Chapter 14: Underwriting Guidelines: Fannie Mae, Freddie Mac, and the Conventional Rulebook

"Nobody in this building wrote the rule you are arguing with. Find out who did, and read what they actually said." — constructed; the most useful sentence an underwriting manager ever says to a first-year loan officer

Overview

Thirteen chapters have been spent assembling facts. You know what the borrowers earn and how much of it counts. You know what they owe and what an underwriter will make of it. You know what is in their accounts and where it came from. You have chosen a structure. The file is real.

Now somebody has to say yes.

The question this chapter answers is the one every new loan officer eventually asks out loud, usually in frustration, usually on a Thursday: who decides what the rules are, and where can I read them myself? The answer is better than you expect. For conventional lending, the rules are written by Fannie Mae and Freddie Mac, they are published in full, they are free, they are searchable, and almost nobody in the business has read more than a few pages of them. A loan officer who will spend two hours a month in the Selling Guide will, within a year, know more about what is actually allowed than most of the people around them — including, on specific questions, the people telling them no.

But the answer has a second half, and it is the half that ruins first-year loan officers. Your employer is allowed to be stricter than the agency, and frequently is, and does not publish those extra rules anywhere you can find them. A borrower is declined. The loan officer, who checked the Selling Guide and found the file eligible, concludes the guide was wrong or the underwriter was unreasonable. Neither is true. The file hit a rule the agency never wrote.

So this chapter teaches four things and one habit. The four things: how the rulebook is organized and how to read it; the difference between eligibility and creditworthiness, which is the distinction that organizes everything else; layered risk, which is the most useful concept in underwriting and the one most often explained badly; and overlays, which are the practical payoff. The habit is asking, every single time a file is declined: is that the agency's rule, or ours?

We will also answer the question underneath all of it. Underwriters are careful to a degree that looks, from the outside, like obstruction. There is a reason, it is written down, and it is called a representation and warranty. Once you understand it, you will stop taking underwriting personally, and you will start writing files that are easy to say yes to.

In this chapter, you will learn to:

• Locate and read the published conventional guidelines, and cite them accurately
• Restate the four Cs as the questions an underwriter actually asks
• Classify any decline as an eligibility problem or a creditworthiness problem
• Describe the structure of credit-event waiting periods without inventing a value
• Explain layered risk as compounding, and inventory it on a real file
• Separate a guideline from an overlay and find out which one you are up against
• Say what manual underwriting is and what changes when it appears
• Assemble compensating factors that an underwriter can actually use
• Explain reps and warrants, and why they explain the underwriter

Learning Paths

🎓 Exam — §14.3 and §14.4. The eligibility/creditworthiness distinction shows up in question stems constantly, and the exam loves the measuring date on a waiting period. Do not memorize the waiting-period numbers from this book; memorize the structure and verify the values. 🏠 New LO — §14.7 and §14.9. Overlays will cost you three deals in your first year unless somebody warns you, and this is the warning. §14.9 is how you argue a close file and win. 🤝 Partner — §14.3 and §14.7. An agent who understands that "declined" has two completely different meanings will stop assuming every decline is final, and will stop assuming any of them are negotiable. 📊 Operations — §14.5, §14.8, and §14.10. The 1008, the manual-underwriting standard, and the repurchase exposure that sets your quality-control appetite.

14.1 Who writes the rules and where to read them

Chapter 1 traced the money from a pension fund to a kitchen table and said, in passing, that the aggregator "publishes the rulebook." This is that rulebook, and it is time to be specific about it.

For conventional conforming lending in the United States, there are two of them.

The Selling Guide is Fannie Mae's. It states, in exhaustive detail, the requirements a loan must meet for Fannie Mae to buy it: who may be a borrower, what property types qualify, what income may be counted and how it must be documented, what credit history is acceptable, what the loan-to-value and ratio limits are, what the appraisal must show, and what the lender must certify at delivery.

The Seller/Servicer Guide is Freddie Mac's, and it does the same job. It is organized differently, it uses different vocabulary for some of the same concepts, and on a meaningful number of specific questions it reaches a different answer than Fannie Mae does.

That last sentence is worth stopping on, because loan officers routinely say "agency guidelines" as if there were one set. There are two. A file that is ineligible under one may be eligible under the other, and knowing which of your lender's products go to which agency is a genuine, billable skill. Chapter 15 covers the automated systems that apply these two rulebooks; they are not interchangeable either.

Both guides are free and public. Anyone with a browser can read either one this afternoon. Both are updated continuously — Fannie Mae publishes Selling Guide Announcements, Freddie Mac publishes Bulletins, and each one amends the guide on a stated effective date. A printed excerpt of either guide begins going stale the day it is printed, which is exactly why this book will not reproduce one.

The stack

The agency guide is not the top of the structure, and it is not the bottom.

THE STACK OF RULES ABOVE ONE FILE                    [constructed teaching diagram]

  STATUTE & REGULATION       TILA/Reg Z (including Ability-to-Repay and the
  ────────────────────────   Qualified Mortgage rule), ECOA/Reg B, RESPA/Reg X,
          │                  HMDA, FCRA, Fair Housing. Federal law. Chapters
          ↓                  24-26. Not waivable by anyone, at any price.
  REGULATOR / CONSERVATOR    FHFA. Regulates Fannie Mae and Freddie Mac and has
  ────────────────────────   been their conservator since 2008 (Chapter 2). Sets
          │                  loan limits, pricing frameworks, credit policy.
          ↓
  THE AGENCY GUIDE           Fannie Mae SELLING GUIDE / Freddie Mac
  ────────────────────────   SELLER-SERVICER GUIDE. Free. Public. Searchable.
          │                  Amended continuously. THE RULEBOOK.
          ↓
  THE INVESTOR / AGGREGATOR  If your employer sells to an aggregator rather than
  ────────────────────────   to the agency directly, that buyer's own purchase
          │                  requirements apply on top.
          ↓
  YOUR LENDER'S OVERLAYS     Stricter than the guide, by choice. Not published
  ────────────────────────   publicly. The binding constraint on a large share
          │                  of declines. Section 14.7.
          ↓
  THE MI COMPANY             Above 80% LTV, a separate underwriter with its own
  ────────────────────────   guidelines and its own right to decline. A third
          │                  "yes" that most loan officers forget exists.
          ↓
  ┌─────────────────────┐
  │   4412 LINDEN ST.   │    Every "no" you will ever receive comes from exactly
  └─────────────────────┘    one of these rows. Naming the row is step one.

Read that diagram from the bottom up and you have the shape of this chapter. A decline is not a mood. It originates in a specific row of that stack, and the row determines everything about what you can do next — whether there is an exception path, who owns it, whether another lender would reach a different answer, and whether the honest thing to tell your borrower is "give me two days" or "not this house, not this year."

How to actually read a guide

Four habits separate people who use the guide from people who quote it.

Search by topic name, not by section number. Both guides use structured identifiers — Fannie Mae's look like B3-5.3-07, and the topic behind that one at the time of writing is significant derogatory credit events and waiting periods. Those identifiers are genuinely useful for citing a rule in an email. They are also renumbered when the guide is reorganized. Search the words.

Read the whole topic, not the sentence somebody sent you. Nearly every requirement in either guide has qualifiers attached — this applies to principal residences only, this applies unless the loan is manually underwritten, this applies except when the borrower has a certain history. A screenshot of one sentence, forwarded three times inside your company, is how a firm belief about a rule that does not exist gets established.

Check the effective date. The guide states when an amendment takes effect and whether it applies to loans with application dates on or after that date, or casefiles submitted on or after it, or loans delivered on or after it. Those are three different populations.

