprise absorbs and then charges everyone else for. The three-year continuance test is a purchase term.

Why does a \$4,900 deposit need a letter and a source document? Because an undocumented deposit might be a loan, an undisclosed loan changes the debt-to-income ratio, and the file was sold on a represented debt-to-income ratio. It is not suspicion of your borrower. It is a warranty the seller made in writing.

Why is there a pre-closing credit refresh — condition 11 on the day-28 approval, written sixteen days before it caught the furniture account? Because the seller warranted the borrower's obligations as of the note date. A \$611 payment that appears on day 41 makes the warranty false. Condition 11 is not an insult to your borrowers; it is a term of sale.

Why does the disclosure timing matter to the penny and the day? Because a loan with a TRID violation may be unsalable, or salable only after a cure, or repurchasable later. Compliance is not merely the license — it is also, quite literally, salability.

Why does the representative score of 706 price differently from 742? Because the enterprise's upfront guarantee-fee adjustments are set on the representative score and the loan-to-value together. That is the g-fee from §28.6, charged upfront instead of monthly, and Chapter 29 shows you the grid.

And why is the underwriter so unreasonable about the thing you consider trivial? Because Chapter 14's representations and warranties are the contractual form of everything in this chapter. Your employer promised the enterprise that this file is what it appears to be. If it is not, the remedy is a repurchase demand: your employer buys the loan back, at par, out of its own capital, possibly years later, possibly on a loan that has already gone delinquent. A repurchase is not a fine. It is being handed back an asset nobody wants at a price nobody would pay. Enough of them kill a mortgage company, and they have.

The underwriter is not being difficult. The underwriter is the last human being who can prevent a warranty from being false.

🔍 Check Your Understanding

1. A borrower asks whether Fannie Mae "approved" their loan. Rewrite that sentence three ways so it is accurate — once from the AUS's point of view, once from the lender's, and once from the borrower's.
2. Your file has a 706 representative score and 95% loan-to-value. Name the single mechanism that connects those two facts to the borrower's interest rate, and name the chapter that computes it.
3. An underwriter conditions for a signed 4506-C that the borrower thinks is pointless. Justify the condition in one sentence that mentions neither "policy" nor "requirement."

(3 is the one worth practicing out loud. Something like: "Whoever buys this loan was told your tax returns say what your application says, and this is the form that lets us prove it if they ever ask.")

What this buys you, practically

Four things, and they are worth real money over a career.

You can explain your own pricing. A loan officer who can say "your rate is what it is because the enterprise charges more for a 706 at 95% loan-to-value, and here is the line item" is a professional. One who says "that's just what the rate sheet says" is a switchboard.

You can tell a guideline from an overlay. Chapter 14 drew that line. This chapter tells you why it exists. An agency guideline is a term of purchase and your lender cannot waive it — nobody at your company has the authority, because the authority belongs to the buyer. An overlay is your employer's own additional rule, adopted for its own reasons, and it can sometimes be excepted, escalated, or avoided by taking the file to a different lender. Knowing which one you are looking at is the difference between an hour of productive escalation and an hour of pointless argument.

You can predict conditions before they are issued. Once you can see the chain, the condition list stops being a surprise. You know what the buyer needs proven, so you know what the underwriter will ask for, so you can collect it on day five instead of day thirty-three. That is the whole discipline of Chapter 19, and it is downstream of this chapter.

You can tell the truth to a borrower about what happens next. Which is not a small thing. The borrower who understands at application that their loan will be sold, that their servicer may change, that nothing in their note can change, and that they should call you if two letters disagree, is a borrower who is not frightened in February and who refers you in April.

That is theme six of this book, and it is the reason Part VI sits where it does. Somebody else's money is at risk. The guidelines are the terms on which that money is willing to show up. Once you can see the chain — kitchen table to closing table to custodian to pool to security to pension fund — the rulebook stops being arbitrary and becomes what it actually is: a price list, published in advance, by people who have never met your borrower and have already decided exactly what they will pay for the promise your borrower is about to make.

🗂️ The Loan File

Chapter 28 contribution: trace the loan out of the building.

The file closed on day 51. Everything below happened after the borrowers stopped thinking about it. Post-closing timing and all execution figures are illustrative and labeled; the loan terms are the frozen file.

