Chapter 29: How a Rate Is Made: Rate Sheets, LLPAs, Pricing Engines, and Lender Credits

"Nobody sets your rate. Six things set your rate, and one of them is what your borrower looks like on paper." — constructed; the working premise of this chapter

Overview

The book opened at 8:40 on a Wednesday with the only question anybody ever asks first — what's your rate? — and with the claim that the number on somebody's website is real for a borrower who is not the one on your phone. Your borrower has a 706 representative score, is putting five percent down, and needs a 30-day lock. The advertised rate does not exist for them.

Twenty-eight chapters later, you can now be told exactly why, in dollars, with the arithmetic shown.

That is this chapter. It is the one that closes the loop. By the end of §29.10 you will have rebuilt the Linden Street file's 6.625% with 0.500 discount point starting from a number printed at the top of a rate sheet at 8:15 in the morning, applying every adjustment this file earns, and landing on \$1,828.75 — not approximately, not close enough, but exactly, with every basis point of price accounted for and none left over.

Two things make this chapter unusual. The first is that it contains more arithmetic than any other chapter in Part VI, and the arithmetic is all the same arithmetic: addition and subtraction of numbers expressed in points of price. There is nothing hard in it. What is hard is the convention those numbers are written in, and §29.3 is devoted entirely to that convention because every loan officer in the history of the business has been confused by it exactly once.

The second is that this chapter is where the book's first theme stops being a slogan. "You don't sell rates, you solve problems" sounds like something a sales manager says on a Monday. It is actually a statement about arithmetic: a rate is an output of a pricing calculation whose inputs are your borrower's score, their loan-to-value, their occupancy, their property, their purpose, and their lock period. You cannot negotiate the output. You can sometimes change an input — and knowing which inputs are movable, and by how much, is a skill nobody who only quotes rates will ever have.

In this chapter, you will learn to:

• Trace a note rate back to the security coupon it delivers into, and name every deduction on the way
• Read a rate sheet page: product block, rate ladder, price column, effective time, footer tables
• Convert price to dollars and dollars to price using the 100.000 par convention, in both directions
• Apply an LLPA matrix and a lock period adjustment to a base price and produce a final price
• Explain where a lender credit comes from and what caps it
• Say what a pricing engine is actually doing, and what makes it return a confident wrong answer
• Explain lender margin, hedging, and pull-through — and why the lock desk behaves the way it does

Learning Paths

🎓 Exam — §29.3 and §29.7 are directly testable and the exam loves them. Know that 100.000 is par, that a price above par is a rebate and below par is a cost, and know what the LO compensation rule did to yield spread premium. §29.4's categories are testable; its values are not. 🏠 New LO — §29.2, §29.6, and §29.10. If you can do nothing else from this chapter, be able to build a quote from a base price without opening the pricing engine. It takes ninety seconds and it is the fastest way to stop being wrong on the phone. 🤝 Partner — §29.3 and §29.4. The single most useful thing you can tell an agent is why two lenders quote differently for the same borrower, and why the borrower's score and down payment change the number by more than the shopping does. 📊 Operations — §29.5, §29.8, and §29.9. Lock period adjustments, pricing engine data hygiene, and pull-through are operational problems wearing a capital-markets costume.

29.1 From MBS coupon to your rate sheet

Chapter 28 left the money one step short of your desk. It established that loans are pooled, guaranteed, and sold as pass-through securities; that those securities trade forward, before the loans in them exist, in the to-be-announced (TBA) market; and that TBA coupons come in half-point increments — 5.0, 5.5, 6.0, 6.5, 7.0 — because a market that traded every eighth of a percent would have no liquidity in anything.

Your borrower's note, however, is not written at 6.0% or 6.5%. It is written at 6.625%, and tomorrow's file will be written at something else. Somebody has to get from a security that trades in halves to a rate sheet that quotes in eighths. That somebody is your lender's secondary marketing desk, and the bridge they build has four planks.

Plank one: the note rate is not the coupon

The rate on the note is not what the investor receives. Two things are removed from the interest stream before the money reaches the security holder.

Servicing. Somebody has to collect the payment, run the escrow account, pay the taxes and the homeowners insurance, mail the statements, and handle the borrower who calls in month forty-one because they lost a job. That work is paid for out of the rate, as a strip of it — conventionally 0.250% per year on agency fixed-rate loans, which is the minimum the agencies permit. Whether your lender keeps the servicing and books it as an asset, or sells it with the loan, the strip comes out of the rate either way.

The guarantee fee. The agency that guarantees timely payment of principal and interest to the investor charges for that guarantee. Chapter 28 named it; the number itself is negotiated between each lender and each agency, is not public per-lender, and changes. Treat any specific figure you see as illustrative, including the one below.

WHAT COMES OUT OF THE RATE                        [constructed teaching example]
                                                  Verify current guarantee fees at the source.

   NOTE RATE the borrower signs .......................  6.625%
     less servicing fee (agency minimum, conventional)  -0.250%
     less guarantee fee (illustrative)                  -0.375%
   ────────────────────────────────────────────────────────────
   = PASS-THROUGH RATE to the security holder .........  6.000%
                                                         ↓
                                          delivers into the 6.0% TBA coupon

That is the first and most important thing to internalize about rate sheets: the note rate and the coupon are roughly half a point apart, and the gap is not profit. It is the cost of two services the borrower is buying whether they know it or not.

Notice also that a range of note rates delivers into the same coupon. The window is about half a point wide — 6.625% sits at the bottom of the 6.0 window — and note rates above the bottom carry excess servicing, value that has to be stripped, sold, or bought down at delivery. This is why the rate sheet's price ladder is smooth across eighths even though the underlying security only exists in halves.

🧮 Run the Numbers

Where the first month's interest actually goes.

The Linden Street borrowers pay \$2,341.94 of principal and interest in month one, of which **\$2,019.24 is interest** (Chapter 4 established this; the principal is \$322.70). Split that interest along the lines above, on the \$365,750 balance:

Piece	Annual rate	Month one
To the security holder (the 6.0 coupon)	6.000%	\$1,828.75
To the servicer	0.250%	\$76.20
To the agency, for the guarantee	0.375%	\$114.30
Total interest	6.625%	\$2,019.24

Check the top line: $\$365{,}750 \times 0.06000 \div 12 = \$1{,}828.75$. The servicing strip is $\$365{,}750 \times 0.00250 \div 12 = \$76.20$ and the guarantee fee is $\$365{,}750 \times 0.00375 \div 12 = \$114.30$. Added before rounding, the three come to \$2,019.24 — the amortization schedule's figure exactly. (Rounded individually first, they add to \$2,019.25. That penny is what happens when you split a payment three ways, and it is why you never rebuild a canonical figure out of rounded pieces.)

One warning, because it will trip you. The \$1,828.75 going to the security holder in month one is the same number as the 0.500 discount point this borrower pays at closing. That is pure coincidence: 6.000% ÷ 12 = 0.500% per month, and 0.500 point is 0.500% of the loan, so both are 0.500% of \$365,750. There is no relationship between them. Do not build one. (Guarantee fee illustrative; verify current pricing at the source.)

Planks two, three, and four: servicing value, margin, and the hedge

The desk now has a security price. It needs a loan price. Three adjustments get it there.

It adds back the value of the servicing it just stripped out, because that strip is worth something — a stream of 0.250% a year for as long as the loan lives, which the market prices as an asset (Chapter 28 covered mortgage servicing rights). It subtracts its own margin, which is what pays for underwriters, processors, closers, technology, compliance, the loan officer's commission, and profit. And it subtracts a reserve for the cost of hedging the pipeline and for the fact that some of the loans it is about to promise rates on will never close. §29.9 takes both of those apart.

THE PRICE WATERFALL — from a trading screen to your rate sheet
                                                  [constructed teaching example]
  TBA 6.0 coupon, this morning, quoted 101-16 ...........  101.500
  + value of the 0.250% servicing strip .................   +0.875
  − lender margin (cost to manufacture + profit) ........   −1.375
  − hedge cost and pull-through reserve .................   −0.250
  ──────────────────────────────────────────────────────────────────
  = BASE PRICE printed at 6.625%, 15-day lock ...........  100.750

  Note: the guarantee fee does NOT appear as a line here. It was
  already removed from the rate stream -- which is exactly why a
  6.625% note delivers into a 6.0% pool and not a 6.5% pool.

  TBA prices are quoted in 32nds. "101-16" is 101 and 16/32 = 101.500.

Two habits to build from this diagram. First, TBA prices are quoted in thirty-seconds and rate sheets are quoted in decimals, so a desk saying "the six is down eight ticks" means the 6.0 coupon fell 8/32 = 0.250 in price. Second, and more useful on the phone: when a borrower asks why your rate is different from a competitor's, three of those four lines are the same for everybody. The TBA price is a public market. The servicing strip is worth about the same to any competent buyer. The guarantee fee varies but not wildly. The line that differs is the margin, and that is a management decision, not a favor anybody can do you. §29.9 explains what moves it.

29.2 Reading a rate sheet, line by line

A rate sheet is the document your lender publishes each morning listing, for every product it offers, the price at which it will buy a loan at each available note rate — together with every adjustment that must be applied to that price before it means anything.

It is not a menu of rates. It is a table of prices, and the rates are the row labels.

Most rate sheets share the same skeleton, and a new loan officer who learns the skeleton can read any lender's sheet in about a minute:

• A header with a date and a time. Not decorative. Pricing is live, and a sheet stamped 8:15 a.m. may be dead by noon.
• Product blocks. Conventional 30-year fixed, 25, 20, 15, 10; high-balance versions of each; FHA; VA; USDA; the ARM products; and often a separate block for each investor a correspondent delivers to. Reading the wrong block is the single most common rate sheet error.
• A loan amount band. Prices frequently differ for small loans, for loans above a threshold, and for high-balance or super-conforming amounts.
• The rate ladder and price column(s). Some lenders print one base column and put lock adjustments in the footer; others print a full grid with a column per lock period. Both are the same arithmetic, one of them pre-done. Know which one you are holding.
• The adjustment tables, usually several pages of them at the back: the loan-level price adjustment matrices, lock period adjustments, product adjusters, and whatever the lender adds on its own account.
• A price cap. Almost every sheet states a maximum price after adjustments, because a loan priced far above par prepays quickly and is worth less than its price suggests.
• A disclaimer. "Prices subject to change without notice. This sheet is not a lock." Both sentences are load-bearing.

