Chapter 37: Purchase vs. Refinance Markets: How to Pivot Your Business When Rates Change

"Nobody builds a referral base in the month they discover they need one. That is the whole lesson, and it costs a great many people their careers to learn it." — constructed; the argument of this chapter

Overview

There are two mortgage businesses in the United States. They share a license, a rulebook, a software platform, and a job title, and they are otherwise almost unrelated. One of them sells houses money. The other one sells money to people who already have houses.

A loan officer who has only ever done one of them will tell you, in complete sincerity, that they know how this business works. They do not. They know how half of it works, in one direction of one rate cycle, and the half they know is the half that is about to disappear.

That is not a moral claim. It is arithmetic. Refinance demand is a function of one variable: the gap between the rate a household already has and the rate available today. When that gap is wide, the phone rings without any effort at all, and it rings from people who found you on a website and will leave for eighth of a point. When the gap closes — and it closes fast, because rates move faster than pipelines do — the population of borrowers who could benefit from refinancing does not shrink by half. It goes to something very close to zero, because everyone who was going to refinance already did, at a rate below today's, and nothing you say will make refinancing at a higher rate sensible. The demand did not soften. It was consumed.

Meanwhile the purchase business, which is slower, harder, and less lucrative per file in a boom, keeps running. Not untouched — affordability falls, inventory tightens, contracts get harder — but it does not go to zero, because people still get married, get transferred, get divorced, have children, and die, and every one of those events moves a household.

This chapter is about that asymmetry and about the two decisions it forces. The first is a borrower decision: is this particular refinance actually good for this particular household, and how do you compute that honestly when three different arithmetic shortcuts all lie in the same direction? The second is a business decision: what do you build, and when, so that the day the refinance market ends you still have a job.

The two decisions are the same decision, seen from opposite ends. A loan officer whose income depends on doing refinances will find reasons to do refinances. A loan officer with a purchase business does not need this borrower's refinance badly enough to argue themselves into a bad one. That is not a character difference. It is a structural one, and it is the most honest argument for diversification anyone will ever make you.

In this chapter, you will learn to:

• Describe purchase and refinance origination as two different businesses with different lead sources, calendars, competitors, and failure modes
• Define rate-and-term, cash-out, and streamline refinances and state what each does to LTV, pricing, and eligibility
• Compute a refinance break-even correctly, and catch the three errors that make one look better than it is
• Show the amortization reset in dollars on a specific file
• Apply a net tangible benefit test and document it
• Explain the harm mechanism in serial refinancing and the prevention discipline
• Explain why refinance volume collapses rather than declines, and build a purchase business before you need one

Learning Paths

🎓 Exam — §37.3 and §37.4. Know the rate-and-term versus cash-out distinction cold, know that the right of rescission applies to a refinance of a principal residence and not to a purchase, and know that "streamline" is a program-specific term with program-specific tests. 🏠 New LO — §37.5 and §37.9. §37.5 is the arithmetic you will be asked to do in your first refinance conversation; §37.9 is the reason you will still be employed in three years. 🤝 Partner — §37.8 and §37.9. If you are a real estate agent, §37.8 explains why your lender partner's attention swings with the ten-year Treasury, and §37.9 tells you which lenders will still be there next cycle. 📊 Operations — §37.1 and §37.8. Purchase and refinance files consume operations capacity in completely different shapes, and the staffing mistake in §37.8 is made by somebody in every cycle.

37.1 Two businesses wearing the same name

Start with what is identical. Both businesses take a Uniform Residential Loan Application. Both pull credit, document income and assets, order a valuation and a title product, run automated underwriting, issue a Loan Estimate, satisfy conditions, and close under the TILA-RESPA Integrated Disclosure rule. Both are licensed activity. Both are subject to the same fair lending obligations, the same compensation rule, and the same ability-to-repay requirements.

Now look at everything else.

The lead

A purchase loan arrives attached to a transaction that already exists. Somebody wrote an offer. Somebody accepted it. There is a contract with dates in it, an earnest money deposit at risk, a listing agent, a buyer's agent, a seller, and a closing date that other people's moving trucks are scheduled around. You were introduced into that transaction by a referral partner or by the borrower's own choice, and the reason you are in it is that somebody vouched for you.

A refinance loan arrives attached to nothing. There is no contract, no counterparty, no deadline other than a rate lock you created yourself, and no third party who will be inconvenienced if the borrower changes their mind. The borrower found you through an advertisement, a mailer, a call center, a rate table, or the servicer's own retention department. Nobody vouched for you. Nobody will be embarrassed if they leave.

That single structural difference produces almost every other difference between the two businesses.

The clock

The purchase business has an externally imposed clock. Chapter 20 works the purchase contract in detail and Chapter 30 works the rate lock; between them, a purchase loan officer lives inside a calendar somebody else wrote. The Linden Street file is a fair example: a contract naming a day-45 closing that actually left forty-one days, a lock taken on day 12 for thirty days, and a file that took fifty-one days and cost a lender-paid extension to do it. Every day was chargeable to something.

A refinance has no external clock. The lock is the only deadline, and the lock exists because you created it. Nothing happens on the day a refinance does not close except that it closes later. This sounds like a relief and is in fact the source of the refinance business's characteristic failure: files drift, borrowers restart their shopping mid-process, locks expire, and volume that felt enormous in April does not fund in June.

The competition

In a purchase transaction, you are usually competing for the referral, not for the loan. Once the borrower has chosen you and the offer is accepted, changing lenders mid-contract is possible but costly and slow, and most borrowers will not do it unless you give them a reason. Your competitor is whoever the agent might have called instead.

In a refinance, you are competing for the loan continuously, right up to the closing table, against every advertised rate in the country. Nothing binds the borrower to you. A borrower who is three weeks into a refinance and sees a lower number has lost nothing by starting over except three weeks they were not spending anyway. Refinance shopping is rational in a way purchase shopping is not, which is why refinance pricing runs thinner and why refinance shops live and die on speed and marketing spend.

The emotional register

This one gets underrated and it changes how you talk. A purchase borrower is frightened. They are making the largest commitment of their life, on a deadline, with somebody else's house on the line, and they are frequently doing it for the first time. They want reassurance, competence, and somebody who will pick up the phone.

A refinance borrower is doing optional arithmetic on an asset they already own. They are not frightened; they are skeptical. They believe — correctly — that you make money when this closes and that you would therefore like it to close. Your job is not reassurance. It is credibility, and the fastest way to earn it is to be visibly willing to tell them not to do it.

THE SAME LICENSE, TWO BUSINESSES

                       PURCHASE                     REFINANCE
  ────────────────────────────────────────────────────────────────────────────
  where the lead       a referral partner,          advertising, a database,
  comes from           a past client                a servicer's retention desk
  what already         a contract, an earnest       nothing
  exists               deposit, a closing date
  the clock            written by someone else      written by you (the lock)
  the counterparty     seller, two agents,          none
                       title, sometimes an HOA
  competing against    other loan officers, for     every advertised rate,
                       the referral                 continuously
  the borrower is      frightened                   skeptical
  what kills the file  the contract deadline,       drift, and a lower number
                       the appraisal, a stip        somewhere else
  volume is driven by  household formation,         the gap between their rate
                       inventory, affordability     and today's rate
  volume changes       slowly                       in a step
  ────────────────────────────────────────────────────────────────────────────

Read the last two rows again, because they are the business argument of this chapter and everything in §37.8 and §37.9 follows from them.

Purchase volume is driven by demographic and economic forces that move on the scale of years. Households form, jobs relocate, families grow. Affordability tightens and loosens and changes what people buy more than whether they buy. Inventory constrains the market, sometimes severely, but the underlying demand does not evaporate.

Refinance volume is driven by a single spread, and the population it draws from is exhaustible. Every household with a mortgage has a note rate. When market rates fall, some fraction of those households move into the money — their existing rate exceeds today's rate by enough to justify the cost of a new loan. Those households refinance. And then they are no longer in the population, because their new rate is today's rate. The pool refills only when rates fall further. If rates stop falling, the pool empties and stays empty. If rates rise, the pool does not merely stop refilling; every loan written during the low-rate period becomes permanently out of the money, and the population of refinance candidates approaches zero and stays there for years.

Secondary-market analysts have a word for the emptying of that pool: burnout. Chapter 28 discusses it from the investor's side, where it is a prepayment-modeling problem. From your side of the desk it is simpler. It is the reason the phone stops.

📞 On the Phone

Borrower: "I got a letter saying rates dropped and I could save four hundred dollars a month. Is that real?"

The wrong answer: "It could be! Let me get you an application started and we'll see what we can do." You have just committed to finding a way to say yes.

The other wrong answer: "Those mailers are junk." Sometimes they are. But rates did drop, the savings might be real, and dismissing the question makes you the person who did not want to look.

What actually works: "Maybe. Four hundred a month is what it costs to make the sentence interesting, not what it costs to make it true. I need four things: what rate are you at right now, what's your balance, do you have mortgage insurance on that loan, and how long do you plan to be in the house. Give me those and I'll tell you in ten minutes whether this is worth doing, and I will tell you if it isn't."