Learn the matrices. Fannie Mae publishes an Eligibility Matrix — a compact grid of the maximum loan-to-value, combined loan-to-value, and related limits by transaction type, occupancy, property type, and underwriting method. Freddie Mac publishes comparable summary exhibits. These are the single most efficient pages in conventional lending, and they are the pages you will actually have open when a borrower is on the phone.

⚖️ Compliance Check

The Selling Guide is not a law, and complying with it is not compliance.

The agency guides are contracts. They state the terms on which a specific buyer will purchase a specific loan. Meeting them makes a loan saleable. It does not make it legal.

Underneath the guides sits federal law that applies whether or not anyone ever buys your loan: the Truth in Lending Act and Regulation Z, including the Ability-to-Repay requirement and the Qualified Mortgage standards built on it; the Equal Credit Opportunity Act and Regulation B; the Real Estate Settlement Procedures Act and Regulation X; the Home Mortgage Disclosure Act; the Fair Credit Reporting Act; the Fair Housing Act. The relationship between the QM standards and agency eligibility has been restructured more than once, and the current arrangement is Chapter 24's subject.

Two practical consequences. First, a loan can be perfectly agency-eligible and still be a compliance problem — the guide has nothing to say about whether your disclosure was delivered on time. Second, state law sits alongside all of it and varies enormously.

Verify current requirements with your compliance department and your state regulator. Nothing in this book is legal advice, and the rules discussed here change on a published schedule.

14.2 The four Cs, restated for people who actually underwrite

Every introductory course in this business teaches the four Cs: capacity, credit, capital, collateral. It is a decent mnemonic and a poor model, and you have now done enough work to be given the better version.

Here is what the four Cs actually are, stated as the four questions an underwriter is asking while reading your file:

Three things the mnemonic hides, all of which matter.

The four Cs are not equally weighted, and the weights are not fixed. On a 95% loan-to-value purchase, capital and credit carry far more weight than they do at 60%, because there is very little equity standing between the investor and a loss. On a self-employed file, capacity absorbs most of the underwriter's attention because it is the number most likely to be wrong. The Cs are not four boxes to tick; they are four dials that move each other.

The four Cs are not independent. This is the whole content of §14.6, and it is the single most important idea in this chapter. A weak C and another weak C do not make a file that is twice as risky. They make a file that is categorically different from either weakness alone.

The four Cs describe creditworthiness only. They say nothing whatsoever about whether this loan is the kind of loan the investor buys. That is a different question, it is answered first, and it is answered by a completely different part of the guide. Which brings us to the distinction that organizes this chapter.

There is also a fifth thing an underwriter checks, which no mnemonic includes and which is responsible for a startling share of conditions: does the file agree with itself? The employer on the application, the employer on the paystub, and the employer on the verification of employment should be the same employer. The address history on the 1003 should be consistent with the address history on the credit report. The deposit on the bank statement should be the amount on the earnest money check. An underwriter reads for contradictions, because a contradiction is either a mistake — in which case something else in the file may also be a mistake — or it is not a mistake, which is worse. Chapter 27 handles the second case. Your job on the first is to make it not happen: read your own file, once, all the way through, before you submit it.

14.3 Eligibility vs. creditworthiness

Two words. Learn them properly and you will spend the rest of your career arguing the right case.

Eligibility asks a categorical question: is this loan — this property, this occupancy, this purpose, this product, this borrower's legal status, this loan amount — of a kind the investor will purchase at all? It is close to binary. It is largely non-negotiable. It is answered by matrices, lists, and definitions, and it is answered before anyone looks at a paystub.

Creditworthiness asks a different question: will this borrower repay? It is evaluated holistically, on a gradient, across the four Cs. It is where judgment lives, where compensating factors live, where automated evaluation lives, and where a well-built file can win an argument that a badly-built one loses.

THE TWO GATES                                        [constructed teaching diagram]

     ┌────────────────────────────┐        ┌────────────────────────────┐
     │      GATE 1 — ELIGIBLE?    │        │  GATE 2 — CREDITWORTHY?    │
     ├────────────────────────────┤        ├────────────────────────────┤
     │ Is this the KIND of loan   │        │ Will THESE PEOPLE repay    │
     │ this investor buys?        │        │ THIS loan?                 │
     │                            │        │                            │
     │ occupancy                  │        │ credit history and score   │
     │ property type              │        │ capacity and the ratios    │
     │ project eligibility        │        │ reserves and cushion       │
     │ loan purpose               │        │ employment stability       │
     │ loan amount vs. limit      │        │ income continuance         │
     │ product and term           │        │ payment shock              │
     │ max LTV / CLTV for the     │        │ how the factors LAYER      │
     │   transaction type         │        │                            │
     │ borrower legal status      │        │                            │
     │ number of financed props   │        │                            │
     │ how title will be held     │        │                            │
     ├────────────────────────────┤        ├────────────────────────────┤
     │  BINARY. Mostly no path.   │        │  GRADIENT. Arguable.       │
     │  Compensating factors are  │        │  Compensating factors are  │
     │  IRRELEVANT here.          │        │  the entire toolkit.       │
     └────────────────────────────┘        └────────────────────────────┘
              ↓ pass                                ↓ pass
              └──────────────── both, or no loan ───┘

The reason to hold these apart is that a file can fail either gate independently, and the failures look identical from the borrower's side and require completely different responses from you.

A file can be gorgeously creditworthy and flatly ineligible. Two physicians with 800 scores, forty percent down, three years of reserves, and a twenty-two percent debt-to-income ratio, buying a unit in a condominium project that fails project eligibility — litigation, owner-occupancy, budget, insurance, whatever the specific defect is. Their creditworthiness is irrelevant. There is no amount of reserves that makes an ineligible project eligible. What might work is a different investor with different project requirements, or a different property, and both of those are conversations you should be having within an hour of learning the problem — not three weeks later, after you have written a heartfelt letter about how strong the borrowers are, which nobody needed and nobody read.

And a file can be perfectly eligible and not creditworthy. Single-family detached, primary residence, purchase, conventional thirty-year fixed, loan amount well inside the limit, standard down payment. Every eligibility box passes. The borrower has a fifty-three percent debt-to-income ratio, no reserves, and a sixty-day late from eight months ago. Nothing about the kind of loan is the problem. Here, an argument is available: pay down a card, restructure the deal (Chapter 13), document income that was missed, produce reserves that exist but were not verified.

🎓 NMLS Exam Watch

This distinction is testable and the exam likes to bury it in a stem that sounds like a math problem. A question describes a borrower with excellent credit and substantial assets purchasing a property that does not meet the program's requirements, then asks what the loan officer should do. Candidates who are pattern-matching on "strong borrower" pick the answer about compensating factors. Compensating factors have no bearing on eligibility. The correct answer involves changing something structural — the program, the property, the investor — or telling the borrower the truth.

A second, related trap: the exam distinguishes between a borrower being approved and a loan being eligible for sale. Approval is your lender's decision about a borrower. Eligibility is a purchaser's decision about a loan. The same file can pass one and fail the other, and the exam tests whether you know that these are two different sentences.

A third: watch for stems where the "problem" is something no amount of borrower quality touches — occupancy misrepresentation, a property type outside the program, a purpose the product does not allow. Those are eligibility. There is no letter of explanation for them.

Now the honest caveat, because a distinction taught without its blur is a distinction you will misapply. Some limits sit on the boundary. A maximum loan-to-value for a transaction type is an eligibility cap: above it, the loan does not exist for that investor. Below it, where you sit within the range is a creditworthiness input. A minimum representative credit score works the same way — it functions as an eligibility floor, and every point above the floor is creditworthiness. When you meet a limit like that, ask which side of it you are on. If you are outside the cap, you are having an eligibility conversation and no letter will help. If you are inside it and merely uncomfortable, you are having a creditworthiness conversation and there is work to do.