📄 Read the File

```text FIGURE 29.1 — "The top of the sheet" [constructed teaching example] THE DOCUMENT Page 3 of an 11-page daily rate sheet, conventional conforming 30-year fixed block, effective 8:15 a.m. on day 12. Constructed; modeled on the structure of a lender rate sheet. NOT current pricing. THE CONTEXT The Linden Street file. The borrowers approved the structure yesterday; the lock will be taken this morning at 6.625% with 0.500 point. This page is where that number starts. WHAT IT SHOWS

CONVENTIONAL 30-YEAR FIXED -- CONFORMING          Effective 8:15 a.m.
Loan amounts $175,001 to the conforming limit      Page 3 of 11
BASE PRICING IS QUOTED FOR A 15-DAY LOCK.
Apply ALL adjustments, pages 9-11, before quoting.
Prices subject to change without notice. This sheet is not a lock.

   RATE        BASE PRICE      cost of the next 1/8 down
   ─────────────────────────────────────────────────────
   7.000%        102.000
   6.875%        101.625              0.375
   6.750%        101.250              0.375
   6.625%        100.750              0.500
   6.500%        100.250              0.500
   6.375%         99.625              0.625
   6.250%         98.875              0.750
   6.125%         98.000              0.875

   Maximum price after all adjustments: 103.000
   Extensions: 15 days = 0.250 pt. 30 days = 0.500 pt.
   Expired locks reprice at worst case. See page 11.

WHAT IT DOESN'T It does not show a single thing about your borrower, so no number on it is a quote. It does not show the mortgage insurance premium, which is priced separately (Chapter 16). It does not show the origination charge your branch adds. It is not a lock and confers no rights. And it says nothing about whether the file is ELIGIBLE -- price and eligibility are different questions and the sheet only answers one. THE DECISION Do not read a number off this page and say it out loud. Turn to pages 9-11, apply this file's adjustments, and then quote. That takes about ninety seconds and it is the entire difference between a quote and a guess. THE LESSON Read the right-hand column of that table. Each successive eighth of a percent costs MORE than the one above it -- 0.375, then 0.500, then 0.625, then 0.750. Price is cheap near the coupon the loan delivers into and expensive away from it. This is why "just buy the rate down another quarter" gets rapidly worse, and why the borrower who wants a number far below the market is not being quoted a rate, they are being quoted a purchase. ```

Constructed teaching figure. Rate sheets are republished daily and revised intraday; every value above is illustrative and must be verified against your lender's current sheet.

Look hard at the right-hand column of that figure, because it is doing more work than it appears to. From 7.000% down to 6.750% the sheet charges 0.375 per eighth. From 6.750% to 6.500% it charges 0.500. Below that it charges 0.625, then 0.750, then 0.875. The ladder is not linear and was never going to be: you are buying an increasingly unusual coupon, and the further you get from where the security actually trades, the fewer buyers there are and the more they charge.

The practical consequence is that the cheapest eighth is always the first one, which is why the honest answer to "how much to get me to five and a half?" is usually a number that ends the conversation, and why the loan officer who has read this column knows that before the borrower asks.

One more thing about the header. That 8:15 a.m. stamp is the reason your pricing engine says a different number at 2:00 p.m. than it said at 9:00. The desk republishes when the market moves enough to matter — Chapter 30 owns what to do about that, including what a reprice is and how to protect a borrower from one. For now, note only this: a rate sheet is a photograph of a moving thing, and quoting from yesterday's sheet is the same error as quoting from last year's credit report.

29.3 Price vs. rate: the 100.000 convention

Here is the concept everyone finds confusing exactly once. Read this section slowly and it will never confuse you again.

Rate sheets quote price, not cost.

A price is what a buyer will pay for the loan, expressed as a percentage of the loan amount. A price of 101.500 means a buyer will pay 101.5% of the loan amount for it. Chapter 1 already used this number: a \$365,750 loan sold at a price of 101.500 brings

$$\$365{,}750 \times 1.01500 = \$371{,}236.25$$

which is \$5,486.25 more than the loan amount. That extra money exists because a loan paying 6.625% is worth more than face value in a market where similar loans pay less.

100.000 is par. Par is the pivot. At a price of exactly 100.000, the loan is worth exactly the loan amount, and no money changes hands between lender and borrower on account of pricing.

From that single pivot, everything follows:

• A price above 100.000 is a premium. The lender receives more than the loan amount when it sells. That surplus is a rebate, and it can be paid out as a lender credit toward the borrower's closing costs. Pricing a loan at a rate high enough to generate a rebate is premium pricing.
• A price below 100.000 is a discount. The lender will receive less than the loan amount, so the borrower makes up the difference in cash at closing. That cash is discount points.

And now the two formulas that are the whole section:

THE PRICE-TO-DOLLARS CONVERSION                          [the Linden Street file]

     points = 100.000 − price
     dollars = points × loan amount ÷ 100

  On $365,750:  1.000 point = $3,657.50
                0.500 point = $1,828.75
                0.125 point =   $457.19

  A POSITIVE points figure is money the BORROWER PAYS.
  A NEGATIVE points figure is money the LENDER PAYS (a rebate/credit).

Applied across the whole price range this file could occupy:

Read that table in both directions until it is boring. Given a price, you can state the dollars. Given the dollars, you can state the price. That is the entire skill, and it is worth about ten seconds of hesitation per phone call for the rest of your career.

The two sentences that cause all the trouble

Here is why this trips people, stated as plainly as it can be:

A high price is good for the borrower. But a low rate produces a low price.

Both sentences are true, they point in opposite directions, and holding them at the same time is what the confusion actually is. Untangle it this way: the price is the loan's value to a buyer, not its cost to the borrower. A high-rate loan is more valuable to an investor, so it prices higher, so there is surplus to hand the borrower. A low-rate loan is less valuable to an investor, so it prices lower, so the borrower has to top it up to par with cash.

Which means the direction of the whole system is simple and never changes:

                 RATE UP  →  PRICE UP  →  borrower RECEIVES money (credit)
                 RATE DOWN → PRICE DOWN →  borrower PAYS money (points)

  and the pivot in the middle, where neither happens, is the PAR RATE
  (defined in Chapter 4): the rate whose final price is exactly 100.000.

The relationship between rate and price is called the yield spread: the fact that a lender can pay more for a loan carrying a higher rate, because it will earn more over time. When the surplus from that spread was paid to a mortgage broker as compensation, it was historically called yield spread premium (YSP) — a term you will still hear from people who were in the business before 2011. What the Loan Originator Compensation rule did to that practice is Chapter 26's subject, and it is one of the most consequential rule changes in the modern industry. The arithmetic of yield spread did not go anywhere. Only who is permitted to receive it changed.

🎓 NMLS Exam Watch

This material is heavily tested and the traps are predictable.

"A loan is priced at 99.250. Is the borrower paying or receiving?" Paying. Below par. The answer is 0.750 point. On a \$200,000 loan that is \$1,500.

"What is par?" Not "the market rate" and not "the best rate." Par is the rate at which the price is exactly 100.000 — no points paid, no credit received. Candidates who memorize par as "no closing costs" get the question wrong, because there are always third-party closing costs at par; what there is not is a discount point or a lender credit.

Discount points versus origination fee. A discount point buys down the rate. An origination fee is a charge for making the loan and buys down nothing. They are both expressed as a percent of the loan amount, they can be the same size, and the exam will present them in a fee list and ask which one is prepaid finance charge (both generally are — Chapter 4) or which one lowers the rate (only the discount point). On the Linden Street file the two happen to sit right next to each other on the Closing Disclosure: origination \$3,657.50** and **points \$1,828.75.

Premium pricing and YSP. Know that a rebate exists, that it may fund a lender credit to the borrower, and that under the LO compensation rule an originator's compensation may not vary based on the terms of the loan. That last clause is the one the stem will hinge on.

29.4 Loan-level price adjustments

Now we can name the thing that makes an advertised rate untrue for a specific person.

A loan-level price adjustment (LLPA) is a price adjustment applied for a specific risk characteristic of a specific loan. It is charged by the agency to the lender at delivery, and the lender passes it through to the borrower as price. It is published in advance, in matrices, and it is not negotiable at the loan officer's level — or at the branch manager's, or at the lender's.

The critical structural fact, and the reason this chapter follows §29.3 rather than preceding it: LLPAs are charged in price, not in rate. The matrix does not say "add 0.375% to the rate." It says "reduce the price by 1.125." What that becomes for the borrower — cash at closing, or a higher rate, or some of each — depends on where on the rate ladder they choose to sit. That choice is Chapter 13's subject. The adjustment itself is fixed the moment the file's characteristics are known.

What gets adjusted

The categories are stable even though the values are not:

Two of those deserve a sentence each because they surprise people. Cash-out refinances are expensive — they carry their own matrix and it is materially worse than the purchase grid at every score and LTV, which is why a borrower who wants \$40,000 out at today's rate is often better served by a second lien (Chapter 35). And subordinate financing is priced on CLTV, not LTV, which is exactly why the Harlow Street file's county down-payment-assistance second matters to pricing and not just to structure.

📄 Read the File

```text FIGURE 29.2 — "The cell your borrower lands in" [constructed teaching grid -- modeled on the structure of published LLPA matrices; the grids are revised periodically and these values are illustrative]

THE DOCUMENT Page 9 of the same rate sheet: the credit score / LTV price adjustment matrix for a purchase, 1-unit, primary residence, 30-year fixed. Constructed. NOT current agency pricing. THE CONTEXT Day 12, the Linden Street file, minutes before the lock. The representative score has been 706 since the credit pull on day 1 (Chapter 10) and the LTV has been 95.00% since the structure was chosen (Chapter 13). Both inputs were known well before this moment. WHAT IT SHOWS Adjustments in POINTS OF PRICE. All figures subtract from price.