Notice the fourth question. Nobody in a refinance advertisement asks how long the borrower intends to stay, because the answer is the single input most likely to make the pitch collapse. Ask it first, write the answer down, and put it in the file. §37.4 explains why that note is worth more than anything else you will write that day.

37.2 The refinance calculation, honestly

A refinance is a decision about total cost over a horizon, and every shortcut that reduces it to something simpler makes it look better than it is.

Here is the honest test, stated once, in full. It is not complicated; it is merely inconvenient, which is why it goes undone.

The net position test. Pick a horizon $H$, in months — realistically how long this household will keep this loan. Then, for each of the two loans:

1. Add up every dollar the borrower pays out over those $H$ months: principal and interest, plus mortgage insurance if the loan carries it, plus any closing costs paid at the table.
2. Add the loan balance still owed at the end of month $H$.

That sum — cash paid plus balance owed — is the borrower's net position. The loan with the lower number is the better loan over that horizon. There is no fifth consideration and no adjustment factor. Cash out the door plus debt still standing, both loans, same date.

Taxes and property insurance appear in both columns identically and cancel; you may include them or omit them so long as you are consistent. Escrow funding is not a cost at all — it is a transfer, and the old escrow account is refunded, typically within a matter of weeks after payoff. Including the new escrow deposit as a "cost" while ignoring the old escrow refund is a small, common, and entirely avoidable distortion.

The four inputs

Everything in the test reduces to four facts, and you should have all four before you quote anything.

One: the rate on the loan they have. Not the rate they remember. Borrowers misremember their note rate constantly, usually downward when they are proud of it and upward when they are hoping. Get the note, the mortgage statement, or the servicer's payoff. While you are looking, get the payment and find out what is inside it, because a borrower quoting you "my payment is \$3,033" is quoting principal, interest, taxes, insurance, and mortgage insurance, and only two of those five change when they refinance.

Two: what the new loan will actually cost. All of it. Origination, discount, appraisal, credit, flood, tax service, title, settlement, recording, and whatever your state adds. If the offer is a "no-cost" refinance, the costs are in the rate, and you should be able to show the borrower both versions side by side — Chapter 29 builds the rate from the price so you can.

Three: how long they will keep the loan. Ask. Write it down. This is the input that decides more refinance questions than the rate does, and it is the only one you cannot look up. Household tenure in a home is far shorter than thirty years on average, and the loan's life is shorter still, because a rate move ends it. Do not substitute a statistic for the borrower's own answer; ask the household in front of you.

Four: what happens to the term and the mortgage insurance. These two are where the errors live, and §37.5 and §37.10 are about nothing else.

The rules of thumb are all wrong

You will hear that a refinance makes sense at a one-point rate improvement, or two points, or whenever the payment drops by \$100. Every one of these is wrong in the same way: it ignores loan size, cost structure, remaining term, mortgage insurance, and horizon. A rate improvement is worth dollars, and dollars are a function of the balance it is applied to.

Two households, identical 100-basis-point improvement — 6.500% to 5.500%, thirty-year fixed — and completely different answers.

On a \$150,000** balance, the payment falls from \$948.10 to \$851.68: a saving of **\$96.42 a month. At closing costs of \$4,250 (1.5% plus about \$2,000 of fixed charges), the naive break-even is $\$4{,}250 \div \$96.42 = 44.1$ months.

On a \$500,000** balance, the same rates give \$3,160.34 and \$2,838.95: a saving of **\$321.39 a month. Costs of \$9,500 on the same structure give a naive break-even of $\$9{,}500 \div \$321.39 = 29.6$ months.

Same rate move. Same cost formula. Forty-four months versus thirty months — and that is before anyone has looked at the term or the mortgage insurance. A household planning to move in three years should do one of these refinances and not the other, and the "one percent rule" cannot tell them which. (Illustrative; all figures constructed.)

Loan size matters because the benefit scales with the balance while a meaningful share of the cost does not. Appraisals, credit reports, flood certifications, settlement fees, and recording charges are roughly fixed. This is a real and under-discussed equity issue in refinance lending: the same rate relief is worth less, per dollar of cost, to a borrower with a smaller loan — which correlates with a smaller house, a lower income, and a market with less appreciation. Nothing in this chapter fixes that, but a loan officer who understands it will stop pushing marginal refinances at borrowers for whom the arithmetic genuinely does not work, and will start saying so out loud.

And the appraisal is not free of consequence

One more input that purchase-trained loan officers underweight and refinance-trained loan officers learn the hard way. A refinance requires a value, and unlike a purchase there is no contract price to anchor it and no seller to renegotiate with.

The Cypress Court file is the book's standing lesson on short valuations: a \$540,000 contract where the appraisal came back at \$505,000, \$35,000 low, and the 80% loan fell from \$432,000 to \$404,000, opening a \$28,000 gap eleven days before closing. In a purchase, that gap has three possible resolutions — the buyer brings more money, the seller reduces the price, or the parties walk — and Chapters 18 and 20 work all three.

Run the same short appraisal through a refinance and notice what is missing. There is no seller. There is no price to renegotiate. There is no earnest money at risk and no contract to terminate. What there is instead is a loan-to-value ratio that moved against the borrower, and every consequence of that ratio arriving at once: a different mortgage insurance factor or the sudden requirement of mortgage insurance where none was contemplated, a worse loan-level price adjustment, possibly a program the borrower no longer qualifies for at all, and a reconsideration-of-value process (Chapter 18) that takes days the borrower's lock may not have. The refinance does not blow up with a bang. It just quietly stops being worth doing, usually after the borrower has already paid for the appraisal.

Order the value early, and tell the borrower before you order it exactly what value the analysis requires. That sentence is the entire operational lesson of this subsection, and §37.10 puts a specific dollar figure on it.

37.3 Rate-and-term, cash-out, and streamline

Three names, three different products, three different rulebooks. Confusing them is the most expensive vocabulary error in refinance lending, because the differences run to loan-to-value limits, pricing, documentation, and sometimes eligibility itself.

Rate-and-term (limited cash-out)

A rate-and-term refinance replaces an existing loan with a new one to change the rate, the term, or both. The new loan pays off the balance of the existing first lien, pays the closing costs of the new transaction, and delivers essentially no money to the borrower.

Fannie Mae calls this a limited cash-out refinance; Freddie Mac calls it a no cash-out refinance; the industry says "rate-and-term." They are the same thing. "Limited" is the operative word: the agencies permit a small amount of incidental cash back at closing — currently the lesser of a small percentage of the new loan amount or a small dollar cap — precisely so that a rounding difference in the payoff does not reclassify the whole transaction. Verify the current limit in the Fannie Mae Selling Guide or the Freddie Mac Seller/Servicer Guide before you rely on it; it is exactly the kind of figure that gets revised.

Two things commonly surprise people about what still counts as rate-and-term. First, paying off a purchase-money second lien — a second taken out at the same time as the original purchase, to buy the same property — is generally treated as rate-and-term rather than cash-out, because the borrower is not extracting equity, merely consolidating debt that was always part of buying the house. Second, paying off a second lien that was not purchase-money — a home equity line drawn after closing, for instance — is generally cash-out, no matter what the borrower spent the money on. Both rules have conditions and seasoning requirements attached and both get revised; treat the principle as durable and the specifics as perishable.

Cash-out

A cash-out refinance replaces an existing loan with a larger one and delivers the difference to the borrower in cash. This is home equity extraction: converting accumulated equity, which is illiquid and does not have to be repaid until the house is sold, into money, which is liquid and now carries a payment for thirty years.

Cash-out is priced and underwritten as a materially riskier transaction, and it is treated that way in three simultaneous places:

• Maximum loan-to-value falls. Conventional cash-out on a one-unit primary residence is currently capped at a substantially lower LTV than rate-and-term — the industry benchmark for years has been 80% — and the caps drop further for second homes, investment property, and multi-unit. Chapter 35 owns occupancy and product as a subject; do not learn the grid here, learn that there is one and that it is not the rate-and-term grid.
• Pricing worsens. Cash-out carries its own loan-level price adjustment, layered on top of the score and LTV adjustments Chapter 29 builds. On a marginal file the cash-out adjustment alone can move the rate a quarter point or more.
• Documentation and seasoning tighten. Ownership seasoning requirements, restrictions on properties recently listed for sale, and limits on using a newly obtained value are all normal.

Streamline

A streamline refinance is a program-specific, reduced-documentation refinance available only within certain government programs, in which some combination of appraisal, income verification, and credit documentation is waived because the investor already carries the risk on the existing loan and is not increasing its exposure.

The two you will meet are the FHA Streamline Refinance, which Chapter 16 owns, and the VA Interest Rate Reduction Refinance Loan (IRRRL), which Chapter 17 owns. Go to those chapters for the mechanics, the eligibility, the mortgage insurance premium treatment, and the tests. What belongs here is the structural point: a streamline is cheap and fast because it does not re-underwrite the risk, and a product that does not re-underwrite the risk is a product that can be sold to a borrower who should not have it. That is not a hypothetical concern. It is the exact history that produced the net tangible benefit requirements in §37.4 and the anti-churning rules in §37.7, and it is why streamline programs carry the strictest benefit tests in the entire refinance universe.