Where the money came from. Three answers, all true at different moments. At the closing table, a warehouse bank. In the medium term, the investors who bought the security in November. In the sense that matters to a loan officer, from an investor who had already published — before day 0 — the exact terms on which they would fund a 30-year fixed loan to a 706 representative score at 95% loan-to-value. Those terms are why the rate was 6.625%.

Who owns the debt now. Not your employer. The loan sits behind a Fannie Mae–guaranteed security; investors hold certificates entitling them to shares of the pool's cash flows. The Linden Street loan is 0.736% of a \$49,700,000 pool and one loan out of 142. The recorded lien is still at the county in the name shown at closing, with subsequent transfers of the beneficial interest tracked electronically rather than by recording a new assignment for each sale.

Who the borrowers pay. The servicer — on this file, still the originating lender, because servicing was retained. The MSR is carried at an illustrative \$4,114.69.

Where the first payment goes. Follow the \$3,033.72 on December 1:

Check the P&I line: \$1,828.75 + \$114.30 + \$76.20 + \$322.70 = \$2,341.95, one cent above the \$2,341.94 payment, for the rounding reason explained in §28.6. Check the whole column: \$2,341.94 + \$385.00 + \$130.00 + \$176.78 = \$3,033.72. ✓

What this settles: the Chapter 1 question. You can now name every party that touched this money and say what each of them took and why.

What it does not settle: what your employer actually earned on this file, and what you earned. Chapter 29 rebuilds the 6.625% quote from base price so you can see where each basis point came from, Chapter 30 covers what the day-42 lock extension really cost, and Chapter 31 puts the warehouse line and the business model together.

Open questions carried forward:

• Q. Exactly which adjustments moved this file's price, and by how much? (Chapter 29)
• Q. What did the lock, the extension, and the six-day overrun cost, and who paid? (Chapter 30)
• Q. Would a broker or a correspondent have had a different execution on this same loan? (Chapter 31)

Your task. In the Appendix C workbook, complete the "Where the money went" page: name the pool, the guarantor, the coupon, the servicer, the payment dates, and the three claims on the interest. Then write two sentences you would actually say to these borrowers in February when they call about a letter they do not understand. Read them out loud. If either sentence contains the word "securitization," rewrite it.

Conclusion

The money was never your employer's. It was borrowed on a warehouse line to reach the closing table, and it was repaid within weeks by selling the loan into a market that had already priced it. Follow it forward and you get a specific chain: closing agent, post-closing, document custodian, capital markets desk, Fannie Mae, a pool, a guarantee, a CUSIP, a broker-dealer, an investor, and a servicer collecting \$3,033.72 on the first of every month for thirty years.

Fannie Mae and Freddie Mac buy loans, pool them, guarantee the securities, and publish the rulebook. They are shareholder-owned corporations under federal charter, in conservatorship since September 2008, and they do not carry the full faith and credit of the United States. Ginnie Mae buys nothing and issues nothing; it guarantees securities issued by approved private issuers and backed by government-insured or guaranteed loans, and its guarantee does carry full faith and credit. Merge those three and you will miss exam questions and mislead borrowers.

The borrower's 6.625% note rate is not one number. It is a 6.000% security coupon, a guarantee fee, and a servicing fee, stacked. The servicing fee is itself an asset that gains value when rates rise and loses it when they fall, which is why lenders who both originate and service own two businesses that fail in opposite weather. And the whole apparatus can quote a borrower a rate six weeks before closing only because the To-Be-Announced market lets a lender sell a security containing loans that do not yet exist.

Which is the real point of Part VI. Guidelines are not obstacles. They are the terms on which somebody else's money is willing to show up, published in advance, enforced by a repurchase obligation, and priced to the basis point. Know whose money it is and what they require, and the rulebook stops being arbitrary.

Next: you now know where the price comes from. Chapter 29 shows you how it is built — base price, the loan-level adjustments that turned an advertised number into 6.625% with a half point for these borrowers, and how to read a rate sheet well enough to quote from it honestly.