Rep.       <=60.00  60.01-  70.01-  75.01-  80.01-  85.01-  90.01-   >95.00
score               70.00   75.00   80.00   85.00   90.00   95.00
─────────────────────────────────────────────────────────────────────────────
>=780        0.000   0.000   0.000   0.000   0.125   0.125   0.125    0.125
760-779      0.000   0.000   0.125   0.125   0.250   0.250   0.250    0.125
740-759      0.000   0.125   0.125   0.250   0.375   0.500   0.500    0.250
720-739      0.000   0.125   0.250   0.500   0.625   0.750   0.750    0.500
700-719      0.000   0.250   0.500   0.750   0.875   1.000  [1.125]   0.875
680-699      0.000   0.375   0.750   1.000   1.250   1.375   1.500    1.125
660-679      0.125   0.625   1.000   1.375   1.625   1.750   1.875    1.375
640-659      0.375   1.000   1.500   1.875   2.250   2.375   2.500    1.875
<640         0.500   1.250   1.875   2.375   2.875   3.000   3.125    2.375

[1.125] = 706 representative score at 95.00% LTV. This file.

WHAT IT DOESN'T It does not show occupancy, property type, purpose, subordinate financing, or product adjusters -- those are separate tables that STACK on top of this one. It does not show the waivers and caps that apply to certain affordable products and to certain first-time buyers at or below area median income (see case study 1). It is not current, and the published matrices were substantially restructured in 2023. THE DECISION Price this file at 1.125 points of adjustment from the score/LTV grid, and say so out loud to the borrowers before quoting, in dollars: "your score and your down payment are costing about $4,100 of price on this loan, and here is what we can and cannot do about it." THE LESSON Walk one row and one column from this file's cell and watch what happens. Same 95% LTV, a 720 score instead of 706: 0.750 instead of 1.125 -- a 0.375 improvement worth $1,371.56. Same 706 score, 90% LTV instead of 95%: 1.000 instead of 1.125, worth $457.19. The score is worth three times the down payment here, and fourteen points of score is the difference. THAT is why Chapter 10 spends a whole section on rapid rescoring, and why you pull credit before you quote. ```

Constructed teaching grid. The published matrices are revised periodically — they were substantially restructured in 2023 — and must be verified at the source before any live quote.

Reading a matrix without hurting yourself

Four disciplines, all learned the hard way by somebody:

One: the bands are cliffs, not slopes. A 719 score and a 720 score are in different rows. One point of FICO can be worth 0.375 of price — on this loan, \$1,371.56 — and the borrower has no idea. This is the entire economic argument for the credit conversation in Chapter 10, and it is why "they're around 715" is not a usable input.

Two: LTV bands are cliffs too, and LTV is computed on the lesser of price or appraised value. Which means the Cypress Court file's short appraisal was not only a cash problem. When value falls and the loan amount does not, the LTV rises, and a file can fall off a pricing cliff on the same afternoon it falls off a down payment cliff.

Three: the highest-LTV column is often cheaper than the one below it. Look at the grid: at 706, the 90.01–95.00 column is 1.125 and the >95.00 column is 0.875. That looks like a typo and it is not. Loans above 95% LTV carry deeper mortgage insurance coverage, and the pricing recognizes it. If you ever see a structure where adding down payment makes the price worse, check whether you have just crossed that boundary before you tell the borrower they found a loophole.

Four: adjustments stack, and they stack in price. A cash-out refinance on a condominium at 80% LTV with a 690 score is not one adjustment. It is four, added together, and the total can exceed anything the borrower imagined a "rate" could contain.

⚖️ Compliance Check

Pricing is a fair lending surface, and it is one of the most heavily examined ones.

The LLPA matrices themselves are risk-based and applied mechanically — that is a feature, because a published grid applied to a file's characteristics is the opposite of discretionary. The exposure is in what happens around the grid: pricing exceptions, discretionary concessions, and negotiated fee waivers. If a lender permits exceptions and grants them unevenly, the pattern can produce disparate outcomes even when no individual decision was made with any improper intent. Equal Credit Opportunity Act and Regulation B liability does not require intent, and Home Mortgage Disclosure Act data — which includes rate spread — is one of the places examiners look first.

What this means at your desk, concretely:

• Apply the sheet. Every file, every time, the same way.
• If your lender allows pricing exceptions, follow the written policy, document the business reason, and never let the reason be about the borrower rather than about the transaction.
• Never discuss pricing in terms of what you think a borrower will accept or notice. Beyond the fair lending exposure, that framing is the beginning of every discipline case in this business.
• Quote from the current sheet with the file's real characteristics — because a quote you cannot reproduce is a quote you cannot defend.

Requirements change and state law varies. Verify current rules with your compliance department and your regulator.

29.5 Lock period adjustments

Everything so far has priced the loan. The last adjustment prices time.

A lock period adjustment is a price adjustment for how long the lender must hold the rate before the loan funds. It exists for two reasons, and both of them are real money.

The first is carry. The lender is hedging a rate it has promised on a loan that does not exist yet, and the instrument it hedges with settles on a specific date. A longer lock means a later settlement month, and forward settlement months trade at a different price than nearby ones. Longer means worse, almost always.

The second is fallout. The longer a lock is outstanding, the more chance the loan never closes — the borrower's offer falls through, the appraisal comes in short, the borrower shops away, the underwriter declines. The lender has hedged a loan that will not arrive. §29.9 quantifies what that costs.

LOCK PERIOD ADJUSTMENT       [constructed teaching grid -- verify at the source]
Base pricing on this sheet is quoted for a 15-day lock.

    LOCK PERIOD        PRICE ADJUSTMENT       on $365,750
    ──────────────────────────────────────────────────────
    15 days                 0.000  (base)          $0.00
    30 days                -0.125                $457.19
    45 days                -0.250                $914.38
    60 days                -0.375              $1,371.56
    75 days                -0.500              $1,828.75
    90 days                -0.625              $2,285.94

  EXTENSIONS (page 11 of the sheet)
    15-day extension        0.250 pt              $914.38
    30-day extension        0.500 pt            $1,828.75
    Expired lock: reprices at WORST CASE -- the worse of the
    original price or current market. See Chapter 30.

Some lenders publish this as a footer table, as above. Others fold it into the rate sheet as a column per lock period, so that the 30-day column is simply the 15-day column minus 0.125 all the way down. The arithmetic is identical. Find out which convention your sheet uses before you quote, because reading a 60-day column and quoting it as a 30-day price is a quarter point of error you will discover at the closing table.

On the Linden Street file the lock taken on day 12 was a 30-day lock, and it therefore cost 0.125 of price — \$457.19. Was a 15-day lock available? Technically. The contract closing date was day 45 and the lock was taken on day 12, so a 15-day lock would have expired on day 27, eighteen days before the file could possibly have closed. The 30-day lock was not a choice between saving \$457.19 and not saving it — a 15-day lock was never a real option.

But it was not the only option, and the table above says so. From day 12, a 30-day lock reaches day 42 — three days short of the day-45 contract date, before anything went wrong. The 45-day lock on that same sheet reaches day 57, covers the closing with room, and costs 0.250 of price, \$914.38**. The lock expired on day 42 and bought a 15-day extension at 0.250 point — **\$914.38.

The same number. The longer lock and the extension cost exactly the same amount; the difference is that one was a decision made calmly on day 12 with the calendar in front of you, and the other was bought under pressure on day 42, absorbed by the lender as a tolerance cure, on a file that had already missed its closing date. Chapter 30 examines that decision. §29.5's part is narrower and harder to argue with: the price of the lock that would have worked was printed on the sheet the whole time.

That is worth sitting with, because it is the chapter's quiet lesson about time. A lock period adjustment is the price of a promise about the calendar, and the calendar on this file was wrong from the day the offer was written — Chapter 6 corrected the "45-day contract" to the 41 days that actually remained. Chapter 30 owns the lock decision itself: when to lock, whether to float, extensions, float-downs, and what to do when the desk reprices. What §29.5 owns is narrow and mechanical: the lock period is an input to price, exactly like the credit score, and it belongs in every quote.

29.6 Building a quote from base price to final rate

You now have every piece. Assembling them is a five-line procedure, and once you have done it a dozen times you will do it faster in your head than the pricing engine loads.

THE FIVE LINES

  1.  Find the right product block and loan amount band on today's sheet.
  2.  Read the BASE PRICE at the rate you are considering.
  3.  Subtract every LLPA the file earns (score/LTV, occupancy, property,
      purpose, product, subordinate financing, escrow).
  4.  Subtract the lock period adjustment.
  5.  FINAL PRICE. Then: points = 100.000 − final price, and
      dollars = points × loan amount ÷ 100.

Done on the Linden Street file at 6.625%:

BUILDING THE QUOTE — 6.625%, 30-year fixed, $365,750         [the Linden Street file]
Sheet effective 8:15 a.m., day 12. All adjustment values constructed.

  BASE PRICE, 6.625%, conventional 30-yr fixed, 15-day lock ....  100.750
  ───────────────────────────────────────────────────────────────────────
  ADJUSTMENTS -- every one subtracts from price
    LLPA  credit score / LTV: 706 rep score at 95.00% LTV .....   -1.125
    LLPA  occupancy: primary residence ........................    0.000
    LLPA  property type: 1-unit single-family detached ........    0.000
    LLPA  purpose: purchase ...................................    0.000
    LLPA  product/term: 30-year fixed .........................    0.000
    LLPA  subordinate financing: none (CLTV = LTV = 95.00%) ...    0.000
    ADJ   escrow account: established, not waived .............    0.000
    ADJ   lock period: 30 days (base is quoted at 15) .........   -0.125
  ───────────────────────────────────────────────────────────────────────
  TOTAL ADJUSTMENTS ............................................   -1.250
  ───────────────────────────────────────────────────────────────────────
  FINAL PRICE ..................................................   99.500

  points  = 100.000 − 99.500 = 0.500
  dollars = 0.500 × $365,750 ÷ 100 = $1,828.75

  QUOTE: 6.625%, 0.500 discount point, $1,828.75, 30-day lock.

Notice something about the total. The adjustments come to exactly 1.250 points — 125 basis points of price — of which 112.5 are the score/LTV cell and 12.5 are the lock. Nothing else on this file moved the price at all. It is a primary residence, a detached single-family home, a purchase, a 30-year fixed, with no second lien and a normal escrow account. Every one of those inputs is a zero. This is a boring file, in the specific technical sense that pricing engines mean by boring, and it still prices 1.250 points below the sheet.

In dollars, the adjustments are worth $1.250\% \times \$365{,}750 = \$4{,}571.88$ — of which \$4,114.69 is the score/LTV cell and \$457.19 is the lock.