That last row is not a detail. Every refinance secured by a borrower's principal residence with a lender other than the current holder gives the borrower a right to rescind the transaction for a defined period after consummation, under Regulation Z. Purchase-money loans have no such right. Chapter 23 teaches the mechanics, the timing, and what a defective notice does to the period — go there. What you need from this chapter is the operational consequence: a refinance does not fund at the signing table. There is a waiting period, disbursement happens after it, and any borrower you have told "you'll have the money Friday" needs to be told the truth instead.

🎓 NMLS Exam Watch

Four things the test likes here, and the trap in each.

1. Rescission. The stem will describe a transaction and ask whether the right of rescission applies. Purchase-money loans: no. Refinance of a principal residence with a new lender: yes. Second homes and investment property: no — the right attaches to the borrower's principal dwelling. The trap is a stem that says "primary residence" while describing a purchase.

2. Rate-and-term versus cash-out. The stem will hide the answer in what is being paid off. A new loan paying off the first lien and the closing costs is rate-and-term. A new loan paying off the first lien and a home equity line the borrower drew last year to buy a boat is cash-out, even though the borrower receives no money at the table. Cash-out is defined by what the proceeds retire, not by whether a check is written.

3. Streamline. Candidates answer "no appraisal" and stop. The tested distinction is that a streamline is program-specific — there is no such thing as a conventional streamline — and that it carries a mandatory benefit test precisely because the underwriting is reduced.

4. Net tangible benefit. Know it as a requirement imposed by programs, investors, and a number of states, not as a courtesy. The trap: the stem offers a payment reduction achieved by extending the term and asks whether the benefit test is satisfied. Extending the term to lower a payment is precisely what the test exists to catch.

37.4 Net tangible benefit

Net tangible benefit is the requirement that a refinance leave the borrower measurably better off, in a way that can be stated and checked, rather than merely differently off.

It exists because the alternative was tried. The refinance transaction has an unusual property among consumer credit products: it can be repeated on the same borrower indefinitely, it generates origination revenue every time, and each repetition can be made to feel beneficial to the borrower because the payment goes down. Nothing about a lower payment guarantees a lower cost. §37.5 shows exactly how far those two can diverge, and §37.7 covers what happened when an industry discovered the gap and industrialized it.

Where the requirement comes from

Net tangible benefit is not one rule. It is a family of rules arriving from at least four directions, and a given file may be subject to several at once:

Program rules. The government streamline programs carry explicit, quantified tests. The FHA Streamline Refinance requires a defined improvement — expressed in terms of a reduction in the combined rate, or a specified change in term — and HUD Handbook 4000.1 is the authority; Chapter 16 works it. The VA's IRRRL and cash-out programs carry statutory requirements added by federal legislation in 2018 aimed squarely at repeat refinancing: seasoning of the loan being refinanced, recoupment of all fees and costs within a defined number of months, and a minimum rate reduction that differs depending on whether the borrower is going fixed-to-fixed or adjustable-to-fixed. Chapter 17 works those. Verify the current thresholds with the agency — every one of these numbers has been revised and will be again.

Investor and lender rules. Aggregators and lenders impose their own benefit tests as overlays, sometimes stricter than the program's, because a loan that refinances too quickly damages the value of the security it was pooled into. Chapter 28 explains why an investor cares.

State law. A number of states impose net tangible benefit tests by statute, typically attached to refinances of owner-occupied property, sometimes with prescribed worksheets and sometimes with private rights of action attached to a violation. These vary enormously and change. Your compliance department knows which ones apply to you. Ask before you need to know.

The ability-to-repay rule. Chapter 24 owns ATR. Note only that a refinance is a new extension of credit and gets the full analysis; there is no "they already have this loan" exception.

The practitioner's version

All of that is the floor. Here is the discipline that actually protects a borrower and a career:

Write the benefit down, in one sentence, in the file, before the borrower signs anything — and write it so that a stranger reading the file in three years could evaluate whether it was true.

A sentence that survives that test looks like: "The borrower's total cost of credit over their stated four-year horizon falls by \$8,400 net of all closing costs, the loan term is not extended, and the mortgage insurance obligation is unchanged." A sentence that does not survive it looks like: "Lower payment."

There are, in the end, only three legitimate benefits, and it is worth being able to name them:

1. Lower total cost over the borrower's horizon, net of every cost, with the term held constant or shortened.
2. A materially safer structure — most commonly an adjustable-rate loan converted to a fixed rate before it adjusts, or a balloon retired before it matures. This can be worth doing even at a higher total cost, because the borrower is buying the removal of a risk. Say so explicitly when that is the trade; it is honest and it is defensible.
3. A use of equity with no cheaper source — a genuinely necessary expense the household cannot fund any other way at a better price. §37.6 is about which ones those are.

Everything else is a payment illusion, a term extension, or someone's commission.

📄 Read the File

text FIGURE 37.1 — "The benefit, written down" [constructed teaching example] THE DOCUMENT Refinance benefit worksheet, one page, dated and signed by the loan officer at application and placed in the file. Not a required agency form; a shop-level discipline that also satisfies several state tests. THE CONTEXT A household six years into a 30-year loan. Original amount $300,000 at 7.000%; P&I $1,995.91; 72 payments made; balance $278,074; 288 payments remain. Today's achievable rate is 6.000%. Total closing costs quoted at $6,000, which the borrower intends to finance. WHAT IT SHOWS Line 1 Current note rate 7.000% Line 2 Proposed note rate 6.000% Line 3 Current P&I $1,995.91 Line 4 Proposed P&I, new 30-year term $1,703.17 Line 5 Monthly payment reduction $292.74 Line 6 Total closing costs $6,000.00 Line 7 Payment-based break-even (L6 / L5) 20.5 months Line 8 Borrower's stated horizon "5 to 7 years" Line 9 Term: current remaining / proposed 288 / 360 months Line 10 Mortgage insurance: current / proposed none / none WHAT IT DOESN'T Line 7 is not a break-even. It is the ratio of one cost to one saving, and it silently assumes the two loans amortize identically, that the $6,000 is repaid on the day it is recovered, and that a payment reduction is a saving. None of the three is true. Line 9 records the 72-month term extension and then does nothing with it. Nowhere on this page does the total cost of either loan appear. THE DECISION Do not present Line 7 to the borrower. Add three lines before you print it: total remaining payments on the existing loan, total payments on the proposed loan, and the same comparison at a term that is not extended. Then present all three options and let the household choose. THE LESSON A benefit worksheet that ends at the payment reduction is not a benefit worksheet. It is a sales document with a compliance heading on it, and the arithmetic that makes it wrong is in the next section.

Constructed. Line 8 is the single most important entry on the page and the one most frequently left blank.

37.5 Break-even, and the three ways it is computed wrong

This is the technical core of the chapter.

The break-even of a refinance is the point at which the borrower is better off having done it. Every loan officer knows the formula:

$$\text{break-even (months)} = \frac{\text{closing costs}}{\text{monthly payment saved}}$$

and that formula is wrong in three separate ways, each of which pushes the answer in the same direction — toward making the refinance look better than it is. They compound. A refinance can clear a twenty-month break-even by that formula and be a net loss to the household after five years.

Work the file from Figure 37.1 all the way through and watch it happen.

The file. Original loan \$300,000 at 7.000%, thirty-year fixed, P&I \$1,995.91. Seventy-two payments made — six full years. Balance \$278,074. Two hundred eighty-eight payments remain. Today's achievable rate is 6.000%. Closing costs \$6,000, financed into the new loan, so the new loan amount is \$284,074 and the new thirty-year payment is \$1,703.17.

The pitch. "Your rate drops a full point, your payment drops \$292.74, and you break even in twenty and a half months. After that it's \$292.74 a month in your pocket for the next twenty-eight years." Every number in that sentence is arithmetically correct. The conclusion is false.

Error one — ignoring the reset of amortization

A thirty-year mortgage does not distribute its interest evenly. In the early years almost the entire payment is interest and almost none of it is principal; that ratio inverts slowly across the term. Chapter 4 taught the mechanism and drew the curve. The consequence nobody carries forward is what happens when a borrower who has climbed part of that curve returns to the bottom of it.

Our borrower has made seventy-two payments. On payment 73 of the existing loan, the split is:

$$\text{interest} = \$278{,}074 \times \frac{0.07}{12} = \$1{,}622.10$$

leaving \$1,995.91 − \$1,622.10 = \$373.81 of principal, or 18.73% of the payment.

Now look at payment 1 of the new loan:

$$\text{interest} = \$284{,}074 \times \frac{0.06}{12} = \$1{,}420.37$$

leaving \$1,703.17 − \$1,420.37 = \$282.80 of principal, or 16.60% of the payment.

Read that twice. After refinancing to a rate a full percentage point lower, a smaller share of a smaller payment goes to principal. They were retiring \$373.81 of debt a month. Now they retire \$282.80. The rate went down and the debt reduction went down with it, because they gave back six years of progress along the curve.

🧮 Run the Numbers

What the amortization reset costs in the first twelve months.

Keep the existing loan. Twelve more payments at 7.000% take the balance from \$278,074 to \$273,441.51 — a principal reduction of **\$4,632.49**.

Refinance into a new thirty-year at 6.000% on \$284,074. Twelve payments take the balance to \$280,585.50 — a principal reduction of **\$3,488.50**.