🔍 Check Your Understanding

Classify each as an eligibility problem or a creditworthiness problem:

1. The borrower's debt-to-income ratio is 49% on a conventional purchase.
2. The subject property is a condominium in a project whose homeowners association is in litigation over construction defects.
3. The borrower has two 30-day mortgage lates in the last twelve months.
4. The borrower intends to occupy the property, but their employment is 400 miles away and they are not relocating.
5. The borrower has 1.2 months of reserves after closing on a 95% loan-to-value purchase.

(2 and 4 are eligibility — the project and the occupancy classification. 1, 3, and 5 are creditworthiness, and 5 in particular is arguable. Note that 4 is also potentially a fraud question, which is Chapter 27; an occupancy statement that the facts do not support is not a paperwork issue.)

14.4 Credit event waiting periods

A waiting period is a required interval between a significant derogatory credit event and the new loan. Chapter 10 taught you to read the credit report and identify the event. This section is about what the rulebook does with it.

Start with the shape, because the shape is stable even when the numbers are not.

Every waiting period has four moving parts.

1. The event type. A Chapter 7 bankruptcy, a Chapter 13 bankruptcy, a foreclosure, a deed-in-lieu of foreclosure, a preforeclosure or short sale, and a charge-off of a mortgage account are treated as different events with different periods. They are not interchangeable, and a borrower who says "I lost a house" has told you almost nothing.
2. The measuring date. This is what loan officers get wrong. It is a specific, documented, verifiable date — not the year the trouble started and not the date the borrower moved out.
3. The ending date. Commonly the disbursement or note date of the new loan, not the application date. A file that clears the period by two weeks on the day you take the application does not clear it if the closing slips.
4. The exception path. Most of these periods are shortened where the borrower documents extenuating circumstances, and the shortened period is a different number for each event.

CREDIT-EVENT WAITING PERIODS - THE STRUCTURE, NOT THE ANSWER

[COMMONLY CITED conventional structure, stated as a framework to verify. Fannie Mae
 and Freddie Mac state these differently from each other, both have revised them more
 than once, and only the guide in force on your loan's note date governs. Look up
 every one of these before you say it out loud to a borrower.]

  EVENT                        MEASURED FROM                 COMMONLY CITED
  ─────────────────────────────────────────────────────────────────────────────
  Chapter 7 or 11 bankruptcy   the DISCHARGE date, or the    4 years
                               DISMISSAL date if the case
                               was dismissed rather than
                               discharged

  Chapter 13 bankruptcy        DISCHARGE date --- shorter    commonly 2 years
                               DISMISSAL date --- longer     commonly 4 years

  Multiple bankruptcy filings  the most recent discharge     LONGER than a
    within the past 7 years    or dismissal                  single filing;
                                                             commonly 5 years

  Foreclosure                  the COMPLETION date of the    the LONGEST of
                               foreclosure action            the set;
                                                             commonly 7 years

  Deed-in-lieu of foreclosure  the date the deed was         shorter than
                               transferred                   foreclosure;
                                                             commonly 4 years
  Preforeclosure / short sale  the sale date

  Charge-off of a mortgage     the charge-off date           commonly 4 years

  ─────────────────────────────────────────────────────────────────────────────
  RUNNING TO:  commonly the DISBURSEMENT or NOTE date of the new loan.
               NOT the application date. This distinction costs files.

  EXTENUATING CIRCUMSTANCES:  a documented exception path exists for most of
               these. It commonly shortens the period substantially -- and for
               foreclosure it commonly comes with ADDITIONAL restrictions on
               loan-to-value, occupancy, and loan purpose.

  ALSO REQUIRED: clearing the waiting period is necessary, not sufficient. The
               guides separately require RE-ESTABLISHED credit -- a demonstrated
               history since the event -- and the file still has to satisfy
               everything else in this chapter.

Three practitioner notes that the table cannot carry.

Get the document, not the story. A borrower's recollection of when their bankruptcy "finished" is wrong roughly as often as it is right, and the gap between discharge and dismissal is not a nuance — it is frequently the difference between two years and four. Ask for the discharge order. Ask for the trustee's deed, the settlement statement from the short sale, the letter from the servicer. Chapter 10 taught you that the credit report's dates on old derogatory events are frequently unreliable; this is the situation where that unreliability is most expensive.

A mortgage that went through a bankruptcy is its own puzzle. When a mortgage debt was discharged in a bankruptcy and the property was foreclosed later, there is a specific rule about which clock governs, it is not obvious, and it has changed. There is an answer in the guide. Look it up every time; do not carry the version you learned three years ago.

Extenuating circumstances is a defined term, not a sympathy standard. The guides describe it, in substance, as a nonrecurring event beyond the borrower's control that resulted in a sudden, significant, and prolonged reduction in income or a catastrophic increase in financial obligations. The important word is nonrecurring. Divorce and a failed business are genuinely devastating and are commonly not accepted, on their own, as extenuating — the reasoning being that they are not outside the borrower's control in the way an illness or a job loss from a plant closing is. Whatever the borrower's circumstance, it must be documented with third-party evidence, and the file must show the borrower has recovered.

📞 On the Phone

Borrower: "We lost our house back in 2021. Somebody told me it's seven years, so we're stuck until 2028?"

The wrong answer, version one: "Yes, seven years." You have just told a household to keep renting for two more years based on a number you did not verify and an event you have not classified.

The wrong answer, version two: "Oh, there are exceptions, we can probably work with that." You have now created hope you cannot support, on a file with no documents in it.

What actually works: "Maybe, maybe not, and the difference is a document. Was it an actual foreclosure that went all the way through, or did you sell it short, or hand the deed back? Those are three different rules with three different clocks. And I need the exact date it finished — not when you moved out, when the transfer happened. If you can find me the paperwork from the closing or the letter from the lender, I'll have you an answer tomorrow, and it may be a better answer than seven years."

Three things happened there. You did not guess. You told them the answer depends on a classification they did not know existed. And you gave them one finite task. That last part is the whole trick with a borrower carrying a credit event — they have been carrying shame about it for years, and a specific errand is enormously easier to receive than a verdict.

14.5 DTI, reserves, and the limits that move together

Chapter 4 taught you to compute the ratios. Chapter 12 taught you what reserves are and how they get verified. This section is about what the rulebook does with them, and the answer is not what a new loan officer expects.

The guideline maximum is not a constant. It is a surface.

New loan officers carry around a single number — "the max DTI is X" — and it is the wrong shape of knowledge. The maximum debt-to-income ratio available to a given file depends on the loan-to-value, the representative credit score, the occupancy, the property type, the product, the underwriting method, and in some cases the reserves. This is precisely why Fannie Mae publishes an Eligibility Matrix rather than an eligibility list. It is a grid. You read down and across.

The same is true of reserves. There is no universal conventional reserve requirement. Reserve requirements attach to specific situations — certain occupancies, certain property types, multiple financed properties, manually underwritten loans, and others — and in many transactions the file's required reserves are set by the automated evaluation rather than by a printed number (Chapter 15).

Which means reserves play two different roles, and confusing them is a classic error:

• Reserves as a requirement. A stated minimum for this transaction. Below it, the file does not qualify. This is an eligibility-shaped constraint.
• Reserves as a compensating factor. Cushion beyond the requirement, offered to offset a weakness somewhere else. This is a creditworthiness argument, and §14.9 covers how to make it.