And here is the part that answers Chapter 1. The base price at 6.625% was 100.750. For a file with none of this file's adjustments, 6.625% would have come with a \$2,743.12 lender credit. For this file it costs \$1,828.75. Same sheet, same morning, same rate, opposite direction. The advertised number was real. It was priced for a different borrower.

Because the total adjustment is a single number that applies to every row of the sheet, you can now regenerate the entire rate/point grid this file was offered — the one Chapter 13 chose from — in one pass:

THE WHOLE GRID FROM ONE SUBTRACTION                          [the Linden Street file]
Every row: FINAL PRICE = BASE PRICE − 1.250

  RATE     BASE PRICE   ADJ      FINAL PRICE   POINTS   $ TO BORROWER    P&I
  ─────────────────────────────────────────────────────────────────────────────
  7.000%     102.000   -1.250      100.750     -0.750     ($2,743.12)  2,433.34
  6.875%     101.625   -1.250      100.375     -0.375     ($1,371.56)  2,402.72
  6.750%     101.250   -1.250      100.000      0.000     par  $0.00   2,372.25
  6.625%     100.750   -1.250       99.500     +0.500      $1,828.75   2,341.94  <-
  6.500%     100.250   -1.250       99.000     +1.000      $3,657.50   2,311.79
  6.375%      99.625   -1.250       98.375     +1.625      $5,943.44   2,281.80
  ─────────────────────────────────────────────────────────────────────────────
  This file's PAR RATE is 6.750%: the rate whose final price is exactly 100.000.

That is the whole thing. One base column, one adjustment total, and the entire menu the borrower was shown falls out of it. If you can reproduce that table from a rate sheet and a file, you can price a loan.

🧮 Run the Numbers

What each eighth actually buys, and what it costs.

The price ladder gets steeper as you buy down. The payment ladder does not — each eighth of rate is worth almost exactly the same monthly dollars. Put them side by side on this loan:

Move	Price cost	Dollars	P&I saved / month	Break-even
6.750% → 6.625%	0.500	\$1,828.75 | \$30.31	60.3 months
6.625% → 6.500%	0.500	\$1,828.75 | \$30.15	60.7 months
6.500% → 6.375%	0.625	\$2,285.94 | \$29.99	76.2 months

Check the first row: \$2,372.25 − \$2,341.94 = \$30.31 saved per month, and \$1,828.75 ÷ \$30.31 = 60.3 months. Check the third: \$2,311.79 − \$2,281.80 = \$29.99, and 0.625% of \$365,750 = \$2,285.94, so \$2,285.94 ÷ \$29.99 = 76.2 months.

Now run it the other way — buying up the rate and collecting the rebate:

Move	Price received	Dollars	P&I added / month	Payback
6.750% → 6.875%	0.375	\$1,371.56 | \$30.47	45.0 months
6.875% → 7.000%	0.375	\$1,371.56 | \$30.62	44.8 months

Read the two tables together and the asymmetry jumps out. Buying the rate down takes about sixty months to earn back and gets worse the further you go. Buying it up returns cash that the higher payment takes about forty-five months to consume. The break-even calculation is Chapter 4's; the decision is Chapter 13's, which chose 6.625% and a 60.3-month break-even for reasons that had to do with this household and not with this table. What §29.6 adds is where the numbers came from: not a negotiation, not a promotion, but a base column and a subtraction.

29.7 Premium pricing and lender credits

Premium pricing is choosing a note rate whose final price is above 100.000, so that the lender receives a premium when it sells the loan and can hand part or all of it to the borrower as a lender credit toward closing costs. Chapter 4 defined the lender credit as a line item. This section says where it comes from, which is a different and more useful thing to know.

It comes from the yield spread. Nothing else. A lender offering a "\$3,000 credit" is not being generous, running a promotion, or eating a cost. It is selling a loan at a higher rate for more money and passing some of the surplus back. The borrower is buying that credit with thirty years of higher payments, and the honest version of the conversation says so.

Run the trade on the Linden Street file. Chapter 13 selected 6.625% with a half point. Suppose the borrowers had gone the other direction to 7.000%:

The cash swing is \$1,828.75 + \$2,743.12 = \$4,571.87**, and the payment rises \$91.40 a month. Break-even (Chapter 4's method): \$4,571.87 ÷ \$91.40 = 50.0 months**. If these borrowers were certain they would sell or refinance inside four years, taking the credit would have been the cheaper path. They were not, and Chapter 13 explains what they decided and why.

(A note on the half cent, since this book promises its numbers resolve: 1.250 points on \$365,750 is exactly \$4,571.875. The 0.750 credit is \$2,743.125, which the sheet rounds to \$2,743.12, so the two published dollar figures add to \$4,571.87 while the points arithmetic gives \$4,571.88. Rate sheets round. The Closing Disclosure carries the rounded dollars, and those are the numbers the borrower lives with.)

Trace the credit all the way through and you can see how much of this file it touches. The lender charges of \$3,657.50 origination and \$1,828.75 in points would become \$3,657.50 and nothing, so total closing costs of \$9,720.25 fall to \$7,891.50; add the \$4,406.09 of prepaids and the total is \$12,297.59; subtract the \$2,743.12 credit and it is \$9,554.47. Cash to close falls from \$25,376.34 to \$20,804.47 — the same \$4,571.87 — and reserves rise from \$12,623.66 to \$17,195.53.

And then the constraint nobody expects: at 7.000% the PITI is \$3,125.12 rather than \$3,033.72, so the back-end ratio moves from 42.66% to 43.53%. Buying a credit costs debt-to-income capacity. On a thinner file it can price you out of the approval you were trying to fund.

📞 On the Phone

Borrower: "The other place said they'd do it with no closing costs. Can you do that?"

The wrong answer: "No, somebody always pays closing costs." True, condescending, and it ends the conversation with you on the losing side of it.

The other wrong answer: "Sure, I can do no-cost too." You have just agreed to something you have not priced, at a rate you have not named.

What actually works: "Yes — that's called premium pricing and I can build it for you. Here's the trade so you can see it. At the rate we talked about you'd bring about \$1,800 to closing for the point. If we move the rate up, the lender pays me a rebate instead, and I apply it to your costs. At 7.000% on your loan that rebate is about \$2,743, and your payment goes up \$91.40 a month. So the question isn't whether there are closing costs — there always are. The question is whether you'd rather pay them now or pay \$91.40 a month for thirty years. If you're likely to move or refinance in the next four years, the credit wins. Past that, it doesn't. How long do you think you'll be in this house?"

The last question is the whole call. You have converted "who has the better deal" into "what is true about your life," which is a question you can actually answer together — and which the website cannot ask.

Three things that cap a credit

A lender credit cannot exceed the borrower's actual costs. Credits offset closing costs; they are not cash back on a purchase. If premium pricing generates more rebate than there are costs to absorb, the excess cannot simply be handed over. It has to go somewhere legitimate — typically by reducing the rate to a lower premium — and the loan officer who does not check this in advance discovers it at the closing table. Chapter 22 covers how credits are disclosed and Chapter 24 covers the rules.

The sheet caps the price. Figure 29.1 states a maximum price after adjustments of 103.000. High- coupon loans prepay quickly — a borrower paying well above market refinances at the first opportunity — so a security backed by them is worth less than its coupon suggests, and no one will pay unlimited premium. In practice the rate ladder simply runs out of useful rungs.

Compensation cannot vary with the terms. The rebate is the lender's, not the loan officer's. An originator's compensation may not be based on the terms of a transaction, which is precisely why the old yield spread premium arrangement no longer exists in the form it took before 2011. Chapter 26 owns this rule in full, including what it means for a broker's lender-paid versus borrower-paid compensation election. Know for now that you cannot make more money by putting your borrower in a higher rate, and that the rule exists because the industry demonstrated what happens when you can.

29.8 The pricing engine and what it is really doing

Every lender of any size runs a product and pricing engine (PPE) — usually just called the pricing engine. Loan officers treat it as an oracle. It is not an oracle. It is a lookup table with a rules layer, and understanding that is what lets you catch it when it is wrong.

Here is what it does, in order, every time you hit "search":

1. Takes the loan characteristics you entered. Amount, value, LTV, CLTV, representative score, occupancy, property type, units, purpose, term, product, lock period, escrow election, subordinate financing, and a dozen more.
2. Determines eligibility. For every product from every investor the lender delivers to, it asks whether this file is even permitted. Ineligible is an answer, and often the most useful one.
3. Pulls the base price from the current sheet for each eligible product at each rate.
4. Applies every adjustment — the LLPA matrices, the lender's own adjusters, the lock period adjustment — and totals them.
5. Applies the branch or company margin, which may differ from the base sheet's if your shop prices at the branch level.
6. Sorts and returns a grid of rate/price combinations, usually best execution first.

PRICING ENGINE -- SEARCH RESULTS                     [constructed teaching example]
Run 9:42 a.m. day 12   Loan L-2214   Sheet effective 8:15 a.m.

  INPUTS AS ENTERED
    Loan amount 365,750     Value 385,000    LTV 95.000   CLTV 95.000
    Rep score 706           Occupancy PRIMARY      Property SFD 1-UNIT
    Purpose PURCHASE        Term 360               Product CONV FIXED
    Lock 30 DAYS            Escrows YES            Sub financing NONE
    First-time buyer YES    AUS APPROVE/ELIGIBLE

  ADJUSTMENT DETAIL (applies to every row below)
    LLPA  score 700-719 / LTV 90.01-95.00 ...............  -1.125
    LLPA  occupancy PRIMARY .............................   0.000
    LLPA  property SFD 1-UNIT ...........................   0.000
    LLPA  purpose PURCHASE ..............................   0.000
    LLPA  subordinate financing NONE ....................   0.000
    LLPA  term 360 CONV FIXED ...........................   0.000
    ADJ   escrow account ESTABLISHED ....................   0.000
    ADJ   lock period 30 DAYS (base = 15) ...............  -0.125
                                             TOTAL         -1.250

  RESULTS -- CONV 30-YEAR FIXED
  ──────────────────────────────────────────────────────────────────
  RATE      BASE      ADJ      PRICE    POINTS    $ BORROWER     P&I
  7.000%   102.000  -1.250   100.750   -0.750   (2,743.12)  2,433.34
  6.875%   101.625  -1.250   100.375   -0.375   (1,371.56)  2,402.72
  6.750%   101.250  -1.250   100.000    0.000        0.00   2,372.25
  6.625%   100.750  -1.250    99.500    0.500    1,828.75   2,341.94
  6.500%   100.250  -1.250    99.000    1.000    3,657.50   2,311.79
  6.375%    99.625  -1.250    98.375    1.625    5,943.44   2,281.80
  ──────────────────────────────────────────────────────────────────
  ELIGIBILITY: ELIGIBLE. MI required; coverage 30%; BPMI monthly.
  MI premium quoted separately -- see Chapter 16.
  THIS SCREEN IS NOT A LOCK. Prices expire when the desk reprices.