After twelve months	Keep	Refinance
Payments made	12 × \$1,995.91 = \$23,950.92	12 × \$1,703.17 = \$20,438.04
Principal retired	\$4,632.49 | \$3,488.50
Balance owed	\$273,441.51 | \$280,585.50

The refinanced borrower has paid \$3,512.88 less in cash over the year. They also owe \$7,143.99 more than they would have. That gap decomposes exactly:

$$\$6{,}000.00 \ (\text{financed costs}) + \$1{,}143.99 \ (\text{foregone amortization}) = \$7{,}143.99$$

Net position at twelve months, cash paid plus balance owed:

• Keep: \$23,950.92 + \$273,441.51 = \$297,392.43
• Refinance: \$20,438.04 + \$280,585.50 = \$301,023.54

The refinanced household is \$3,631.11 worse off at the one-year mark. Check it against the components: \$6,000 of costs, minus \$3,512.88 of cash saved, plus \$1,143.99 of lost principal reduction, equals \$3,631.11.

Run the same test forward and the true crossover — the month at which the refinance finally puts this household ahead on net position — arrives at month 32. The pitch said twenty and a half. The honest answer is a year and a half longer, and that is with a full point of rate improvement. (Constructed file; amortization computed on the stated balances.)

The reason the naive formula misses this is subtle and worth stating precisely. The formula compares a cost (closing dollars) to a cash flow (monthly payment reduction), and treats the cash flow as if it were profit. It is not. Part of the "saving" is simply money the borrower is no longer paying toward their own principal — money that was never an expense in the first place, because it was going into the house rather than to the lender. A payment reduction achieved by paying down less debt is not a saving. It is a transfer from the borrower's balance sheet to the borrower's checking account, and it costs interest to make.

Error two — comparing payment to payment instead of total cost

The second error is the same mistake with the horizon removed, and it is the more spectacular of the two.

Compare what this household will actually pay, all in, under each choice.

Keep the existing loan. Two hundred eighty-eight payments of \$1,995.91:

$$288 \times \$1{,}995.91 = \$574{,}822.08$$

Refinance into a new thirty-year. Three hundred sixty payments of \$1,703.17:

$$360 \times \$1{,}703.17 = \$613{,}141.20$$

The refinance costs **\$38,319.12 more**. A full point of rate improvement, a payment \$292.74 lower, and the household pays thirty-eight thousand dollars more for the privilege — because they bought back six years of loan.

Now do it honestly. Refinance at the same rate into a term that does not extend: 288 months, so the loan retires on exactly the same date it would have. The payment on \$284,074 at 6.000% over 288 months is \$1,863.46.

$$288 \times \$1{,}863.46 = \$536{,}676.48$$

That is \$38,145.60 less than keeping the existing loan, over the identical horizon, ending on the identical date.

The two refinance columns differ by \$76,464.72 — on the same borrower, the same house, the same day, the same rate, and the same lender. The only difference is a number typed into the term field.

That is the entire argument for total-cost comparison, and it is why the payment reduction is a worse guide than no guide at all. The option with the larger payment reduction is the option that costs the household seventy-six thousand dollars more. A borrower shown only the payments will choose exactly wrong, every time, and will thank you for it.

Error three — ignoring the costs rolled into the balance

The third error hides inside the phrase "no money out of pocket."

Our borrower financed \$6,000 of closing costs. That was presented as convenience — nothing due at closing — and it is convenient. It is not free. That \$6,000 is now principal, borrowed at the note rate, amortized over three hundred sixty months.

Price it. The new loan is \$284,074 with a payment of \$1,703.17. The same loan without the financed costs — \$278,074 at 6.000% for 360 months — carries a payment of \$1,667.19. The difference:

$$\$1{,}703.17 - \$1{,}667.19 = \$35.98 \text{ per month, for 360 months}$$

$$360 \times \$35.98 = \$12{,}952.80$$

Twelve thousand nine hundred fifty-two dollars and eighty cents to avoid writing a check for six thousand. The interest alone is \$6,952.80 — more than the costs themselves. The household will pay for these closing costs 2.16 times over, and will still be paying for them in year twenty-nine.

🧮 Run the Numbers

The three errors, stacked, on one file.

The pitch: costs \$6,000, saving \$292.74/month, break-even 20.5 months.

What the naive formula says	What is actually true
Break-even 20.5 months	Net-position break-even month 32
"\$292.74/month saved" | \$91.01/month of that is principal they stopped paying
"\$6,000 in costs" | \$12,952.80 paid over 360 months to finance \$6,000
"Total saving \$292.74 × 360" | Total cost **\$38,319.12 higher** than doing nothing

And the option nobody was shown:

288-month refinance at 6.000%
Payment	\$1,863.46 (−\$132.45)
Total P&I	\$536,676.48
Versus keeping the existing loan	\$38,145.60 saved
Versus the 360-month refinance	\$76,464.72 better

One conversation, one number typed into one field, seventy-six thousand dollars. (Constructed file. Every figure derives from the stated balance of \$278,074 at the stated rates.)

So what is the correct calculation?

Return to §37.2. Cash paid plus balance owed, both loans, same horizon. Everything above is a consequence of that test being skipped.

If you want a shortcut you can defend, use this one instead of the payment formula:

Compare total remaining cost over the term that ends on the same date. Take the existing loan's remaining payments and add any remaining mortgage insurance. Take the proposed loan's payments over the same number of months, add its mortgage insurance, add its closing costs, and add whatever balance is still outstanding on the proposed loan at that date if its term is longer. Lower wins.

The term-matched comparison is not just a fairness device. It is how you find the option in the third column of the table above — the one that saves the household thirty-eight thousand dollars and would never have appeared on the worksheet in Figure 37.1.

Two practical notes before we leave the arithmetic.

Most lenders will write a custom term. You are not restricted to 30, 20, and 15. If a borrower has 288 months remaining, ask for 288 months. Many lenders quote custom terms as a standard offering and price them off the nearest bracket. If yours will not, the 25-year is usually closer to honest than the 30-year.

The "skipped payment" is not skipped. Borrowers are frequently told that a refinance lets them miss a month's mortgage payment. What actually happens is that the payoff includes interest through the payoff date, the new loan collects prepaid interest from closing to the end of the month, and the first payment on the new loan is not due until the month after that. The month is not skipped; the interest for it is paid at the closing table, and the calendar merely moved. Selling a refinance on a skipped payment is selling a borrower their own money.

Three questions before you go on

Answer these out loud before you read §37.6, because everything after this section assumes them.

One. A borrower's payment drops \$300 a month and their monthly principal reduction drops \$110. How much of the \$300 is a saving? *(\$190. The other \$110 is money they have stopped paying toward their own principal — a transfer from their balance sheet to their checking account, made with borrowed money.)*

Two. Why does financing the closing costs make the payment-based break-even formula more wrong rather than less? This is the one people get backwards. Financing drives the formula's numerator toward zero — nothing out of pocket — while simultaneously raising the payment, which shrinks the denominator. Both halves of the ratio move at once, in opposite directions, for the same reason. The result stops measuring anything at all.

Three. A loan officer says "we can do this with no closing costs." What did they change, and where would you look to find it? (They moved the costs into the rate as a lender credit. Compare the quoted rate against the day's par pricing for the same lock period, LTV, score, and product — Chapter 29 builds the rate from the price so you can. Then note that this is not a trick: for a borrower with a short horizon it is frequently the correct structure. It is misrepresented only when it is sold as free.)

37.6 Cash-out: the good reasons and the expensive ones

Everything above assumed the borrower wanted a better loan. Now assume they want money.

A cash-out refinance converts equity into cash. That is a real and sometimes valuable financial transaction, and mortgage debt is, for most households, the cheapest borrowing available to them by a wide margin. It is also the transaction most likely to leave a household worse off in a way they will not notice for years, because it does three things at once and only one of them is visible.

What the borrower sees: money.

What the borrower does not see: the loan just got bigger, the term just reset, and debt that used to be unsecured — enforceable only against the borrower's income and credit — is now secured by their house.

That third one is not an accounting distinction. Unsecured debt in default produces collection activity, credit damage, and sometimes a lawsuit. Secured debt in default produces a foreclosure. Any conversation about consolidating unsecured debt into a mortgage must include that sentence, out loud, and a loan officer who skips it is not doing the job.

The reasons that hold up

A necessary expense with no cheaper source. A roof, a failing system, a medical obligation sitting at punitive interest, a legal necessity. The test is comparative: what is the alternative, what does the alternative cost, and is the mortgage genuinely cheaper over the horizon the borrower will carry it?

Home improvement that is both wanted and durable. Note the honest framing. Not "improvements that pay for themselves," which is a claim you are not licensed to make and which the renovation literature does not support as a general rule; Chapter 35 covers renovation products where the value question is underwritten explicitly. The defensible version is that the household wants the improvement, will use it for years, and the mortgage is the cheapest way to fund it.

Removing a worse loan. Retiring a high-rate second, a balloon about to mature, or an adjustable lien before it adjusts, where the arithmetic works over the horizon.

A structural need in the household's balance sheet. A divorce buyout, an estate settlement, a business obligation with documented terms. These are real and they are common, and the loan officer's job is to make sure the number requested is the number actually needed.