The dollars that satisfy the requirement cannot also be the dollars you offer as a compensating factor. Only the excess counts as strength.

🧮 Run the Numbers

What one dollar of debt costs, and why the ratios and the reserves move in opposite directions. [the Linden Street file]

Qualifying income is \$10,500.00 a month. So every dollar of monthly debt is worth $1 \div 10{,}500 = 0.00952\%$ of back-end debt-to-income — call it 0.95 percentage points per \$100 of monthly payment. That single conversion is worth memorizing, because it turns every "should they finance the couch?" question into arithmetic.

Total obligations today are \$4,479.72** — PITI plus mortgage insurance of \$3,033.72, plus \$1,446.00 in other monthly debts — giving a back-end ratio of 42.66%**.

Now suppose the ceiling that applies to this file is 50%. (Verify the applicable maximum in the guide and in your lender's matrix; do not take that number from this book.)

$0.50 \times \$10{,}500.00 = \$5{,}250.00$ allowed, and $\$5{,}250.00 - \$4{,}479.72 = \$770.28$.

\$770.28 a month of headroom. Feels like a lot. Now run it against a 45% ceiling instead, which is a very ordinary overlay:

$0.45 \times \$10{,}500.00 = \$4{,}725.00$, and $\$4{,}725.00 - \$4{,}479.72 = \$245.28$.

\$245.28. One furniture payment. Same file, same borrowers, same day — and the difference between comfortable and one purchase away from a problem is a rule your borrower cannot read.

Now the trade nobody warns you about. Suppose you decide to buy down the ratio by paying off the smaller auto loan — \$429.00 a month with 19 payments remaining. The ten-month rule does not reach it (Chapter 4), so it cannot simply be excluded; it has to actually be paid.

$19 \times \$429.00 = \$8{,}151.00$ to retire the loan.

The ratio improves beautifully. Remaining debts \$1,446.00 − \$429.00 = \$1,017.00; obligations \$3,033.72 + \$1,017.00 = \$4,050.72, so $\$4{,}050.72 \div \$10{,}500.00 = 38.58\%$ — down from 42.66%.

And the reserves collapse. Post-closing reserves of \$12,623.66 become \$12,623.66 − \$8,151.00 = \$4,472.66, so $\$4{,}472.66 \div \$3{,}033.72 = 1.47$ months — down from 4.16.

You bought 4.08 points of debt-to-income for 2.69 months of cushion. On a 95% loan-to-value file with a 706 representative score, that is very probably a bad trade — the reserves are one of the few genuine strengths this file has, and §14.6 explains why removing a strength from a layered file is more dangerous than it looks.

For completeness: the larger auto, \$487.00 with 31 payments left, would cost $31 \times \$487.00 = \$15{,}097.00$ to retire — more than the borrowers have. It is not an option, and knowing that in ten seconds is why you do this arithmetic before you suggest anything.

The general principle is in that last calculation and it generalizes far beyond auto loans. Guideline limits are coupled. Improving one almost always spends something that was satisfying another. Down payment improves loan-to-value and destroys reserves. Paying off debt improves debt-to-income and destroys reserves. A larger gift improves cash to close and may bump against minimum-contribution rules. Buying discount points improves the payment and the ratio and consumes cash. There is no free move, and the loan officer's actual skill is knowing which constraint is binding.

📄 Read the File

text FIGURE 14.1 - "The whole file on one page" [the Linden Street file] THE DOCUMENT Uniform Underwriting and Transmittal Summary -- Fannie Mae Form 1008 / Freddie Mac Form 1077. One page. Prepared by the underwriter when the decision is issued, day 28. It travels with the loan when the loan is sold. Field names below follow the published form; the values are this file's. THE CONTEXT Day 28 of 51. The file has been in underwriting five days. Everything Chapters 9 through 13 produced is on this page, compressed to about sixty numbers. WHAT IT SHOWS BORROWER AND PROPERTY Property address ......... 4412 Linden Street, Ridgeview Property type ............ single-family detached, built 1994 Occupancy ................ primary residence Sales price .............. 385,000 Appraised value .... 385,000 MORTGAGE INFORMATION Loan type ................ conventional Lien position ... first Amortization ............. fixed rate Term ............ 360 mo. Loan purpose ............. purchase Original loan amount ..... 365,750 Note rate ....... 6.625% Initial P&I .............. 2,341.94 STABLE MONTHLY INCOME Base ..................... 8,120.00 Other .................... 2,380.00 (shift differential/OT, and commission -- both 24-mo. avg.) TOTAL .................... 10,500.00 PROPOSED MONTHLY PAYMENT First mortgage P&I ....... 2,341.94 Hazard insurance ......... 130.00 Taxes .................... 385.00 Mortgage insurance ....... 176.78 HOA / other .............. 0.00 TOTAL HOUSING EXPENSE .... 3,033.72 All other obligations .... 1,446.00 TOTAL ALL PAYMENTS ....... 4,479.72 QUALIFYING RATIOS LTV 95.00% CLTV 95.00% HCLTV 95.00% Housing expense ratio .......... 28.89% Debt-to-housing gap ratio ...... 13.77% Total debt ratio ............... 42.66% BORROWER FUNDS TO CLOSE Required ................. 25,376.34 Verified ................. 38,000.00 Source ................... savings, 2 accounts, plus 10,000 gift Months of reserves ....... 4.16 RISK ASSESSMENT Representative credit score .... 706 Underwriting method ............ automated Underwriter comments ........... see condition sheet, 11 items WHAT IT DOESN'T It does not show a single document. Not one paystub, bank statement, or tradeline appears -- only the conclusions drawn from them. It does not show that 2,380.00 of the 10,500.00 is variable income, because "Other" is one line. It does not show that 706 is the LOWER of two middle scores 36 points apart (Chapter 10). It does not show payment shock at all. And the one word "automated" stands in for an entire findings report, which is Chapter 15's subject. THE DECISION Read your own 1008 the moment it exists, and read it against the file you submitted. If a number on this page is not a number you recognize, something was interpreted differently than you intended, and the time to find that out is day 28 -- not day 44, when a re-underwrite has a closing date attached. THE LESSON The 1008 is the file as the investor will see it. Everything you spent fifty-one days documenting is represented here by about sixty numbers, and those numbers are what somebody warrants to be true. Most loan officers have never read one. Read every one of yours.

Constructed, using this book's frozen figures. Form 1008 is a real, published, jointly-used agency form; the field set has been revised over time — check the current version.

Notice the debt-to-housing gap ratio on that form: the difference between the total debt ratio and the housing ratio, here $42.66\% - 28.89\% = 13.77\%$, which is just the \$1,446.00 of non-housing debt expressed against income. It exists on the form because a wide gap and a narrow gap describe genuinely different households. A borrower at 42% total debt whose housing ratio is 38% has almost no consumer debt and a large mortgage payment. A borrower at 42% whose housing ratio is 24% is carrying a lot of other obligations against a modest house. Same ratio, different risk, and an underwriter reads that gap deliberately.

14.6 Layered risk

This is the most useful concept in underwriting and the one most often explained badly. Here is the bad explanation, which you will hear: "layered risk means the file has a lot of risk factors." True and useless.

Here is the real one.

Layered risk means risk factors compound rather than add. Two adverse characteristics on the same file produce more risk than the sum of what either produces alone, because the mechanisms that make each one dangerous reinforce each other.

Why that is true is not a mystery. A high loan-to-value means there is very little equity between the borrower and being underwater — so if anything goes wrong, selling the house does not solve it. A moderate credit score means the borrower has, historically, been somewhat more likely to have something go wrong. A high debt-to-income ratio means less monthly slack, so a smaller shock is enough. Thin reserves mean no buffer at all when the shock lands. Each factor makes each other factor more consequential. That is not four separate problems; it is one problem with four sources.