Compare that screen with the hand build in §29.6. They are the same arithmetic. The engine is faster than you and it is not smarter than you, and the reason to be able to do it by hand is not nostalgia. It is that a loan officer who can reproduce the engine's answer can also tell when the engine has been fed something false — and the engine will never tell you.

⚠️ Where Deals Die

The pricing engine will price, quickly and confidently, a loan you have described incorrectly. It has no way to know. Here are the five inputs that go wrong, in rough order of how often:

The score. You priced at what the borrower told you. "I think we're around 740." The pull comes back and the representative score is the lower of the two middle scores (Chapter 10), which on the Linden Street file is 706 and not the 742 the stronger borrower would have quoted you. On this grid that is the difference between 0.500 and 1.125 — 0.625 of price, \$2,285.94. Pull credit before you quote. There is no version of this that ends well otherwise.

The LTV. You priced at the LTV implied by the contract price, and the appraisal came in low. LTV is computed on the lesser of price or value, so the Cypress Court file's \$35,000 short appraisal moved LTV and moved the pricing cell (Chapter 18).

The occupancy. A second home or investment property priced as a primary is not a pricing error, it is potentially occupancy fraud (Chapter 27), and it is caught at delivery.

The property type. Condominium versus planned unit development versus detached is a distinction the borrower cannot reliably make and the appraisal will settle. Ask what the deed says, not what the neighborhood looks like.

The lock period. You priced a 30-day lock on a file that will need 45. The 0.125 of difference is small; the extension you will buy at 0.250 is not.

There is a sixth that is not an input error at all: quoting from a stale run. The desk repriced at 11:20 and your quote is from 9:42. Chapter 30 covers reprices; the discipline here is simply to re-run before you commit to anything in writing.

Two more things a good loan officer knows about their engine.

"Best execution" is not always best. The engine sorts by price. The cheapest investor may have a guideline overlay that will decline this file (Chapter 14), a turn time that will not make the closing date, or a condition set your processor will spend two days on. Price is one input to a decision that includes the calendar.

The engine prices; it does not underwrite. An "eligible" flag reflects product parameters, not an automated underwriting decision and certainly not a human one. Chapter 15 owns the findings. Files that price beautifully and do not close are a genuine category, and pricing screens are where they begin.

29.9 Lender margin, hedging, and pull-through

Everything so far has been the borrower's side of the sheet. This section is the lender's, and it is the part almost no loan officer learns — which is a shame, because it explains most of the behavior that new originators find arbitrary.

Margin

Lender margin is the spread the lender builds into the base price to cover the cost of manufacturing a loan and to make a profit. In the §29.1 waterfall it was the line marked −1.375.

That number is a management decision, and it moves for reasons that have nothing to do with your file. When the pipeline is full and operations cannot absorb more volume, the desk widens the margin — raising prices, slowing the flow, and protecting turn times. When volume is thin, it narrows. Two lenders reading the same TBA screen on the same morning will print different rate sheets almost entirely because of this line, which is the real answer to "why is their rate better than yours today?" It is usually not a better cost of funds. It is a different appetite for volume this week.

What does the lender actually earn on the Linden Street loan? Two pieces:

The 0.500 discount point is not on that list, and this is the part that surprises people. The borrower's \$1,828.75 does not go into anybody's pocket as profit. It exists to raise a final price of 99.500 up to par. It is the borrower buying the loan back to 100.000, and the lender is exactly as whole at 6.625%-with-a-half-point as it would have been at 6.750%-at-par. That is what it means to say the rate ladder is priced.

Out of that \$8,686.56 comes the loan officer's commission (Chapter 26), the processor, the underwriter, the closer, the technology stack, the compliance function, corporate overhead, and the hedge result. A retail lender's fully loaded cost to manufacture a loan runs to several thousand dollars per file and swings sharply with volume; the Mortgage Bankers Association publishes current figures and they are worth looking up, because a loan officer who knows their employer's cost per loan understands their own compensation conversation considerably better.

Hedging

Here is the lender's actual problem. On day 12 it promised the Linden Street borrowers 6.625% for thirty days. It does not have that loan. It will not have that loan for weeks. It has sold something it does not own.

If rates rise before the loan funds, the loan the lender is obligated to make becomes worth less than 100.750 — but the lender is stuck at the promised price. That is a straightforward loss, repeated across every locked file in the pipeline.

So the lender hedges: it takes an offsetting position, most commonly by selling TBA securities forward (Chapter 28 described the instrument), so that if rates rise and its pipeline loses value, the short position gains roughly the same amount. It is not speculation. It is the opposite of speculation — the desk is trying to be indifferent to the direction of rates, so that the company earns its margin on manufacturing loans rather than on guessing.

Pull-through, and why it makes hedging hard

Now the complication that makes the whole thing genuinely difficult.

Pull-through is the share of locked loans that actually fund. It is never 100%. Borrowers walk, contracts fall apart, appraisals come in short, underwriters decline, and — critically — borrowers renegotiate or leave when rates improve.

If the desk hedges 100% of its locked pipeline and only 80% funds, it has hedged loans that never existed and must unwind the excess at whatever the market has done in the meantime. So the desk hedges its expected fundings: pipeline times estimated pull-through, estimated by lock age, product, purpose, rate movement since lock, and often by branch and by loan officer.

And here is the cruel part: pull-through moves against the hedge in both directions.

WHY FALLOUT IS A ONE-WAY STREET

  RATES FALL                            RATES RISE
  ──────────────────────────            ──────────────────────────
  Pipeline gains value          │       Pipeline loses value
  Hedge (short) loses           │       Hedge (short) gains
  Borrowers renegotiate or      │       Borrowers cling to their
    walk to a lower quote       │         locks; nobody leaves
  PULL-THROUGH FALLS            │       PULL-THROUGH RISES
  → the desk is OVER-hedged     │       → the desk is UNDER-hedged
  ──────────────────────────────┴──────────────────────────────────
  Whichever way the market moves, fallout moves the wrong way.

🧮 Run the Numbers

What being wrong about pull-through costs. (Illustrative; constructed teaching example.)

A mid-size lender locks \$100,000,000 of new loans in a week. Historical pull-through for that mix is 85%, so the desk sells \$85,000,000 of TBAs forward.

Over the next two weeks rates fall. MBS prices rise 0.500. Borrowers who locked at the old rate start calling, and pull-through comes in at 70% — only \$70,000,000 funds.

The desk is short \$15,000,000 of securities it does not need, and it must buy them back after the price has risen half a point:

$$\$15{,}000{,}000 \times 0.00500 = \$75{,}000$$

of loss on the over-hedge alone, before any renegotiation the lender grants to keep the borrowers it still has.

Now run it the other way. Rates rise, prices fall 0.500, and pull-through comes in at 95% — \$95,000,000 funds against an \$85,000,000 hedge. The \$10,000,000 that was never hedged is worth half a point less:

$$\$10{,}000{,}000 \times 0.00500 = \$50{,}000$$

The lesson is not the dollar amounts, which are constructed. It is the shape. The desk loses money when it is wrong in either direction, and its estimate of pull-through is the thing it is most likely to be wrong about. That is why the hedge reserve is a line in the price waterfall, and why it is priced into every loan the lender makes.

What this means at your desk

Almost everything a lock desk does that seems unhelpful is downstream of this section:

• Extension fees exist because fallout is expensive. The 0.250 point — \$914.38 — that carried the Linden Street lock fifteen days past day 42 came out of exactly the budget the price waterfall reserves 0.250 for. It is not an accounting identity, since the reserve is managed across the whole pipeline rather than loan by loan, but it is the right mental model: the extension was paid for in advance, by every loan on the sheet.
• Worst-case repricing on expired locks exists because a free option is worth money. A borrower who could re-lock at the better of the old rate and the new one after letting a lock expire would hold a valuable option the lender never sold.
• The desk asks whether a file is real — is there an executed contract, is the appraisal ordered, is the borrower committed — because pull-through is the question, and a loan officer with a reputation for locking speculative files is an expensive customer of their own secondary desk. That reputation is tracked. It shows up as slower exception approvals and less flexibility on extensions when you need them.

Chapter 30 takes the lock decision itself — when to lock, when to float, float-downs, extensions, renegotiations, and reprices. This section is the reason those rules are shaped the way they are. Read them knowing that on the other side of every one of them is a desk trying to be indifferent to rates while people keep changing their minds.

29.10 Rebuilding the Linden Street quote from scratch

Put the sheet away. Build it from nothing.

The file, as of 9:42 a.m. on day 12. Conventional 30-year fixed. Contract price \$385,000, 5% down (\$19,250), loan amount **\$365,750, LTV 95.00%, CLTV 95.00% with no subordinate lien. Representative score 706 — the lower of two middle scores, 742 and 706, established on day 1. Primary residence. Single-family detached, 1,780 square feet, built 1994. Purchase. Escrows established. 30-day lock**, because the contract closing is day 45 and today is day 12.

Step 1 — the base price. Turn to the conventional conforming 30-year fixed block on the sheet effective 8:15 a.m. The loan amount falls in the \$175,001-to-conforming-limit band. At 6.625% the base price is 100.750, quoted for a 15-day lock.

Step 2 — the score/LTV cell. A 706 representative score is in the 700–719 row. A 95.00% LTV is in the 90.01–95.00 column — and note that 95.00 is the top of that band, not the bottom of the next one. The cell is 1.125. Subtract it.

$$100.750 - 1.125 = 99.625$$

Step 3 — every other LLPA. Primary residence: 0.000. One-unit detached: 0.000. Purchase: 0.000. Thirty-year fixed: 0.000. No subordinate financing: 0.000. Escrows established rather than waived: 0.000. Six lines, six zeros. The running price is unchanged at 99.625.