Key Terms

Fannie Mae (Federal National Mortgage Association) — a government-sponsored enterprise, shareholder-owned under a federal charter, that purchases conventional loans from approved sellers, pools and securitizes them, guarantees timely principal and interest to investors, and publishes the Selling Guide. In conservatorship under FHFA since September 2008. (Ch.28)

Freddie Mac (Federal Home Loan Mortgage Corporation) — the parallel government-sponsored enterprise, operating the Single-Family Seller/Servicer Guide and Loan Product Advisor. Also in conservatorship under FHFA since September 2008. (Ch.28)

Ginnie Mae (Government National Mortgage Association) — a wholly owned government corporation within HUD that does not buy loans and does not issue securities; it guarantees the timely payment of principal and interest on securities issued by approved private issuers and backed by government-insured or guaranteed loans (FHA, VA, USDA RD, §184). Its guarantee carries the full faith and credit of the United States. (Ch.28)

Mortgage-backed security (MBS) — a security whose payments are supported by the principal and interest collected on a pool of mortgage loans. (Ch.28)

Pool — the specific group of loans backing one security, identified by a pool number and a CUSIP and described to investors by weighted-average statistics rather than by individual loans. (Ch.28)

Pass-through rate — the rate at which interest actually passes through to security holders: the borrower's note rate minus the servicing fee minus the guarantee fee. On the Linden Street illustration, 6.625% − 0.250% − 0.375% = 6.000%. (Ch.2 gave the pass-through structure; Ch.28 gives the arithmetic.) (Ch.28)

Coupon — the stated interest rate a security pays its holders; for an agency pass-through, the pass-through rate. Never the borrower's note rate. (Ch.28)

Guarantee fee (g-fee) — what an enterprise charges for guaranteeing timely payment to investors; an ongoing fee in basis points per year on the outstanding balance plus upfront loan-level price adjustments. Embedded in the rate; never a line item to the borrower. (Ch.28)

To-Be-Announced (TBA) market — the forward market in agency mortgage-backed securities, in which a trade specifies only agency, maturity, coupon, price, par amount, and settlement date; the specific pools are identified later. It is what allows a lender to sell a security before the loans in it exist, and therefore what makes a forward rate lock possible. (Ch.28)

Agency / non-agency — agency loans are eligible for purchase by Fannie Mae or Freddie Mac or for a Ginnie Mae pool; non-agency loans are everything else, and go to a portfolio lender or a private-label securitization. (Ch.28)

Whole loan sale — the sale of the loan itself for cash at a price expressed as a percentage of the balance, as distinct from delivering it into a security. (Ch.28)

Servicing released / servicing retained — whether the originating lender sells the right to service the loan (usually for a servicing released premium) or keeps it and collects the servicing fee. (Ch.28)

Mortgage servicing right (MSR) — the contractual right to service a loan and collect the servicing fee, together with its obligations; valued on expected future cash flows, quoted as a multiple of the annual fee or in basis points of balance, and worth more when rates rise and less when rates fall. (Ch.28)

Capital markets desk — the group inside a lender (also called secondary marketing) that prices locks, hedges the pipeline, chooses each loan's execution, and delivers closed loans for sale. (Ch.28)

Document custodian — the neutral third party that holds the original note and certifies that the physical collateral file matches the data delivered to the agency. (Ch.28)

Spaced Review

1. (Ch.2 + Ch.28) Chapter 2 gave you the history: Ginnie Mae was created in 1968 and Freddie Mac in 1970. Using this chapter's mechanics, explain in three sentences why two institutions created two years apart do such structurally different jobs — and say which of them has never owned a mortgage loan.

2. (Ch.14 + Ch.28) A repurchase demand arrives on a loan your company sold three years ago: the borrower's income was overstated in the file. Explain, in the chain of parties from §28.1, exactly who is demanding what from whom, and why the enterprise's guarantee is the reason the demand exists at all.

3. (Ch.28) A borrower's note rate is 6.625% and the security backing it pays a 6.000% coupon. Account for the missing 0.625% and say who receives each piece. Then state what happens to each piece when the borrower's balance falls.

4. (Ch.28) Rates drop 100 basis points. Your company's origination pipeline triples. What happens to the value of your company's servicing book, and why is that a feature of the business model rather than a problem?

5. (Ch.14 + Ch.28) A processor tells you a condition is "just an overlay." Describe how you would determine whether that is true, and explain what your options are in each case — first if it is an overlay, then if it is an agency guideline.