ADDITIVE THINKING vs. COMPOUND REALITY

[ILLUSTRATIVE ARITHMETIC ONLY. The multipliers below are invented to demonstrate
 the shape of compounding. They are NOT any agency's model, NOT a measured default
 rate, and NOT to be quoted. The lesson is the gap between the two lines.]

  Take a baseline file's chance of trouble as 1 unit.
  Add four factors, each of which "only" raises risk by 40%:

    ADDITIVE  -- how an untrained reader adds up a file:
      1.00 + 0.40 + 0.40 + 0.40 + 0.40  =  2.60 units

    COMPOUND  -- how a risk model actually accumulates them:
      1.00 x 1.40 x 1.40 x 1.40 x 1.40  =  3.84 units

    ███████████████████████████                      2.60   "it's fine, look
    ███████████████████████████████████████          3.84    at each one"

  The distance between those two bars is layered risk. Nothing in the file
  changed. Only the arithmetic of how the factors combine.

Now the point of the section: your file has layers whether or not you counted them. So count them. Out loud, before you submit, every time. It takes ninety seconds and it is the difference between being surprised by an underwriting decision and having predicted it.

The layered-risk inventory on the Linden Street file

This is a genuinely approvable file. It is not a slam dunk, and saying so honestly is the entire point of this exercise.

LAYERED RISK INVENTORY                                  [the Linden Street file]

  FLAGS -- what an underwriter will see
  ─────────────────────────────────────────────────────────────────────────────
  1  Loan-to-value 95.00%          high. 5% down, $19,250. Very little equity
                                   between the investor and a loss, and MI is
                                   required, which means a second underwriter.
  2  Representative score 706      adequate, not strong. And it is the LOWER of
                                   two middle scores (742 / 706) -- the file
                                   prices and evaluates at the weaker borrower.
  3  Back-end ratio 42.66%         moderate-to-high. $4,479.72 of $10,500.00.
  4  First-time buyers             no prior mortgage performance to observe.
                                   Nobody has ever watched these people make a
                                   housing payment of this size.
  5  Payment shock +64.0%          rent $1,850.00 -> PITI+MI $3,033.72.
                                   3,033.72 / 1,850.00 = 1.64x.
                                   An extra $1,183.72 out the door every month.
  6  Variable income $2,380.00     of $10,500.00 -- 22.67%, nearly a quarter of
                                   qualifying income, is overtime, shift
                                   differential, and commission. All of it is a
                                   24-month average of something that varies.
  ─────────────────────────────────────────────────────────────────────────────
  OFFSETS -- what is genuinely on the other side
  ─────────────────────────────────────────────────────────────────────────────
  A  Reserves 4.16 months          $12,623.66 after closing. Real cushion, and
                                   verified.
  B  No derogatory credit at all   no lates in 24 months, no public records, no
                                   collections. Nothing to explain.
  C  Stable employment             3 years and 4 years, same employers, both W-2.
  D  36-month verified rental      they have paid $1,850.00 on time for three
     history                       years and it is documented, not asserted.
  E  Properly sourced gift         $10,000, documented, not a mystery deposit.
  F  Housing ratio 28.89%          under 29%. The house itself is not the
                                   problem; the consumer debt is what pushes the
                                   back end up.
  ─────────────────────────────────────────────────────────────────────────────

Read those two lists together and notice how the offsets actually engage the flags, which is what makes this a strong file rather than merely a file with a long list.

The reserves (A) engage the payment shock (5) and the variable income (6) directly: a household absorbing an extra \$1,183.72 a month, with nearly a quarter of income that fluctuates, is exactly the household for which four months of cushion is the answer. The verified rental history (D) engages the first-time-buyer flag (4) — they have no mortgage history, but they do have three documented years of paying a housing obligation on time. The clean credit (B) engages the 706 (2): a 706 that reflects thin file depth is a different animal from a 706 that reflects recent damage, and this one has no damage anywhere in it. And the low housing ratio (F) tells the underwriter that the back-end ratio (3) is a consumer-debt story, not a too-much-house story — which matters, because consumer debt amortizes away and a mortgage payment does not. Both autos will be gone inside three years.

What remains genuinely unaddressed is the combination of 95% loan-to-value with a 706 score. There is no offset in the file for that pairing, because the only real cures are more down payment or a better score, and neither is available today. That combination is why this file will be priced the way it is priced (Chapter 29) and why the mortgage insurer's own review matters (§14.7).

The transferable move. When a file is dense with layers, do not try to fix all of them and do not try to argue all of them. Find the cheapest layer to remove. On some files it is a rapid rescore that moves a 698 to 702. On some it is one more percent down that crosses a pricing tier. On some it is documenting an income source that was left out because it seemed small. Removing one layer does not subtract one unit of risk — it divides the stack. That is the practical consequence of compounding, and it is why an apparently minor improvement sometimes changes a whole file.

14.7 Overlays: your lender's rules on top of the agency's

Here is the practical payoff of the chapter.

A guideline is the agency's requirement — published, free, readable by you and by anyone else.

A guideline overlay is your employer's additional, stricter rule, stacked on top of the agency's. It is a rule about which loans your company is willing to make, not a rule about which loans the agency is willing to buy.

Overlays are not published anywhere public. There is no website. There is no aggregator that lists them. They live in an internal product matrix, a credit-policy bulletin, an email from the underwriting manager, or — genuinely, in some shops — in an underwriter's head. And they are the binding constraint on a large share of declined files.

Why they exist

Not arbitrariness. Overlays come from real pressures, and understanding the pressure tells you whether an exception is possible.

• Repurchase exposure. §14.10. If a category of loan has historically produced repurchase demands, credit policy will tighten it beyond the guide.
• Servicing cost. Delinquent loans are expensive to service even when they never default. A lender that retains servicing has a direct financial reason to be stricter than a lender that sells it.
• The buyer above your lender. If your employer sells to an aggregator rather than to the agency directly, the aggregator's requirements come along too, and they may be tighter.
• Warehouse line covenants. The bank financing your employer's loans at closing (Chapter 31) has opinions about what may be funded on its money.
• Mortgage insurance. Above 80% loan-to-value the file needs an MI company's approval as well, and the MI company has its own guidelines and its own right to say no.
• Operational capacity. Some loans require expertise a shop does not have. A lender that does not want to staff for manufactured housing simply does not do manufactured housing.
• Appetite. Sometimes it is a business decision, and that is legitimate. A lender is allowed to decide what business it wants.

What overlays typically look like

GUIDELINE vs. OVERLAY -- THE SAME FILE, TWO RULEBOOKS   [constructed illustration]

  DIMENSION            THE AGENCY GUIDE          A TYPICAL LENDER OVERLAY
  ──────────────────────────────────────────────────────────────────────────
  minimum score        a published floor for  →  a higher floor, sometimes
                       the transaction type      keyed to LTV or product
  maximum DTI          set by the matrix and  →  a hard cap several points
                       the automated system      below it
  reserves             required in defined    →  required on every file above
                       situations                a stated LTV
  property types       a defined eligible     →  a shorter list: no manufac-
                       set                       tured, no non-warrantable
                                                 condo, no 5+ acres
  documentation        a stated minimum       →  more: full written VOE, tax
                                                 transcripts, extra months of
                                                 statements
  appraisal            waiver available in    →  full appraisal on everything,
                       defined cases             regardless
  gift funds           permitted under        →  a minimum borrower own-funds
                       stated rules              contribution
  ──────────────────────────────────────────────────────────────────────────
  Every right-hand cell is LEGAL, COMMON, and INVISIBLE to your borrower.
  None of them means the agency said no.