Step 4 — the lock period adjustment. Base pricing on this sheet is quoted at 15 days. This lock is 30. Subtract 0.125.

$$99.625 - 0.125 = 99.500$$

Step 5 — the final price, converted to dollars.

$$\text{points} = 100.000 - 99.500 = 0.500$$ $$\text{dollars} = 0.500\% \times \$365{,}750 = \$1{,}828.75$$

The quote: 6.625%, 0.500 discount point, \$1,828.75, 30-day lock, P&I \$2,341.94.

Every basis point of price is accounted for. Base 100.750, less 112.5 basis points for the score/LTV cell, less 12.5 basis points for the lock, equals 99.500. Total adjustment 125 basis points — **\$4,571.88** on this loan, of which \$4,114.69 is the credit score and down payment and \$457.19 is the calendar. Nothing else on this file cost a single basis point of price.

Repricing an advertised rate — the method, on a different file

The reason to learn this is not to reproduce a quote you already have. It is to be able to take a number somebody else advertised and find out what it actually means for a specific person. Here is that skill on a file that has nothing to do with Linden Street.

The advertisement: "6.500%, no points." In rate-sheet terms, that is a claim that the final price at 6.500% is exactly 100.000.

The file [constructed teaching example — not one of this book's four running files]: a \$280,000 conventional 30-year fixed purchase, 720 representative score, 85.00% LTV, primary residence, single-family detached, 45-day lock.

REPRICING AN ADVERTISED RATE                  [constructed teaching example]

  Base price, 6.500%, 30-yr fixed, 15-day lock ..........  100.250
    LLPA  score 720-739 / LTV 80.01-85.00 ...............   -0.625
    LLPA  occupancy PRIMARY .............................    0.000
    LLPA  property SFD 1-UNIT ...........................    0.000
    LLPA  purpose PURCHASE ..............................    0.000
    ADJ   lock period 45 DAYS ...........................   -0.250
  ───────────────────────────────────────────────────────────────
  TOTAL ADJUSTMENTS .....................................   -0.875
  FINAL PRICE ...........................................   99.375

  points  = 100.000 − 99.375 = 0.625
  dollars = 0.625% × $280,000 = $1,750.00

  So "6.500% with no points" is, for this borrower, 6.500% plus $1,750.00.

And now the second half of the skill, which is the part that keeps the borrower rather than just winning the argument: tell them what their real choices are. Run the same subtraction up the ladder for this file. At 6.625% the base is 100.750, so the final price is $100.750 - 0.875 = 99.875$ — a cost of 0.125 point, \$350.00. At 6.750% the base is 101.250, so the final price is $101.250 - 0.875 = 100.375$ — a credit of 0.375 point, \$1,050.00.

The honest quote is therefore not "they're lying." It is: "6.500% is real, and on your file it costs \$1,750. If you'd rather not write that check, 6.625% costs \$350 and 6.750% pays you \$1,050 back. Here's what each does to your payment." That is a conversation the advertisement cannot have, and it is worth more than a quarter point of rate.

The Linden Street borrowers were also shopping an online lender with a lower advertised rate. Chapter 40 reprices that quote for this file using precisely the method above and assembles the complete comparison. Do not guess at the answer. Build it.

29.11 Why the grid says what it says

§29.4 handed you the matrix and told you to apply it mechanically — every file, every time, the same way. That is the correct professional instruction and it is not an explanation. A loan officer who knows why each cell holds the number it holds can do two things the one who merely knows the grid exists cannot: tell a borrower which of their facts are worth changing, and recognize a quote that could not have come from any published grid at all.

Underneath every adjustment sit the same two questions — how likely is this loan to default, and how much is lost when it does? The quantity being priced is expected loss, and expected loss is a product, not a sum:

$$\text{expected loss} = \text{probability of default} \times \text{loss given default}$$

An LLPA is a one-time charge in price covering an expected lifetime loss. The credit score axis is mostly the first term — how often. The loan-to-value axis is almost entirely the second — how badly. That is why the largest table on the sheet is two-dimensional, and why its axes multiply rather than add.

🧮 Run the Numbers

The two axes, taken apart. (Constructed teaching model. The agencies set the published matrices from their own loss experience and their capital requirements, not from this arithmetic. The model reproduces the shape, not the values.)

Start with severity. Take the Linden Street house at \$385,000 and assume a total cost of default of 25% of value — unpaid interest, legal fees, preservation, brokerage, and the concession a distressed sale carries. That figure is constructed and round; the real one swings with state foreclosure procedure. It leaves net proceeds of $\$385{,}000 \times 0.75 = \$288{,}750$. Run four loan sizes against that one number:

LTV	Loan	Net proceeds	Shortfall	As a % of the loan
95%	\$365,750 | \$288,750	\$77,000	21.05%
90%	\$346,500 | \$288,750	\$57,750	16.67%
80%	\$308,000 | \$288,750	\$19,250	6.25%
60%	\$231,000 | \$288,750	none — a \$57,750 surplus	0.00%

One input moved and the loss went from a fifth of the loan to nothing. That is the entire economic content of the loan-to-value axis, and it is why the ≤60.00% column of Figure 29.2 holds zeros at every score of 680 and above: at that leverage there is nothing left for a borrower's credit history to damage.

Now the other axis. Assume — constructed again, for shape only — lifetime default probabilities of 1.0% at 780 and above, 2.5% at 700–719, and 6.0% at 640–659, and multiply each by the severity above:

Score band	At 95% LTV (21.05% severity)	At 80% LTV (6.25% severity)
780+, 1.0%	0.211 points	0.063 points
700–719, 2.5%	0.526 points	0.156 points
640–659, 6.0%	1.263 points	0.375 points

Read the corners. The strongest borrower at low leverage costs six hundredths of a point; the weakest at high leverage costs about twenty times that. Neither axis alone produces that spread — the product does, which is why the grid is a grid and not two lists.

Two caveats, because an uncaveated model turns into a fact. Every loss above is stated before mortgage insurance — which is precisely why a conventional loan over 80% LTV must carry it, and why the deeper coverage at the top of the ladder is what §29.4 meant about the >95.00% column. And these values do not match the published grid's and are not meant to. The takeaway is narrower: when you move one band and the price moves, you now know which quantity you moved.

One thing the frequency axis is not. A credit score makes no statement about any individual — it says that among borrowers whose files look like this one, some proportion will go ninety days past due. Nothing in the grid is a judgment about the household on the other end of the phone, and a loan officer who explains it as one has made the conversation both wrong and worse.

Occupancy: which payment gets made first

A second home or an investment property prices worse than a primary residence at every score and every LTV, and the mechanism is not moral. It is a statement about the order in which a household under pressure makes its payments: when money runs short, people pay for the roof they sleep under, and the weekend place is the payment that stops. On an investment property a second mechanism stacks on top — the payment depends on a tenant, and a vacancy removes the income supporting the loan while the taxes, insurance, and maintenance keep arriving.

It is also why occupancy misrepresentation is treated so seriously — Chapter 27 owns what that is and what it costs. The pricing point is narrower: the adjustment is small enough that no borrower should ever be tempted, and the consequence is large enough that no loan officer should ever help.

Purpose: why cash-out is its own matrix

§29.4 said cash-out refinances carry a separate matrix, materially worse than the purchase grid at every score and LTV. Three mechanisms produce that, and they compound.

It consumes the cushion directly. Run the severity arithmetic above in reverse. A borrower at 60% loan-to-value sits in the column where a default costs the lender nothing; take cash out to 80% and they have moved into a column where it costs 6.25% of the loan. The borrower experiences a cash-out as receiving money. The grid experiences it as the deliberate removal of the thing protecting the loan.

There is no arm's-length price. On a purchase, two unrelated parties negotiated a number and one wrote a check for it; the appraisal checks that number rather than being its only source. On a cash-out there is no buyer and no seller — the value rests entirely on one appraiser's opinion, produced for a borrower who benefits from a higher one. A softer value input makes the LTV less reliable, which makes the severity estimate less reliable, and that is why valuation scrutiny on a cash-out is heavier than borrowers expect.

Adverse selection. Averaged across a population, households converting equity into cash are under more financial pressure than those that are not — though many individually are not, and pricing is done on populations.

Together they give the reason behind the instruction §29.4 offered: a borrower who wants \$40,000 out of a house is often better served by a second lien (Chapter 35). A cash-out reprices the entire first mortgage into a worse matrix. A second lien prices only the \$40,000.

The rest of the stack, in one line each

The same two questions answer the remaining adjusters. A condominium carries correlated risk: a unit's value depends on an association the borrower does not control, so a structural assessment moves every unit at once, and losses arriving together are worse for a portfolio than the same losses arriving independently. Subordinate financing is priced on combined loan-to-value because a borrower at 80% LTV behind a 15% second has 5% of their own money at stake and behaves like it. An escrow waiver prices a lien that outranks yours — in most jurisdictions a property tax lien is senior to the mortgage, so an unescrowed borrower who stops paying taxes puts the first lien position itself at risk; verify how that works in your state. Small loans price worse because manufacturing cost does not shrink when the loan does. Shorter terms price better because they amortize fast, so the severity model starts improving in month one and never stops.

Which facts are movable

None of this lets you argue with a grid. What it buys is the answer to the only pricing question a borrower can act on — which of my facts are movable? Score is (Chapter 10). Down payment is, if the money exists. Purpose sometimes is: a second lien instead of a cash-out. Occupancy and property type are not, and a conversation that treats them as movable has stopped being a pricing conversation and become a fraud conversation.

29.12 Following the loan: who takes the risk, and who is paid for it

The chapter has built the price twice. Follow the loan instead — from the moment somebody says a number on the phone to the moment a pension fund owns a share of the payment — naming at each hand-off who accepts which risk and what they collect for accepting it.

Month one is \$2,341.94 of principal and interest, of which \$2,019.24 is interest, plus \$176.78 of mortgage insurance alongside it. §29.1 split that interest three ways; here is the same split with a column that changes what it means.

Three of those rows deserve more than a line.

The lender's risk does not end when the loan is sold

This is the most misunderstood item on the list, including among people who have been in the business for years. When a lender sells a loan to an agency it does not sell away responsibility for whether the loan was made properly. It sells subject to representations and warranties — promises that the income was documented, the appraisal supported the value, the borrower occupied the property, the file complied. If the loan defaults early or an audit finds a defect, the agency can require the lender to repurchase it: to buy back a non-performing asset at par, months or years later.