The reason that usually does not

Consolidating revolving debt into a thirty-year mortgage. This is the single most common cash-out pitch and the one that most reliably makes households worse off, and the reason is arithmetic rather than morality.

🧮 Run the Numbers

The consolidation, priced honestly.

A household carries \$40,000 of revolving balances at a blended 22.9% and is paying \$1,000 a month against them. They own a \$300,000 home with a \$180,000 first mortgage at 6.000% and are offered a cash-out refinance to \$220,000 — a 73.3% loan-to-value, inside the conventional cash-out cap.

What they are doing now. At \$1,000 a month against \$40,000 at 22.9%, the balances retire in about 76 months — six years and four months — for a total of roughly \$76,240, of which about \$36,240 is interest. Expensive, painful, and finite.

What the consolidation does. The \$40,000 slice, added to a 30-year mortgage at 6.000%, carries a marginal payment of:

$$\$40{,}000 \times \frac{0.005}{1-(1.005)^{-360}} = \$239.82 \text{ per month}$$

$$360 \times \$239.82 = \$86{,}335.20$$

The comparison.

	Pay it off	Roll it into the mortgage
Monthly outlay	\$1,000.00 | \$239.82
Months	~76	360
Total paid	~\$76,240** | **\$86,335.20
Difference	—	about \$10,095 more
Debt is	unsecured	secured by the house

The consolidation is not a saving. It is the purchase of \$760.18 a month of cash flow for about \$10,095 in additional total dollars, plus the closing costs of the refinance, plus the cash-out price adjustment, plus the conversion of unsecured debt into a lien on the family home.

It can still be the right answer. If the household cannot actually sustain \$1,000 a month, the comparison is not "\$76,240 versus \$86,335"; it is "\$86,335 versus default," and the consolidation is obviously correct. The loan officer's job is to know which of those two problems is on the table, and the way you find out is by asking whether the \$1,000 has actually been paid every month for the last year, and looking at the credit report to see whether the answer is true. (Constructed file. Revolving payoff computed at a constant \$1,000 payment; blended rate illustrative.)

There is a second half to that analysis and it decides most of these files. The consolidation only works if the freed cash flow goes somewhere. If the household redirects the \$760.18 to the new mortgage as extra principal, or to savings, or to anything at all, the picture changes completely and the transaction can be excellent. If the revolving balances rebuild — and the reason underwriters and investors treat repeat consolidators as elevated risk is that the balances frequently do — the household ends up with the old unsecured debt and a larger mortgage and less equity, which is strictly the worst outcome available.

You cannot control that. You can name it, once, plainly, and put the borrower's answer in the file: "Asked what the \$760 a month would be used for. Borrower states it will go to the auto loan and then to reserves." That note costs nothing and it is the difference between an originator who advised and one who processed.

Finally, the limits. Cash-out caps loan-to-value materially lower than rate-and-term, prices worse, and often requires ownership seasoning. On a file that is already close to a threshold, the request for cash may be the thing that moves it over — and the borrower who asks for "maybe twenty thousand, if it's easy" should be told what twenty thousand does to the rate before they decide it is easy. Chapter 35 covers occupancy and product; Chapter 29 covers what the adjustment does to the price.

37.7 Churning and serial refinancing

Everything in §37.5 describes how a single refinance can be made to look better than it is. This section is about what happens when someone does it repeatedly, to the same borrower, on purpose.

Serial refinancing — the industry also calls it churning, or loan flipping — is the practice of refinancing a borrower repeatedly, at intervals short enough that each transaction's costs have not been recovered before the next one begins, principally for the benefit of the party originating the loans. It is a documented harm with a real enforcement history, and it is the direct cause of most of the rules in §37.4.

The mechanism

Understand it as an engine with four cylinders, all of which you have now seen in isolation.

One: financed costs compound. Each refinance rolls its costs into the balance. The second refinance's balance includes the first refinance's costs, and finances them again. A household that refinances four times in six years, financing \$6,000 each time, has added \$24,000 to their mortgage before a single dollar of benefit is counted — and is paying interest on the first \$6,000 for the third time.

Two: amortization resets. §37.5 priced this. Every reset returns the household to the front of the curve, where the principal share of the payment is smallest. A borrower who has been "paying a mortgage" for twelve years may have retired almost nothing, because they have never held a loan long enough to reach the part of the curve where principal accumulates.

Three: the mortgage insurance clock resets. §37.10 works this in detail. Mortgage insurance terminates on a schedule keyed to the loan's origination. A new loan is a new schedule.

Four: equity leaves. Every cash-out extracts it and every financed cost erodes it. Equity is the household's protection against a market decline, their ability to sell in a hurry, and their retirement asset. Strip it steadily and a household with an appreciating asset can end a decade owning less of it than when they started.

Add a fifth if the era permits it: prepayment penalties, which historically were used to make each refinance more expensive and to trap the borrower with the originating lender until a fee was paid. The Ability-to-Repay and Qualified Mortgage framework (Chapter 24) sharply restricted prepayment penalties on most residential mortgages; the history is why.

None of the four cylinders is visible in a payment comparison. That is the point. Every mechanism of harm in serial refinancing is invisible to the arithmetic most loan officers do.

The enforcement history, in outline

The pattern has been recognized and legislated against repeatedly, in specific and verifiable ways. Three markers are worth knowing:

The Home Ownership and Equity Protection Act of 1994 amended the Truth in Lending Act to impose heightened requirements on high-cost mortgages. The abuses it was enacted to address were concentrated in the refinance market and centered on equity stripping from homeowners — frequently older homeowners with substantial equity and limited income — through repeated, fee-laden refinances. HOEPA is the first federal statutory recognition that a refinance can be a mechanism of harm rather than a benefit.

The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 created the Consumer Financial Protection Bureau, established the Ability-to-Repay requirement, and produced the Loan Originator Compensation rule. Chapter 26 works compensation properly; the relevant fact here is that the rule prohibits compensating a loan originator based on the terms of a transaction, which removes one specific incentive to steer a borrower into a worse structure.

Federal legislation enacted in 2018 — the Economic Growth, Regulatory Relief, and Consumer Protection Act — added protections aimed specifically at repeated refinancing of VA loans, after a documented pattern of veterans being refinanced at high frequency. The protections require seasoning of the loan being refinanced, recoupment of all fees within a defined period, and a minimum rate improvement that varies by the type of refinance. Ginnie Mae separately imposed seasoning requirements before rapidly-refinanced VA loans could be pooled, which changed the economics for the lenders doing it. And the Consumer Financial Protection Bureau has brought enforcement actions concerning deceptive advertising of VA refinances — mailers designed to appear to come from a government agency, quoting savings figures that could not be delivered. Chapter 17 covers VA lending; the pattern is the point here.

Verify the current form of every one of those requirements before relying on it. Thresholds, seasoning periods, and recoupment windows have all been revised. The durable fact is the direction of travel: every decade, someone industrializes the refinance transaction, and every decade the rules tighten around it.

⚖️ Compliance Check

What this means at your desk, this week.

• Net tangible benefit is a requirement, not a philosophy. Program tests (FHA, VA), investor overlays, and a number of state statutes each impose one, and they are not identical. Know which apply to your files. Your compliance department has the list.
• Seasoning and recoupment rules govern how soon. Government refinance programs impose waiting periods measured from the first payment due date of the loan being refinanced, and require that fees be recouped within a defined number of months. A file that clears the benefit test can still fail the seasoning test.
• Compensation may not vary with terms. Under Regulation Z's Loan Originator Compensation rule, your compensation cannot be based on a transaction's terms. Chapter 26 works it. Its relevance here is that it removes one specific incentive to prefer a worse structure.
• Advertising is regulated. A refinance solicitation that misrepresents the savings, implies a government affiliation it does not have, or quotes a payment that omits mortgage insurance or escrows can be a deceptive act. Do not send mail you have not read against the rules, and do not forward a marketing vendor's template without review.
• Document the benefit contemporaneously. A one-line, dated benefit statement in the file at application costs nothing and is the single best evidence that the decision was made for the borrower's reasons.

Requirements change and state law varies enormously — several states impose their own net tangible benefit tests with their own worksheets and their own remedies. Verify current requirements with your compliance department and your state regulator before you rely on anything in this section.

Writing it from the prevention side

Here is the discipline, stated as things you actually do.

Ask when the current loan closed, every time, first. If the answer is inside a year, the burden of proof is on the refinance, not on leaving it alone. There are legitimate reasons to refinance a loan that is nine months old — a large rate move, a removal of mortgage insurance, an adjustable loan converted to fixed — and you should be able to state which one applies before you take an application.

Ask what they paid in costs last time, and whether those costs were financed. If the borrower refinanced eighteen months ago and financed \$5,800 of costs, you are proposing to finance those costs a second time. Say that out loud.

Compute the recoupment yourself, honestly, using §37.5's method. Not the payment formula.

Never present a refinance as a rescue for a borrower in distress without saying what it does not fix. A household behind on unsecured debt with a payment problem has a payment problem; a refinance moves the problem's due date and adds a lien. Sometimes that is right. It is never right silently.