The question

"Is that the agency's rule, or ours?"

Ask it every time. Ask it politely, ask it of the underwriter or the underwriting manager, and ask it before you deliver bad news to anyone. It is not an aggressive question — it is a routine, professional one, and any competent underwriter will answer it in four seconds because they know the difference perfectly well.

The answer changes everything downstream:

• "It's the agency's." Then the rule is published, you can read it yourself, and every lender selling to that agency faces the same wall. Now the productive questions are whether the other agency says something different, whether a different product exists, or whether the file can be restructured (Chapter 13).
• "It's ours." Then another lender may not have it. If you are a broker (Chapter 1), you can move the file. If you are retail, you can request an exception — and you should know, before you ask, who has authority to grant one and what they need to see.

That single question is the most useful thing a loan officer can ask about a decline, and the reason "but I thought this was allowed" happens over and over in a new originator's first year is that nobody teaches them to ask it.

⚠️ Where Deals Die

The overlay you did not know existed — and the third rulebook nobody mentioned.

A composite, built from a pattern that repeats in every shop in the country.

The file is a 95% loan-to-value conventional purchase, primary residence, single-family detached. The loan officer checks eligibility, finds it clean, and issues a pre-approval. The borrower writes an offer. The offer is accepted. The file goes to underwriting on day 22 and comes back declined on day 27 — not by the lender, which approved it, but by the mortgage insurance company, which has its own guidelines, reviewed the file separately, and declined to insure it.

No mortgage insurance, no 95% conventional loan. The lender did nothing wrong. The agency did not say no. And the loan officer, who was watching two rulebooks, lost the file to a third.

The same shape occurs with pure lender overlays: a score floor keyed to loan-to-value, a hard debt-to-income cap several points under the matrix, a required minimum of the borrower's own funds when a large gift is involved, a property-type exclusion. The mechanism is identical every time — the loan officer verified the rule they knew to check, and was stopped by a rule they did not know existed.

What the disciplined loan officer does instead:

1. Get your own shop's overlay matrix and read it. It exists. Ask your manager. If your company cannot produce one, that is itself important information about where you work.
2. On any file above 80% loan-to-value, remember there are three approvals — the agency's, the lender's, and the mortgage insurer's — and ask early whether MI has been submitted.
3. Before you issue a pre-approval on anything unusual, ask the underwriter one question: "Is there anything about this structure we don't do?" Ninety seconds, before the offer, instead of five days, after it.
4. When you are told no, name the row of the stack it came from. Then act accordingly.

Two boundaries on all of this

First, this is not permission to shop a file around a legitimate risk assessment. Moving a file to a lender without a particular overlay is ordinary, honest brokerage. Moving a file because the first underwriter caught something and you would rather it not be caught again is fraud, and Chapter 27 is unambiguous about it. The difference is whether the facts travel with the file. They must.

Second, overlays are a fair-lending exposure and must be applied consistently. An overlay applied to some applicants and waived for others, on any basis other than a documented, consistently applied credit standard, is a serious problem — and a facially neutral overlay can still produce a disparate impact. Chapter 25 handles this properly. What matters here: exceptions must be requested, documented, and decided through a defined process, not granted by relationship. And you must never discourage an application because you assume an overlay will kill it. Take the application. Let the process produce the answer in writing.

14.8 Manual underwriting and when it appears

Manual underwriting is a human underwriter evaluating a file directly against the guide's manual underwriting requirements, without an automated recommendation to rely on.

That phrasing is deliberate. It is not "underwriting by hand" in the sense of the underwriter doing more work — an underwriter reads every file. It is a different standard, written in a different part of the guide, and it is generally stricter.

When it appears. Chapter 15 covers the automated systems and their outputs in full, so this is kept to the shape:

• The automated evaluation does not return an approval recommendation, and the lender may proceed manually where the guide permits it for that transaction.
• The borrower has no usable credit score, or the file is being built on nontraditional credit.
• The transaction or the borrower falls into a category the guide directs to manual review.
• The lender elects it, or the product requires it.
• Certain government programs have their own manual paths and their own benchmarks — those are Chapters 16 and 17, and their numbers are not these numbers.

What changes. In broad structure, and stated as structure because the specifics are exactly the kind of value this book will not print:

The last two rows are the practical ones for a loan officer. On a manually underwritten file, the compensating factors are not implied — they are written down, by a person, in the underwriter's comments, and they have to be things the file actually supports. Which means a loan officer who submits a manual file with the offsets already identified, documented, and stated in a submission memo has done a meaningful share of the underwriter's work and will get a faster and better answer. A loan officer who submits the same file with nothing but documents will get conditions asking for exactly that.

One honest correction to a common belief: manual underwriting is not a way around the computer. New loan officers hear "manual" and imagine a sympathetic human who will see what the machine missed. Sometimes, on the right file, that is roughly what happens. Much more often, the manual standard is tighter than the automated one, and a file that could not get an automated approval will not get a manual one either. Manual underwriting is a different door, not a lower one.

14.9 Compensating factors that actually work

A compensating factor is a documented strength in the file, offered to offset a documented weakness, in support of a creditworthiness argument.

Every word in that sentence is load-bearing.

Documented. A compensating factor that lives in your belief about the borrower is not a compensating factor. It is an opinion, and the underwriter cannot put an opinion in a file that gets sold.

Strength in the file. Not a strength in the borrower's life. "They're wonderful people and they'll be fine" is true of nearly every borrower and moves nothing.

Offset a documented weakness. A compensating factor has to be aimed. Offering "long job stability" against a property-type problem is a non sequitur. §14.3 already told you why: there are no compensating factors for eligibility. They only work on gate two.

In support of a creditworthiness argument. Which means it goes in writing, into the file, where the underwriter can cite it — not into a phone call.

The ones that carry weight

COMPENSATING FACTORS THAT UNDERWRITERS ACTUALLY USE
[the general shape of what the guides recognize and what practitioners cite. The
 specific factors an underwriter may rely on, and how, are stated in the guide and
 in your lender's policy. Verify before you build an argument on one.]

  FACTOR                              WHAT WEAKNESS IT ADDRESSES
  ──────────────────────────────────────────────────────────────────────────
  Reserves well beyond requirement    payment shock, variable income, high
                                      DTI, thin employment history
  Minimal or negative payment shock   ability to sustain the new payment
  Documented ability to save the      payment shock, specifically -- and this
    difference between rent and the   is the strongest version of the argument
    new payment, over time
  Long, verified housing history      first-time buyer, no mortgage history
    paid on time
  Low housing ratio                   a back-end ratio driven by consumer debt
                                      that will amortize away
  Long, stable employment with        income continuance, variable income
    the same employer
  Significant down payment /          essentially everything -- equity is the
    low loan-to-value                 most powerful single offset there is
  No discretionary use of credit;     credit depth, score
    clean history with real depth
  Consumer debt that retires soon     a high back-end ratio with a near-term
                                      end date
  ──────────────────────────────────────────────────────────────────────────
  AND THE ONES THAT DO NOT WORK, NO MATTER HOW SINCERELY OFFERED:
    "they're good people"            -- not documentable
    "they're getting a raise"        -- future income, not qualifying income
    "the house is worth more"        -- that is the appraisal's job, not yours
    "they've never missed a payment" -- already counted; that IS the score
    "their parents will help"        -- not a party to the loan
    assets nobody verified           -- not in the file, does not exist

The three traps

Double-counting. The dollars satisfying a required reserve cannot also be offered as excess reserves. The employment history the guide already required is not additionally a strength. Offering a requirement as a compensating factor tells the underwriter you do not understand the requirement.