Put that next to the revenue. §29.9 established the lender's gross on this file at \$8,686.56; the repurchase exposure is \$365,750, a loan the lender would own outright and probably worth materially less than par by the time anyone demanded it. Roughly forty-two to one between what a defective file can lose and what a good one earned — and every underwriting question you have found tedious lives inside that ratio. The underwriter is not being difficult about a paystub date; they are the only control the lender has over a liability forty-two times its revenue that outlives the sale. It runs the other way too: a loan officer whose files are consistently clean reduces a real corporate exposure, which is why file quality eventually shows up in a shop's margin — and therefore in the base price column at the top of §29.1.

The servicer advances, and the investor waits

Two rows on that table are quietly doing more than their dollar figures suggest.

The servicer's \$76.20 a month sounds like a rounding error until you see the duty: on an agency pass-through the servicer must advance scheduled principal and interest to the security holder even in months when the borrower has not paid. A servicer with a delinquent book funds somebody else's mortgage payments out of its own liquidity — while running escrow analyses, disbursing \$4,620 of taxes a year, and answering the phone in month forty-one. That is why servicing is a scale business and why mortgage servicing rights trade as an asset (Chapter 28).

The security holder is not exposed to whether these borrowers pay — the agency's \$114.30 a month buys that away, which is the entire reason a pension fund can own this loan without analyzing it. What the investor is exposed to is when they pay. An investor who bought this coupon at a premium is counting on above-market interest for years; if rates fall and the borrower refinances in month eight, the premium is gone and the money returns to be reinvested at the new, lower rate. That is prepayment risk, and it is why §29.7's price cap of 103.000 exists: a loan priced far above par refinances at the first opportunity, so nobody pays unlimited premium however high the coupon goes. The cap is not a rule somebody invented. It is a bond investor's arithmetic, arriving at your desk with the paperwork stripped off.

The shape of the whole chain

Month one is a snapshot, and every proportion in it holds for all 360 payments, because each strip is a fixed share of the same note rate. The guarantee and the servicing together take 9.43% of every interest dollar this household pays — a real cost of having a thirty-year fixed-rate mortgage available at all, not a fee anybody can negotiate away.

Which leaves the asymmetry to carry out of this section: the loan officer is paid once, inside the first fifty-one days, and every other participant is paid across three decades. Most of the incentive problems this industry has had live in that fact, and it is why Chapter 26's compensation rule exists and why Chapter 24's disclosure rules point where they point.

29.13 Running the sheet backwards

Everything so far has run the arithmetic in one direction: base price, minus adjustments, equals final price, converted to dollars. The direction that earns a living is the other one. It answers three questions a loan officer is asked constantly and usually answers badly — what does that quote actually mean?, what does the lender who gave it have to be doing to produce it?, and what is the lowest rate this borrower can reach for the money they have? All three are the same subtraction, moved.

THE THREE REVERSALS

  1.  QUOTE  ->  FINAL PRICE
      A quoted rate with a cost or a credit IS a final price.
          final price = 100.000 - points quoted
      "6.500% with 0.375 point"  ->  final price 99.625
      "6.750%, $4,500 credit on $450,000"  ->  1.000 credit  ->  101.000

  2.  FINAL PRICE  ->  IMPLIED BASE PRICE
      Add this file's adjustments back on.
          implied base = final price + total adjustments
      The agency grid is published and applies to every seller, so the
      adjustments are (mostly) not what differs. The BASE is.

  3.  BUDGET  ->  RUNG
      points affordable = cash available / loan amount x 100
      Then find the lowest rate whose points are at or below it.
      The ladder moves in eighths. There is nothing in between.

Work them on a file that has nothing to do with Linden Street.

The file [constructed teaching example — not one of this book's four running files]: a \$600,000 purchase, 25% down, loan amount **\$450,000, LTV 75.00%, 745 representative score, primary residence, single-family detached, 30-day lock, escrows established. On Figure 29.2's grid the 740–759 row at 70.01–75.00 is 0.125; the 30-day lock is 0.125; everything else is zero. Total adjustments: 0.250.**

THE SAME BASE COLUMN, A DIFFERENT FILE       [constructed teaching example]
$450,000 conventional 30-year fixed, 745 score, 75.00% LTV, 30-day lock.
Base column from Figure 29.1. Every row: FINAL = BASE - 0.250.

  RATE     BASE      FINAL     POINTS    $ TO BORROWER        P&I
  ────────────────────────────────────────────────────────────────────
  7.000%  102.000   101.750    -1.750      ($7,875.00)    2,993.86
  6.875%  101.625   101.375    -1.375      ($6,187.50)    2,956.18
  6.750%  101.250   101.000    -1.000      ($4,500.00)    2,918.69
  6.625%  100.750   100.500    -0.500      ($2,250.00)    2,881.40
  6.500%  100.250   100.000     0.000    par     $0.00    2,844.31
  6.375%   99.625    99.375    +0.625       $2,812.50     2,807.41
  6.250%   98.875    98.625    +1.375       $6,187.50     2,770.73
  6.125%   98.000    97.750    +2.250      $10,125.00     2,734.25
  ────────────────────────────────────────────────────────────────────
  This file's PAR RATE is 6.500%.

Same sheet, same morning, same base column — and the par rate moved from 6.750% to 6.500% because this file's adjustments are 0.250 instead of 1.250. One number relocated the entire menu by a full eighth of rate.

Reversal one and two: what a quote implies

A borrower brings in a written quote from elsewhere: 6.500% with 0.375 point.

Reversal one: 0.375 point of cost means a final price of $100.000 - 0.375 = 99.625$. Your final price at 6.500% is 100.000, so you are 0.375 of price better at the identical rate — \$1,687.50 the borrower does not have to write a check for.

Reversal two is the part almost nobody does. Add this file's adjustments back onto their final price to see what their base column must say: $99.625 + 0.250 = 99.875$. Their base at 6.500% is 99.875; yours is 100.250. The published agency matrix is the same document for both of you — the LLPA is charged to every seller delivering to that agency — so unless one lender is stacking overlay adjusters of its own, the whole 0.375 difference sits in the line §29.1 called margin: not a better cost of funds, not a relationship, not a mistake, but a different appetite for volume this week. Two subtractions got you a sentence you can defend on a phone call.

When the implied base is impossible

Now the version that saves you from an argument you cannot win by rate-shopping.

A borrower has seen an advertisement: 6.125%, no points. No points means a final price of exactly 100.000, so the implied base is $100.000 + 0.250 = 100.250$. Your sheet prints 98.000 at 6.125% — and 100.250 is what it prints three eighths higher, at 6.500%. The implied base is 2.250 points above yours, and nothing on a conventional conforming sheet differs by 2.250 points.

So the honest conclusion is not "they are lying." It is that quote is not describing this loan — it is describing a different product, a different lock period, a temporary buydown somebody is funding, a rate available only to a file with no adjustments at all, or a price from a different morning. Each has a question attached:

• What lock period is it priced for, and what is the effective date on the sheet it came from?
• Is it fixed for thirty years, or an initial rate on an adjustable?
• Is there a temporary buydown — who is funding it, and what does the rate become?
• What score, down payment, occupancy, and property type was it quoted on?

Ask those four and the number usually resolves itself, without you ever having to say anything about the other lender.

Reversal three: from a budget to a rung

"What can I get?" is a budget, so convert it to points first: \$3,000 is 0.667 points on \$450,000. At 6.375% the cost is 0.625 point — **\$2,812.50, inside it. At 6.250% it is 1.375 points — \$6,187.50**, or \$3,375.00 more, which is Figure 29.1's 0.750 step arriving in dollars. Nothing sits in between, because a lender that prints eighths does not quote 6.3125%.

29.14 The four facts, and the conversation that follows

Everything in this chapter fails at the same place: the moment somebody says a number before they know what they are pricing.

The four facts

A quote requires four inputs and there is no honest substitute for any: a score that was pulled rather than remembered; a loan-to-value built from a real price and money the borrower can document, provisional until the appraisal lands; what the property is and how it will be occupied, which the deed answers and the borrower frequently cannot; and how many days you actually need, which §29.5 prices and Chapter 30 decides.

With those four you can build a price in ninety seconds. Without them you can build a number, and it will be wrong in the direction that hurts, because a loan officer guessing at inputs guesses optimistically. Nobody has ever assumed a worse score than the borrower turned out to have.

📞 On the Phone

How to explain a price adjustment without sounding like you are making an excuse.

This is the call that follows the one above, and most loan officers handle it badly.

Borrower: "Wait. You're saying our credit score is costing us four thousand dollars? Our credit is fine. We've never missed a payment in our lives."

The wrong answer: "That's just what the agencies charge, I don't have control over it." True, entirely passive, and it teaches the borrower nothing except that you are not the person to talk to.

The other wrong answer: "Well, if you were at a 720 it would be a lot better." Now their score is a personal failing and you are the one who said so.

What actually works — lead with the number, not the reason. An explanation that arrives before the number sounds like a defense. A number that arrives before the explanation sounds like a fact.

"You're right that your credit is good — you're at a 706 and there's nothing on the report I'd change. Here's what's actually happening. Two things about this loan move the price: the score and the fact that you're putting five percent down. Together they're worth 1.250 points, and the score is 1.125 of that — about \$4,114 on your loan amount. It's a published grid, the same one at every lender that sells to Fannie Mae, and it prices in bands: 700 to 719 is one band, 720 to 739 is the next one up. You're fourteen points from the next band, and that band is worth 0.375 — about \$1,371 on this loan. So the question isn't whether the charge is fair. It's whether fourteen points is reachable before we lock, and that's a question I can actually answer if you'll give me ten minutes with your balances."

Four things happen there. You validated the fact — their credit is good. You gave the number first, in dollars, before any explanation. You named the mechanism as external and published, which is the difference between "this is what your loan costs" and "this is what I decided to charge you." And you ended on the movable part, because a borrower handed a problem with an action attached almost never stays angry. The version that goes wrong is the one where you never say the dollar figure and the borrower finds it on a Loan Estimate — which from their side looks exactly like concealment, even when it was only avoidance.

When an input moves after you have quoted

The four facts are not permanent. Three of them can move between the quote and the closing table, and every move has a price attached.

A score that re-pulls lower. Credit reports expire — the window is a Selling Guide question, so verify it rather than remember it — and this file carries a pre-closing refresh as condition 11 precisely because characteristics drift. Most refreshes change nothing. Some do.