Decline the ones that do not clear. This is the whole thing, and it is a business decision as much as an ethical one. A loan officer who will decline a marginal refinance is a loan officer whose borrowers believe them when they say a refinance is good — and belief is the only durable asset in this business. Chapter 38 turns that into a pipeline.

37.8 What happens to a refinance shop when rates rise

Now the business argument, which is the reason this chapter sits in Part VIII rather than Part III.

Volume does not decline. It collapses.

Return to the exhaustible pool from §37.1. The population of refinance candidates at any moment is the set of outstanding mortgages whose note rate exceeds today's achievable rate by enough to justify the costs. That is not a smooth function of rates. It is closer to a threshold.

Watch what happens as rates move.

THE REFINANCE POOL — schematic, not to scale

  Suppose the outstanding book of mortgages is distributed across note rates
  like this, and today's achievable rate is the line marked NOW.

  note rate   3.0%   3.5%   4.0%   4.5%   5.0%   5.5%   6.0%   6.5%   7.0%
              ████   ████   ███    ██     ██     ███    ████   ███    ██
                                          NOW ↑
                                          └─── in the money ────────────┘

  Rates fall 100bp. The line moves left. Everything to its right refinances,
  once, and re-enters the book AT the new rate:

  note rate   3.0%   3.5%   4.0%   4.5%   5.0%   5.5%   6.0%   6.5%   7.0%
              ████   ████   ███    ██     ██████████████ ·      ·      ·
                                   NOW ↑                (now empty)
                                   └── in the money ──┘

  Rates rise 150bp from there. The line moves right, past everything:

  note rate   3.0%   3.5%   4.0%   4.5%   5.0%   5.5%   6.0%   6.5%   7.0%
              ████   ████   ███    ██     ██████████████ ·      ·      ·
                                                          NOW ↑
                                                    (nothing to the right)

  The pool is not smaller. It is EMPTY, and it stays empty until rates fall
  back below the rates already on the book — which may take years.

That diagram is schematic and the distribution is illustrative, but the shape of the argument is exact and it is why refinance volume behaves the way it does. A rate rise does not reduce refinance demand proportionally. It eliminates it, because the entire book was just rewritten at rates below the new market.

This is not a hypothetical. The 2020–2021 period produced an extraordinary refinance wave as mortgage rates fell to historic lows, followed by one of the sharpest rate increases in the history of the modern mortgage market — the cycle this book's narrator describes as taking the market from under three percent to over seven in nineteen months. The refinance share of originations, which had dominated the market, fell away almost entirely. This is well documented in outline and you should look at the actual data rather than take anyone's summary of it, including this one; the Federal Reserve, the Federal Housing Finance Agency, and the Mortgage Bankers Association all publish relevant series. Do not quote a number you have not sourced. What you should carry from it is the shape: a wave, then a cliff, then a long flat.

And there is a second-order effect that the naive story misses entirely, and that Part VIII's readers need. All those households who refinanced at historically low rates now hold mortgages they will not voluntarily give up. Selling the house means giving up the rate. That reluctance — generally called the lock-in effect — suppressed the supply of existing homes for sale and therefore constrained purchase volume too. The rate cycle did not simply move business from one column to the other. It shrank both, in different ways and on different timetables.

What it does to a business

Application volume — the count of new applications taken in a period — is the leading indicator of everything. Closings lag applications by weeks; revenue lags closings by weeks more. A shop watching closed volume is watching a rear-view mirror. A shop watching applications sees the cliff about six weeks earlier, which is the difference between managing a decline and being surprised by one.

When applications collapse, three things happen in sequence, and they happen to somebody in every cycle:

First, capacity becomes cost. A shop that staffed up for boom volume — processors, underwriters, closers, and the systems and space to hold them — now carries that cost against a fraction of the revenue. Capacity is expensive to add and slow to remove, and the fixed cost per closed loan can double or worse. This is why the mortgage industry reduces headcount so sharply and so publicly in every downturn.

Second, pricing gets worse before it gets better. Excess capacity chasing scarce volume compresses margins across the industry. The loans that remain are fought over.

Third, originators whose pipeline came entirely from inbound volume discover they have no lead source. This is the one that ends careers. Chapter 7 laid out the arithmetic of the funnel: a closing requires a certain number of applications, an application requires a certain number of real conversations, and every conversation has to come from somewhere. In a refinance shop, "somewhere" is the marketing spend, the database, the call queue, or the servicer's retention list — and none of those belong to the loan officer. When the company stops buying leads, the loan officer's funnel does not shrink. It has no top.

A purchase-based originator in the same downturn has a smaller funnel and a harder market, but the top of their funnel is a set of relationships that are still there in the morning.

⚠️ Where Deals Die

The single-channel pipeline is not a pipeline. It is a subscription, and you are not the one paying for it.

The mechanism is slow and it feels like success the entire time it is happening. Volume is arriving. You are busy — genuinely, exhaustingly busy — and the work in front of you is real work that has to be done today. Every hour you might spend on referral development is an hour taken from files that are closing this month. So you don't spend it. Nobody does, in a boom. The originators who look like geniuses in a refinance wave are usually the ones who worked hardest at exactly the thing that was about to stop working.

Then the applications stop. And the specific, brutal detail is this: the moment you need referral partners is the moment every other originator in your market needs them too. Agents who were mildly interested in a new lender relationship last year are now taking four calls a week from loan officers who discovered them the same month you did. You are competing for scarce attention with a large number of people who have exactly your story, at the worst possible moment to be telling it.

What the disciplined originator does instead: treat purchase-side development as a fixed cost of doing business, paid weekly, in a boom especially. Not "when things slow down" — there is no such moment, because when things slow down it is already too late. §37.9 is about what that actually looks like on a calendar, and Chapter 38 builds the whole program.

37.9 Building a purchase business before you need one

The title is the argument. Here is why it is true and what to do about it.

Why it cannot be done later

Referral relationships have a lead time measured in quarters, and that lead time is a property of the relationship rather than of your effort. A real estate agent will not refer you their client until they have watched you do something hard, or until someone they trust has. Watching takes transactions. Transactions take months. The pipeline from "we met" to "they sent me a buyer" runs through at least one file that you did not get from them and did well anyway, at least one piece of genuine usefulness that had nothing in it for you, and enough repeated contact that you are the person they think of at 8:40 on a Wednesday.

None of that compresses. You cannot buy it, you cannot rush it, and you cannot start it in the month the phone stops — because everyone else is starting it that month too, and because an agent can tell the difference between a lender who has been present for two years and a lender who appeared with the downturn.

There is a second lead time, less discussed and equally binding: your own competence. Purchase files are operationally different, and a refinance-trained originator taking their first purchase files makes a specific and predictable set of mistakes. They under-budget the calendar because they have never had one imposed on them. They mis-size the rate lock against the contract's closing date rather than against their own optimism — the exact error the Linden Street file makes on day 12, which Chapters 20 and 30 dissect. They do not know what a financing contingency protects, or when an appraisal contingency expires, or why the listing agent needs to hear from them on day two. They communicate with the borrower and forget that in a purchase there are four other people who need to know the same thing. None of these are hard to learn. All of them take files to learn, and the files you learn them on are files somebody trusted you with.

What "building it" actually consists of

Chapter 38 is the full program and this is not a substitute for it. What belongs here is the counter-cyclical discipline, because the discipline is the part that fails.

Allocate the time as a fixed cost. A defined block, every week, that is spent on purchase-side development regardless of what is in the pipeline. It comes out of the week the way rent comes out of the month. The temptation to raid it when volume is high is exactly the failure mode; the whole value of the practice is that it runs when it feels unnecessary.

Be useful when there is nothing in it for you. Take the hard question from an agent whose file is with another lender. Explain why a self-employed borrower's tax returns produced a number their accountant did not — the Fulton Avenue problem, Chapters 11 and 32. Look at a down-payment assistance layering question at no charge, the way the Harlow Street file demands. Agents remember who helped them when there was no file attached, and it is a very short list.

Solve the files nobody wants. The most reliable way to become an agent's lender is to close something that was already declined somewhere else. This is also, not coincidentally, the best possible use of a slow market: you have the time to work a hard file properly, and hard files generate more loyalty per unit of work than easy ones.

Convert the database you built in the boom. Every refinance borrower is a household who will eventually buy, sell, or refer. That conversion works only if you stayed in contact, and only if your contact was worth receiving. The refinance shop that treated its borrowers as leads has a mailing list. The originator who treated them as clients has a referral base — the same names, completely different asset.

Run both channels, always, and let the mix move. This is the actual answer, and it reframes what "pivot" means. A pivot is not an emergency reallocation performed under duress. It is the ongoing management of a mix you already run: some purchase, some refinance, weighted to the market, with the infrastructure for both maintained at all times. An originator running 80/20 in a refinance wave can go to 20/80 in a quarter. An originator running 100/0 cannot go anywhere, because there is no relationship to weight toward.

What the mix costs in a boom

Be honest about the trade, because pretending it is free is why the advice gets ignored.

In a refinance wave, an hour spent on purchase development produces less immediate income than an hour spent on the refinance in front of you. Measurably less. That is not an illusion and there is no argument that makes it false. The purchase business is slower per file, more operationally demanding, more exposed to third parties who can kill a transaction you did nothing wrong on, and in a boom it is genuinely the worse use of the next hour.