Offering the weakness as the strength. It happens more than you would think: "their income is strong" on a file whose problem is that a quarter of that income is variable. Read your own file first.

Verbal factors. If it is not in the file, it does not exist. This is the same lesson as Chapter 1's second theme, and it never stops being the lesson.

Writing the exception request

Most shops have a defined process for requesting an exception to an overlay, or for asking an underwriter to rely on compensating factors on a close call. It is a memo. It is short. It goes in the file. Here is the structure that gets read:

EXCEPTION / COMPENSATING FACTOR MEMO -- THE STRUCTURE
                                                     [constructed template]

  1. THE ASK, IN ONE SENTENCE
     "Requesting an exception to the 45% DTI overlay to allow 46.8% on this
      file." Not "requesting consideration." Name the rule and the number.

  2. WHICH RULE, AND WHOSE
     Agency guideline or lender overlay? Cite it. If it is an overlay, say so;
     if it is the guide, say which guide and which topic. You are showing the
     reader you know what you are asking them to set aside.

  3. THE WEAKNESS, STATED PLAINLY AND FIRST
     Do not bury it. An underwriter who finds a weakness you did not mention
     stops trusting the rest of the memo. State it, quantify it, and do not
     minimize it.

  4. THE OFFSETS, WITH THE DOCUMENT NAMED FOR EACH
     "4.16 months of reserves post-closing (bank statements, pp. 3-8)."
     "36 months of verified rent paid on time (VOR, dated 10/14)."
     One line each. Document reference on each line. No adjectives.

  5. WHAT THE OFFSETS DO NOT COVER
     Name the residual risk honestly. This is the paragraph that makes the
     memo credible and the one everybody omits.

  6. THE ASK AGAIN, WITH THE DECISION-MAKER NAMED
     Who has authority to grant this, and by when do you need it, and what
     happens on the calendar if the answer is no.

That fifth section is the one that separates a memo that works from one that gets ignored. An underwriter or a credit-policy officer reading an exception request is looking for a reason to trust your reading of the file. Nothing establishes that faster than being the person who names the risk before they have to find it.

14.10 Reps and warrants: why the underwriter is that careful

We can now answer the question that has been sitting under this whole chapter, and under most new loan officers' relationship with underwriting.

A representation and warranty is the lender's promise, made to the investor at the time the loan is sold, that the loan meets the guidelines. Not "we looked at it carefully." Not "we believed in good faith." It conforms. The income was calculated correctly. The assets were verified. The occupancy is what the file says it is. The appraisal met requirements. The waiting period was cleared. The eligibility criteria were met.

It is a contractual promise about facts, made by a company, on a file assembled by people.

What a breach costs

If a representation turns out to be untrue, the investor can issue a repurchase demand: the lender must buy the loan back, at par — the unpaid principal balance plus accrued interest and costs. Sometimes an indemnification or a make-whole payment is negotiated instead. Chapter 23 works the mechanics, along with early payment default; what you need here is the shape of the exposure.

Three features make this severe in a way that is not obvious.

It arrives years later. The demand does not come at closing. It comes when the loan goes delinquent and somebody pulls the file, or when a routine quality-control review samples it. That can be eighteen months or five years after you closed it.

It usually arrives on a loan that has already gone bad. This is the part that matters. A performing loan rarely draws scrutiny. A repurchase is therefore, in practice, a demand to buy back a defaulted asset at the price of a good one — and the difference between those two numbers is the loss.

It attaches to the file, not the market. "Everyone was doing this" is not a defense. "It was a close call" is not a defense. The question is only whether the loan conformed.

THE PROMISE AND WHAT BREAKS IT                     [constructed teaching diagram]

  DAY 51        The lender funds and later SELLS the loan. At delivery it
                REPRESENTS AND WARRANTS that this loan meets the guide.
                     │
                     ↓
  MONTHS 1-n    The loan performs. Nobody looks at it. Selling reps begin to
                age toward relief (see below).
                     │
                     ├──────────► QUALITY CONTROL: the lender's own post-closing
                     │            QC, plus the agency's own sampling -- random
                     │            and targeted. Files get pulled and re-reviewed.
                     ↓
  SOMETHING     Delinquency, or a QC hit. The file is re-underwritten by
  HAPPENS       somebody with hindsight and no deadline.
                     │
                     ↓
  THE FINDING   "The income was overstated." "The occupancy was not primary."
                "The waiting period was not met." "The asset was not sourced."
                     │
                     ↓
  REPURCHASE    Buy it back at par. On a loan that is already in default.
  DEMAND        Or indemnify. Or make whole. Chapter 23.

The relief framework, and what never goes away

After the 2008 crisis, the volume of repurchase demands against lenders was enormous, and the resulting litigation and multi-billion-dollar settlements are a matter of public record (see case-study-01.md). One consequence was that lenders became so afraid of repurchase risk that they tightened credit far beyond what the guides required — which is a direct historical explanation for why overlays are as pervasive as they are.

In response, and under FHFA direction, Fannie Mae and Freddie Mac introduced a Representation and Warranty Framework — announced in 2012 and applying to loans acquired beginning in 2013, and revised repeatedly since. Its structure, which is what you should carry:

• Relief through payment performance. Selling representations and warranties are relieved after the loan demonstrates a defined period of consecutive on-time payments — commonly cited as thirty-six months, with a shorter period for certain refinance programs, and with an alternative path added later that tolerates a small, bounded number of early delinquencies.
• Relief through quality control. A satisfactory independent quality-control review by the agency can also produce relief, without waiting out the payment clock.
• Life-of-loan exclusions. Certain matters are never relieved, no matter how long the loan performs. The commonly cited categories: misrepresentation, misstatement, and omission; specified data inaccuracies; clear title and first-lien enforceability; compliance with the agency's charter requirements; compliance with certain laws, including high-cost and responsible-lending requirements; and unacceptable mortgage products.

Verify the current framework and its terms at the source. It has been revised more than once and the details in the list above are structure, not settled values.

Read that last bullet again, because it is the one with your name on it. Misrepresentation never ages off. A loan can perform flawlessly for a decade, and if the file contained a material misstatement, the exposure is still live. That is why an underwriter who seems relaxed about a judgment call becomes immovable about a document, and it is why "the borrower said" is never an acceptable substitute for "the record shows."

It is also the reason the credit refresh happens shortly before closing on virtually every conventional file — a practice that grew directly out of post-crisis agency quality initiatives requiring lenders to confirm that a borrower had not taken on undisclosed debt between application and closing. On the Linden Street file, that refresh is day 44, it finds something, and Chapter 19 works the resolution.

📞 On the Phone

How to argue with an underwriter, and how not to.

What does not work: "Can you just look at this again? These are really strong borrowers and they're going to lose the house." Every sentence there is about you and the borrower. None of it is about the file. The underwriter has heard it four times today and cannot act on any of it.

What works: "I want to make sure I'm reading the decline right before I call them. Is the reserve requirement here the agency's or ours? — Ours, okay. Then two things. One, I've got another two months of reserves in a retirement account we didn't document because we didn't think we needed it; if I get you the statement and the terms-of-withdrawal page, does that clear it? Two, if it doesn't, who owns the exception and what would they want to see?"

Look at what that call does. It names the row of the stack. It offers a document, not an argument. It asks a closed question that can be answered yes or no. And it identifies the decision-maker if the answer is no. The whole call is ninety seconds and it treats the underwriter as what they actually are — the person who has to make a promise about this file to somebody who will hold them to it.