🧮 Run the Numbers

Seven points of score, after the lock.

Suppose a pre-closing refresh — on any file, at any lender — comes back seven points lower than the pull the loan was priced on: a 706 representative score becomes a 699. On Figure 29.2's grid, 706 sits in the 700–719 row and 699 sits in the 680–699 row. At 90.01–95.00% loan-to-value:

	Score/LTV cell	Total adjustments	Final price at 6.625%	Points	Dollars
Priced at 706	1.125	1.250	99.500	+0.500	\$1,828.75
Repriced at 699	1.500	1.625	99.125	+0.875	\$3,200.31

Check it: $100.750 - 1.500 - 0.125 = 99.125$, so $100.000 - 99.125 = 0.875$ point, and $0.875\% \times \$365{,}750 = \$3{,}200.31$. The increase is $\$3{,}200.31 - \$1{,}828.75 = \$1{,}371.56$ — 0.375 of price, the exact width of one row, and the same \$1,371.56 that fourteen points in the other direction was worth in Figure 29.2's lesson. The grid is symmetric. Nothing else about the situation is.

Put the number where it lands: \$1,371.56 is cash due at closing, on a household whose reserves after closing are \$12,623.66 — just under eleven percent of everything they have left. It arrived from seven points on a credit score, thirty-two days after the rate was locked, on a file where nobody did anything wrong. That is not what happened here; the day-44 refresh found a new debt rather than a lower score, and Chapter 19 owns what that cost. The point of the counterfactual is that the same condition catches both, and only one of them is the one everybody prepares for.

An appraisal that moves the LTV band. This is the Cypress Court file, and its pricing consequence has not been drawn out. A \$540,000 contract with 20% down was a \$432,000 loan at exactly 80.00% LTV; the appraisal returned at \$505,000, and because LTV is computed on the lesser of price or value that same loan is now $\$432{,}000 \div \$505{,}000 = 85.54\%$ — which skips 80.01–85.00 entirely and lands in 85.01–90.00. Through the middle of the grid, from 700 to 759, that move costs 0.250 of price — \$1,080.00** on \$432,000; it costs 0.375 (\$1,620.00) in the 660–699 rows and 0.500 (\$2,160.00) at 640–659, because the weaker the file, the more a column is worth. And even that is the small part: the same appraisal pushed the loan above 80% LTV where mortgage insurance is required and was in nobody's budget, and broke the structure outright, since the maximum 80% loan against \$505,000 is **\$404,000 and the down payment rises from \$108,000 to \$136,000 — the \$28,000 gap Chapters 18 and 20 work through. A short appraisal is a cash problem, an eligibility problem, and a pricing problem in one email.

A property type discovered late. The borrower said townhouse; the appraisal says condominium. That happens constantly, because "townhouse" describes the shape of a building and says nothing about the form of ownership. On the \$450,000 file from §29.13, assume a condominium adjuster of 0.750 above 75.00% LTV [constructed teaching grid — verify current pricing at the source]: discovering it after the lock adds \$3,375.00 to a file quoted without it, and triggers a project review nobody ordered on a calendar with no room for one.

So tell the borrower which facts are load-bearing on the first call. "If the appraisal comes in under contract, or your score moves, or the property turns out to be a condominium, the price changes and I will call you the same day." Said on day 1 that is professionalism; said on day 44 it is an excuse. Behind it sits a mechanical fact — a lock is taken on a file, not merely on a rate — so when a characteristic changes, the desk reprices the existing lock against the new facts. Whether the resulting change in points may be redisclosed is a tolerance question Chapter 22 owns.

🗂️ The Loan File

Chapter 29 contribution: rebuild the quote from a base price, and account for every basis point.

The file already contained the answer — 6.625% with 0.500 discount point, \$1,828.75, locked on day 12 for thirty days. What it did not contain until now was the derivation.

What this settles. Where the rate came from, completely. The 6.625% was not negotiated, discovered or won. It is the fourth rung of a published ladder, moved 1.250 points by two facts about this file — a 706 representative score at 95% loan-to-value, and a thirty-day promise about a calendar. The par rate for this file was 6.750%; every rate below it costs cash and every rate above it pays cash, on a schedule the borrower could have been shown on day 1.

What it does not settle. Whether locking on day 12 was the right day to lock, what to do when the desk reprices, and what the lock expiration on day 42 actually cost and who should pay it — Chapter 30. Nor does it settle whether the borrowers should have taken the credit instead of paying the point; Chapter 13 made that call, and Chapter 40 tests it against what the competition was offering.

Open questions carried forward:

• Q29.1. Day 12 was a lock decision as much as a pricing one. Was it the right day? (Chapter 30)
• Q29.2. The 15-day extension on day 42 cost 0.250 point, \$914.38, and the lender absorbed it. Under what circumstances does that cost land on the borrower instead? (Chapter 30)
• Q29.3. The borrowers had a lower advertised quote in hand. Repriced for this file, what is it actually worth? (Chapter 40)

Your task. In the Appendix C workbook, reproduce the table above from the rate sheet in Figure 29.1 and the grid in Figure 29.2 without looking at this page. Then do it again changing exactly one input: make the representative score 720 instead of 706. Recompute the final price at 6.625%, convert it to dollars, and write one sentence stating what fourteen points of credit score was worth to this household in cash. Then write a second sentence about what you would have done about it on day 1.

Conclusion

A rate is not set. It is assembled, from a base price that came off a trading screen and a stack of adjustments that came off a published grid.

The base price starts with the security coupon the loan will deliver into — the note rate less servicing and less the guarantee fee — and is adjusted by the value of the servicing strip, the lender's margin, and a reserve for hedging and fallout. That produces the rate sheet, which quotes price, not cost, on a scale where 100.000 is par. Above par is a rebate the lender can pay; below par is cash the borrower brings. One subtraction converts between them, and one multiplication turns the answer into dollars.

Onto that base price go the loan-level price adjustments — the score-by-LTV grid first and largest, then occupancy, property, purpose, product, subordinate financing — and then the lock period adjustment, which prices time rather than risk. What is left is the final price, and the final price is the quote.

On the Linden Street file that arithmetic runs 100.750 − 1.125 − 0.125 = 99.500, which is 0.500 point, which is \$1,828.75. And the number that answers Chapter 1 is the one at the top: at a base price of 100.750, that same 6.625% would have paid a \$2,743.12 credit to a file without this file's adjustments. The advertised rate was never a lie. It was a price for somebody else, and the whole of a loan officer's usefulness is the ninety seconds it takes to find out what it is for the person on the phone.

Everything in this chapter has assumed the price was captured. It was not, automatically — somebody had to decide, on day 12, to stop looking and take it.

Next: Chapter 30 is that decision. When to lock and when to float, what a reprice is and how to survive one, what a float-down actually costs, how extensions are priced, and what happens on the day a lock expires with the file not clear to close — which is, as it happens, exactly what day 42 of this file looked like.

Key Terms

Rate sheet — the document a lender publishes each business day listing, by product and note rate, the price at which it will buy a loan, together with the adjustments that must be applied to it. (Ch.29)

Base price — the price printed on the rate sheet at a given rate before any loan-specific adjustment; on the Linden Street file, 100.750 at 6.625%. (Ch.29)

Par — a price of exactly 100.000: the loan is worth exactly the loan amount, so no discount point is paid and no rebate is generated. The par rate is the rate whose final price is 100.000. (Ch.29)

Premium pricing — choosing a note rate whose final price exceeds 100.000 so the loan generates a rebate, typically applied as a lender credit toward the borrower's closing costs. (Ch.29)

Rebate — the surplus generated by a price above par; the source of a lender credit. (Ch.29)

Yield spread — the relationship between note rate and price: a higher rate is worth more to an investor, so it prices higher. Historically, surplus paid to a broker from this spread was called yield spread premium; the LO compensation rule changed who may receive it (Chapter 26). (Ch.29)

Loan-level price adjustment (LLPA) — a published price adjustment charged for a specific risk characteristic of a specific loan — credit score, LTV, occupancy, property type, purpose, product, subordinate financing — applied in points of price and cumulative. (Ch.29)

Lock period adjustment — the price adjustment for the length of the rate lock, reflecting the cost of carrying a hedge for longer and the higher chance the loan never funds. (Ch.29)

Buy-up / buy-down grid — the rate-and-price ladder itself: what each eighth of rate costs in price going down, and pays in price going up. On a typical sheet the cost per eighth rises as the rate moves further below the coupon the loan delivers into. (Ch.29)

Pricing engine / product and pricing engine (PPE) — the software that determines eligibility, retrieves base prices, applies every adjustment, and returns a sorted grid of rate/price combinations; a lookup-and-rules system, only as reliable as the characteristics entered into it. (Ch.29)

Lender margin — the spread the lender builds into the base price to cover the cost of manufacturing the loan and to earn a profit; set by management and widened or narrowed with capacity and appetite. (Ch.29)

Hedging — taking an offsetting market position, typically by selling securities forward, so that the value of a locked pipeline and the value of the hedge move in opposite directions and the lender is roughly indifferent to rate movement. (Ch.29)

Pull-through — the share of locked loans that actually fund. It falls when rates improve and rises when they worsen, which is why it moves against the hedge in both directions and why it is priced into every rate sheet. (Ch.29)

Spaced Review

1. A rate sheet shows a base price of 101.125 at a given rate. The file's adjustments total 2.000 and the lock adjustment is 0.250. What is the final price, is the borrower paying or receiving, and how much on a \$420,000 loan?

2. (Chapter 4) The Linden Street borrowers paid \$1,828.75 for a half point and saved \$30.31 a month, a 60.3-month break-even. Explain, using §29.6, why the next half point would have taken longer to break even than the first — and name the two separate reasons.

3. (Chapter 28) A 6.625% note rate delivers into a 6.0% pass-through coupon. Name the two deductions that account for the gap, say who receives each, and state which of them is negotiated between the lender and the agency rather than fixed by convention.

4. (Chapter 4) A borrower is quoted a rate at par with an origination charge of 1.000% and no discount point. A colleague says, "there are no points on that loan, so the APR equals the note rate." Identify both errors in that sentence.

5. Rates fall three-quarters of a point over two weeks. Explain, in three sentences and without using the word "hedge," why the lock desk becomes noticeably less flexible about extensions that month.