The argument is not that the hour is worth more. It is that the hour is insurance, and insurance is always a bad trade until it isn't. The premium is real income, paid weekly, and the payout is that you still have a business in a market where a great many people do not. Anyone who tells you that purchase development is also the better use of your time in a boom is selling something. It is the better use of your career, which is a different claim and a stronger one.

37.10 The Linden Street refinance question

Fourteen months after closing, rates fall 150 basis points.

The borrowers call. They have seen the same headlines everyone has seen, and the servicer's retention department has already sent them something. The question is simple and the answer is not: should they refinance?

What we know

From the file as closed: a conventional thirty-year fixed of \$365,750 at 6.625%, P&I \$2,341.94**, taxes \$385.00, homeowners insurance \$130.00, borrower-paid monthly mortgage insurance \$176.78** at a 0.58% annual factor, for a total housing payment of **\$3,033.72. The appraisal supported the \$385,000 contract price. Mortgage insurance terminates automatically at payment 137, when the balance reaches 78% of the \$385,000 original value, and may be requested at payment 125, when it reaches 80%.

The call comes in mid-December of the following year. By the time an application is taken, a value is obtained, and a loan could actually close, they will be fourteen payments in.

The balance after fourteen payments

The canon gives the balance after twelve payments as \$361,757.88. Carry it two more months at the servicer's method — interest computed on the outstanding balance and rounded to the cent:

LINDEN STREET — PAYMENTS 13 AND 14                        [the Linden Street file]
  $365,750 original, 6.625%, 360 months, payment $2,341.94.

  payment  opening balance   interest    principal   closing balance
  ───────────────────────────────────────────────────────────────────────
     12         —               —            —        361,757.88   (canon)
     13     361,757.88       1,997.20      344.74     361,413.14
     14     361,413.14       1,995.30      346.64     361,066.50
  ───────────────────────────────────────────────────────────────────────
  interest 13 = 361,757.88 x 0.06625 / 12 = 1,997.2050 -> 1,997.20
  interest 14 = 361,413.14 x 0.06625 / 12 = 1,995.3017 -> 1,995.30

**The balance is \$361,066.50** — about \$361,100. In fourteen months of \$2,341.94 payments, this household has paid \$32,787.16 and reduced the debt by \$4,683.50. That is what the front of the curve looks like from inside it.

What the new loan looks like

A 150-basis-point market move puts a comparable file at roughly 5.125%. On \$361,066.50 over 360 months:

$$M = \$361{,}066.50 \times \frac{0.0042708\overline{3}}{1-(1.0042708\overline{3})^{-360}} = \$1{,}965.96$$

A P&I saving of \$2,341.94 − \$1,965.96 = \$375.98 a month.

Refinance closing costs, constructed for this file: origination at 1.00% of the loan amount \$3,610.67, appraisal \$650.00, credit report \$85.00, flood certification \$14.00, tax service \$78.00, lender's title policy at the reissue rate \$750.00, settlement fee \$595.00, recording \$212.00 — **\$5,994.67 total**. No owner's title policy, no survey, no pest inspection; those were purchase items. The escrow account is re-funded at the new lender and the old one refunded by the old servicer, which is a cash-flow event and not a cost.

The number every loan officer would quote:

$$\frac{\$5{,}994.67}{\$375.98} = 15.9 \text{ months}$$

Sixteen months to break even, then \$375.98 a month for twenty-eight and a half years. That is the pitch, and by now you know it is not the answer.

The amortization reset, explicitly

They have made fourteen payments of a three-hundred-sixty-payment schedule. A new thirty-year loan puts them at month 1 of 360 again, having already made fourteen payments. The final payment date moves from month 346 of today to month 360 of today — fourteen months later, and they will have made 374 mortgage payments on this house instead of 360.

Price the reset directly. Payment 15 of the existing loan would apply $\$361{,}066.50 \times 0.06625 \div 12 = \$1{,}993.39$ to interest and \$348.55 to principal. Payment 1 of the new loan applies $\$361{,}066.50 \times 0.05125 \div 12 = \$1{,}542.05$ to interest and \$423.91 to principal.

Here the reset does not slow the principal down — the rate improvement is large enough that the lower payment still retires more debt than the higher one did, \$423.91 against \$348.55. That is what a 150-basis-point move buys, and it is why this file is a genuine refinance candidate and the \$300,000 file in §37.5 was not. The reset is not automatically fatal. It is automatically uncounted, which is different, and you have to do the arithmetic to know which case you are in.

Where the reset does cost them is at the end. Compare total principal and interest:

(The existing loan's total is $346 \times \$2{,}341.94 = \$810{,}311.24$ less the \$3.07 by which the schedule overshoots zero at the final payment.) The refinance is still \$102,562.57 better in total P&I — a 150-basis-point move is a large move — but note what it cost to reset: refinancing into a term that ends on the same date, 346 months, gives a payment of \$1,999.76 and a total of $346 \times \$1{,}999.76 = \$691{,}916.96$. The fourteen months of reset cost \$15,828.64.

The shorter-term options

Since the term is a field somebody types into, type something better into it.

Look at the last column. For \$65.94 a month more than they are paying right now, this household retires the mortgage eight years and ten months earlier and pays \$232,416.97 less in principal and interest than their current schedule. That option does not appear on any refinance mailer ever printed, because it does not lower the payment. It is very likely the best financial decision available to them, and the only reason it is a hard sell is that the entire refinance industry has trained borrowers to evaluate this decision on one number.

(Fifteen- and twenty-year products typically price below the thirty-year. Holding all four terms at 5.125% is therefore conservative — the shorter terms would in practice look better still. Illustrative rate; verify current pricing.)

The mortgage insurance question

Now the part nobody computes.

The existing loan's mortgage insurance is not permanent and it is not on a clock that follows the borrower. It is on a clock that follows the loan. Under the Homeowners Protection Act, automatic termination occurs when the balance reaches 78% of the property's original value — for a purchase, the lesser of price or appraised value at consummation, which on this file is \$385,000. That threshold is \$300,300, reached at payment 137.

The household has made 14 payments. 123 payments of mortgage insurance remain, at \$176.78:

$$123 \times \$176.78 = \$21{,}743.94$$

Refinance and that schedule is gone, replaced by a new one. For a refinance, "original value" is the value at consummation of the refinance — so a new appraisal is not a formality, it is the input that sets the entire mortgage insurance schedule for the next decade.

And they will have mortgage insurance. At a \$385,000 value and a \$361,066.50 balance:

$$\frac{\$361{,}066.50}{\$385{,}000} = 93.78\%$$

which is nowhere near the 80% at which mortgage insurance can be avoided. The refinance restarts a mortgage insurance obligation on a loan that is already fourteen months into paying one off.

🧮 Run the Numbers

What the mortgage insurance reset actually does on this file — and what it would do on a different one.

The structure that works. Financing all \$5,994.67 of costs gives a loan of \$367,061.17, which is 95.34% of \$385,000 — over the 95% line, into a worse mortgage insurance band, a worse price adjustment, and possibly outside the standard limited cash-out program entirely. Ninety-five percent of \$385,000 is exactly **\$365,750.00 — the original loan amount. So finance \$4,683.50** of the costs, bring **\$1,311.17** to the table, and land the new loan at \$365,750.00 at 95.00% LTV.

At that loan amount and the same 0.58% factor, monthly mortgage insurance is $\$365{,}750 \times 0.0058 \div 12 = \$176.78$ — identical to today's. P&I at 5.125% on \$365,750 over 360 months is **\$1,991.46, so PITI plus MI is $\$1{,}991.46 + \$385.00 + \$130.00 + \$176.78 = \$2{,}683.24$, a saving of **\$350.48 a month.

When does the new mortgage insurance end? The 78% threshold is still \$300,300, but now measured against the new loan's schedule. Solving the amortization for \$365,750 at 5.125%, the balance reaches \$300,300 at payment 118.

	Keep	Refinance at 95.00%
MI payments remaining	123	118
Monthly MI	\$176.78 | \$176.78
Total MI still to pay	\$21,743.94** | **\$20,860.04
MI ends, months from today	123	118

The reset is nearly free here — and that is entirely because of two facts that will not hold on the next file. The rate drop is large, so principal amortizes faster and the new loan reaches 78% sooner. And they are only fourteen payments into the old schedule, so there was very little progress to give back.

Change the second fact and watch it invert. Take the identical structure and put the household at payment 120 instead of payment 14 — seventeen payments from never paying mortgage insurance again:

$$\text{keep: } 17 \times \$176.78 = \$3{,}005.26$$ $$\text{refinance: } 113 \times \$174.52 = \$19{,}720.76$$

They would have bought \$16,715.50 of mortgage insurance they were seventeen months from escaping — a cost that appears nowhere in any payment comparison and that by itself exceeds every closing cost in the transaction.

The rule: the cost of the mortgage insurance reset is proportional to how far into the schedule the borrower already is. Ask the question every time. On this file it is small. On the file after it, it may be the whole answer. (MI factor of 0.58% held constant for comparison and labeled illustrative; factors are LTV-banded and score-banded and are revised — verify the current card.)

The value is the whole risk

Everything above assumes the property appraises at \$385,000. Test that assumption, because it is load-bearing.