Once you genuinely understand reps and warrants, underwriters stop being obstacles and start being the most useful colleagues you have. They know the guide better than you do, they know your overlays better than anyone, and almost all of them would rather help you fix a file than decline it. What they cannot do is warrant something the file does not support. Stop asking them to.

🗂️ The Loan File

Chapter 14 contribution: read the file against the conventional rulebook, and name what an underwriter will flag.

The file went to underwriting on day 23. The conditional approval came back on day 28 with eleven conditions (those are Chapter 19's). This checkpoint is what a competent loan officer should have been able to predict on day 22, before submitting.

Gate one — eligibility. Work the categorical questions first, because they are the ones with no argument attached.

Gate one passes — with one flag that most loan officers would not have written down: at 95% loan-to-value this file requires mortgage insurance, which means the MI company reviews it too. That is a third rulebook (§14.7), and it is the one that is easiest to forget exists until it declines something.

Gate two — creditworthiness. The ratios, stated once, with both the numerator and the denominator:

The honest flag-and-offset reading. This is the exercise. For each thing an underwriter will flag, name what is actually on the other side of it — and where nothing is, say so.

What this settles: the file is eligible, and it is creditworthy, and it is not a slam dunk. Six flags with genuine offsets, one pairing with none. That is a normal, approvable, competently assembled conventional file — and knowing that on day 22 means the day-28 conditional approval is confirmation rather than news.

What it does not settle: everything the layers are exposed to. Nothing in this analysis protects against a new layer being added — which is exactly what happens on day 44.

Open questions carried forward:

• Q14.1. Which of the eleven conditions are documentation and which are decision conditions? (Chapter 19)
• Q14.2. Does the mortgage insurer's review produce anything the lender's did not?
• Q14.3. If one more layer were added to this stack — one more debt, one less month of reserves — which of these six flags would tip first?

Your task. In Appendix C's workbook, build the flag-and-offset table for your own file. Two columns, no adjectives, a document named for every offset. Then write the sentence at the bottom that most loan officers will not write: which flag has no offset, and what would it take to get one? If you can answer that before you submit, you can predict the decision. If you cannot, you are going to be surprised, and surprises in this business always arrive on a Friday.

Conclusion

The rules are written down. For conventional lending they are written in the Fannie Mae Selling Guide and the Freddie Mac Seller/Servicer Guide, they are free, they are public, they are amended continuously, and they are the most under-read documents in the industry. Read them. Not all of them — nobody reads all of them — but the topics that touch your files, and read them yourself rather than accepting a forwarded sentence as the rule.

Every question those guides answer is one of two kinds. Eligibility asks whether this is the kind of loan the investor buys, and it is close to binary; compensating factors have nothing to do with it. Creditworthiness asks whether these borrowers will repay, and it is a gradient where documented strengths genuinely move documented weaknesses. Loan officers who conflate these spend their energy arguing the wrong case — writing letters about a wonderful borrower to a rule that never cared about the borrower.

Within creditworthiness, the concept that does the most work is layered risk. Risk factors compound; they do not add. That is why a file with four individually acceptable characteristics is a different proposition from a file with one, and it is why removing a single layer sometimes changes everything.

And a large share of the walls you will hit were not built by the agency at all. Overlays are your employer's own stricter rules, unpublished, entirely legal, and the binding constraint on a lot of declined files. The question — is that the agency's rule, or ours? — is worth more to your first year than any other sentence in this chapter, because the answer determines whether the file is dead or merely in the wrong building.

Underneath all of it sits the representation and warranty: the lender's promise to the investor that the loan is what the file says it is, enforceable by repurchase, years later, usually on a loan that has already defaulted, and — for misrepresentation — never expiring. The underwriter is not being difficult. They are the person who has to make that promise. Bring them documents, not arguments, and you will find they were on your side the entire time.

Next: the rules in this chapter are not usually applied by a person reading a grid. They are applied, first, by an automated system that takes the file and returns a recommendation and a list of what must be proven. Chapter 15 opens that system up — what it actually evaluates, what its output means, what it does not do, and why the sentence "the system approved it" is one of the most misunderstood in the business.

Key Terms

Selling Guide — Fannie Mae's published, free, continuously updated statement of the requirements a mortgage loan must meet for Fannie Mae to purchase it. (Ch.14)

Seller/Servicer Guide — Freddie Mac's equivalent published rulebook; organized differently from the Selling Guide and reaching different answers on a meaningful number of specific questions. (Ch.14)

Eligibility — the categorical question of whether a loan is of a kind a given investor will purchase at all: occupancy, property type, purpose, product, loan amount, and similar. Largely binary; compensating factors do not apply. (Ch.14)

Creditworthiness — the holistic question of whether a particular borrower will repay, evaluated across capacity, credit, capital, and collateral. Where compensating factors live. (Ch.14)

Guideline overlay — a lender's own requirement, stricter than the agency guideline, layered on top of it. Legal, common, and not published publicly. (Ch.14)

Waiting period (credit event) — a required interval between a significant derogatory credit event — bankruptcy, foreclosure, deed-in-lieu, short sale, mortgage charge-off — and the new loan. Measured from a specific documented date to, commonly, the note or disbursement date. (Ch.14)

Extenuating circumstances — a defined exception standard: a nonrecurring event beyond the borrower's control causing a sudden, significant, and prolonged reduction in income or a catastrophic increase in obligations. Documented with third-party evidence; commonly shortens a waiting period. (Ch.14)

Layered risk — the principle that risk factors on a file compound rather than add, so that several individually acceptable characteristics together present a materially different risk than any one of them alone. (Ch.14)

Compensating factor — a documented strength in the file, aimed at a documented weakness, offered in support of a creditworthiness argument. Never applicable to eligibility. (Ch.14)

Manual underwriting — evaluation of a file by an underwriter directly against the guide's manual requirements, without an automated recommendation; a different and generally stricter standard, in which compensating factors must be explicitly identified and documented. (Ch.14)

1008 / transmittal summary — the Uniform Underwriting and Transmittal Summary (Fannie Mae Form 1008 / Freddie Mac Form 1077): the one-page underwriting summary of a loan file — income, payment, ratios, funds, reserves, score, and the underwriter's comments — that travels with the loan when it is sold. (Ch.14)

Representation and warranty — the lender's contractual promise to the investor, made at delivery, that the loan conforms to the applicable guidelines. Breach can trigger a repurchase demand. (Ch.14)

Life-of-loan exclusion — a category of representation, notably misrepresentation and misstatement, that is never relieved by a loan's payment performance and remains enforceable for as long as the loan exists. (Ch.14)

Spaced Review

1. (Ch. 10 + Ch. 14) Your borrower pair has middle scores of 742 and 706. Which one governs, and name two separate things in this chapter that the governing score drives. Then answer the harder question: does the other borrower's 742 do anything at all for the file?

2. (Ch. 13 + Ch. 14) A file is declined for a debt-to-income ratio 1.8 points over the cap. Before you touch the borrower's debts, name two things about the loan structure you could change that would move the ratio, and state what each one costs. Then classify the decline: eligibility or creditworthiness?

3. (Ch. 10 + Ch. 14) A borrower tells you they "went through a bankruptcy about three years ago." List, in order, the four facts you need before you can say anything about whether they qualify — and say which of the four the credit report is least reliable about.

4. (Ch. 14) An underwriter declines your file for insufficient reserves. Write the two questions you ask, in the order you ask them, and say what each possible answer changes about your next move.

5. (Ch. 13 + Ch. 14) Explain to a real estate agent, in under thirty seconds and without using the word "guideline," why their buyer was approved by one lender and declined by another on the same income, same credit, and same house.