For the balance-only loan to sit at or below 95% LTV, the value must be at least $\$361{,}066.50 \div 0.95 = \$380{,}070.00$. A value below that pushes the loan into the 95.01–97% band, where the mortgage insurance factor is higher, the loan-level price adjustment is worse, and program eligibility narrows.

Suppose the appraisal returns \$375,000. Now three things move at once:

• LTV becomes $\$361{,}066.50 \div \$375{,}000 = 96.28\%$.
• The mortgage insurance factor moves to the higher band. At an illustrative 0.86%, monthly MI becomes $\$361{,}066.50 \times 0.0086 \div 12 = \$258.76$ — \$81.98 more per month than they pay now, cutting the \$375.98 P&I saving to \$294.00 and pushing the naive break-even from 15.9 months to 20.4.
• The 78% threshold falls to \$292,500, which the new loan reaches at payment 124. Mortgage insurance now costs $124 \times \$258.76 = \$32{,}086.24$ against the \$21,743.94 they owe on the current schedule — \$10,342.30 worse.

That is the Cypress Court lesson arriving in refinance form. A short appraisal in a purchase creates a cash gap somebody can negotiate. A short appraisal in a refinance creates a mortgage insurance problem nobody negotiates, and it is worth more than every closing cost on the Loan Estimate.

The three break-evens

One last thing, because the household will ask "when do we break even" and the honest answer is another question: break even on what?

Take the recommended structure — loan \$365,750, \$1,311.17 cash at the table, P&I \$1,991.46, MI \$176.78, total housing \$2,683.24 against today's \$3,033.72.

Out-of-pocket break-even. \$1,311.17 ÷ \$350.48 = 3.7 months. Technically true and thoroughly misleading, because \$4,683.50 of costs went into the balance rather than being paid.

Net-position break-even — cash paid plus balance owed, both loans, same date. At twelve months the refinance is \$825.29 behind**; at twenty-four months it is **\$4,326.39 ahead; the crossover is at month 14.

Balance crossover — the month the household actually owes less than they would have. Running both amortizations forward, the new loan's balance first falls below the old loan's balance at payment 62, a little over five years, where the two sit at \$335,343.27 and \$335,380.71.

Three defensible numbers — 3.7 months, 14 months, 62 months — for one transaction, and they are all correct, because they measure different things. If the question is "when will I have more money," the answer is month 14. If the question is "when will I owe less," the answer is month 62. Say which one you are answering.

🗂️ The Loan File

Chapter 37 contribution: the refinance decision, worked and recommended.

The recommendation: yes — refinance, but not into a new thirty-year loan by default, and not with the costs rolled in until somebody has checked the 95% line.

A 150-basis-point improvement on a 95%-LTV loan fourteen months old is not a close call. What is a close call is the structure, and the structure is where all the money is.

Do this, in this order.

1. Get a value before you get anyone's hopes up. Order the appraisal or check whether the file qualifies for a value acceptance. Tell the household in advance that the analysis needs \$380,070 or better for the loan to sit at or below 95%, and what happens if it does not. Everything else in this recommendation is conditional on that number.
2. Do not finance the full \$5,994.67.** At \$367,061.17 the loan is 95.34% of \$385,000 — over the line. Finance \$4,683.50**, bring **\$1,311.17, and land at exactly \$365,750.00 / 95.00%**. Reserves after the day-44 furniture payoff were \$7,423.66, so fourteen months of rebuilding makes \$1,311.17 a manageable number, and it should be confirmed before it is assumed.
3. Present three terms, not one. Thirty years at \$1,991.46. Term-matched at 346 months. Twenty years at \$2,407.88 — \$65.94 a month more than they pay today, retiring the loan eight years and ten months earlier and saving \$232,416.97 in principal and interest. Let the household choose. Do not choose for them by showing only the lowest payment.
4. Compute the mortgage insurance both ways and put it in writing. On this file the restarted schedule ends at payment 118 versus 123 today, so it is nearly a wash — but that conclusion belongs in the file with its arithmetic, because it is the conclusion that will be wrong on the next file.
5. Write the benefit down (Figure 37.1's discipline, not Figure 37.1's worksheet), dated, at application.
6. Tell them the money does not move at signing. This is a refinance of their principal residence with a new lender; the right of rescission applies. Chapter 23 has the mechanics.

What this settles: that the refinance is worth doing, what it is worth, at what structure, and what the household gives up by taking the lowest payment on offer.

What it does not settle:

• The value. Not one number in this analysis survives an appraisal below \$380,070 unchanged.
• The qualification. Income was held at \$10,500.00 and debts at \$1,446.00, which is conservative — one auto loan has five payments remaining and may now be excludable, which would put back-end debt-to-income near 35.24% rather than 39.33%. Both are comfortable. Neither is verified. Nothing is approved until it is documented (Chapters 10, 11, 12).
• The credit. Fourteen months of clean payment history may have improved the 706 representative score, which would improve the price. It has not been pulled.
• The horizon. The entire net-position analysis turns on how long this household keeps this loan, and nobody has asked them.

What would change the recommendation:

• an appraisal below **\$380,070**, and decisively below about \$375,000
• a plan to sell or relocate within roughly two years
• a credit event in the last fourteen months that reprices the file
• closing costs materially above \$5,994.67, or a "no-cost" offer whose rate has not been checked against par (Chapter 29)
• a request for cash, which makes this a different transaction with different limits and different pricing (§37.6, Chapter 35)

Open questions carried forward:

• Q37.1. How long do these borrowers intend to keep this house and this loan? (Nobody has asked. It is the most important unanswered question in the file.)
• Q37.2. Does the property still support \$385,000, and what is the plan if it supports \$375,000?

Your task. In Appendix C's workbook, complete the refinance analysis at three horizons — 24 months, 62 months, and to payoff — using the net position test from §37.2. Then write the one sentence you would put in the file as the net tangible benefit statement, and a second sentence naming the fact that would make it false.

Conclusion

Purchase and refinance origination share a license and almost nothing else. One arrives attached to a contract, a deadline, and a person who vouched for you; the other arrives attached to nothing, competes against every advertised rate in the country, and disappears entirely when the spread that created it closes.

The refinance decision itself is a total-cost question over a horizon, and the standard formula — costs divided by payment saving — gets it wrong three times in the same direction. It ignores that a new loan returns the borrower to the front of the amortization curve. It treats a payment reduction achieved by extending the term as a saving when it is the opposite. And it prices financed closing costs at face value when they are borrowed at the note rate for thirty years and cost more than twice what they appear to. On one constructed file those three errors turned a "twenty-month break-even" into a thirty-two-month one, and turned a full point of rate improvement into \$38,319.12 of additional cost — while the term-matched version of the identical refinance saved \$38,145.60.

Net tangible benefit requirements exist because that gap is exploitable and was exploited. Serial refinancing strips equity through financed costs, resets amortization, resets the mortgage insurance clock, and does all three invisibly to anyone comparing payments. The prevention is not complicated: ask when the current loan closed, compute the recoupment honestly, write the benefit down, and decline the ones that do not clear.

On the Linden Street file, a 150-basis-point move fourteen months after closing is a genuine opportunity — worth \$350.48 a month at the right structure, and worth \$232,416.97 over the term if the household takes the twenty-year instead of the payment. But the structure matters more than the rate: financing all the costs would push the loan to 95.34% and into a worse mortgage insurance band, and the whole analysis dies on an appraisal below \$380,070.

And behind the arithmetic is a business fact. Refinance volume does not decline when rates rise; it is consumed and then it is gone, and the originators who leave the business in every cycle are the ones whose entire funnel belonged to somebody else. A purchase business takes quarters to build and can only be built before you need it, because the month you need it is the month everyone in your market is competing for the same handful of agents with the same story.

Next: Chapter 38 takes the thing this chapter says you must build and builds it — where referral relationships actually come from, what a partner is worth in dollars, and how to run a book of business that does not depend on which way rates moved this quarter.

Key Terms

Purchase market — origination of loans to finance the acquisition of real property; demand is driven by household formation, inventory, and affordability, and changes gradually. (Ch.37)

Refinance market — origination of loans that replace existing mortgage debt on property the borrower already owns; demand is driven by the spread between borrowers' existing note rates and today's rate, and changes in steps. (Ch.37)

Rate-and-term refinance — a refinance that changes the rate, the term, or both, paying off the existing lien and closing costs with only incidental cash to the borrower; called a limited cash-out refinance by Fannie Mae and a no cash-out refinance by Freddie Mac. (Ch.37)

Cash-out refinance — a refinance for more than the payoff of the existing lien and costs, with the difference delivered to the borrower; defined by what the proceeds retire, not by whether a check is written. (Ch.37)

Streamline refinance — a program-specific reduced-documentation refinance within a government program, in which appraisal, income, or credit documentation is waived because the investor's risk is not increased; carries mandatory benefit tests. (Ch.37)

Break-even (refinance) — the point at which a refinance leaves the borrower better off; correctly computed as the horizon at which cash paid plus balance owed is lower under the new loan than under the old one, not as closing costs divided by payment reduction. (Ch.37)

Net tangible benefit — the requirement that a refinance leave the borrower measurably and demonstrably better off, imposed variously by program rules, investor overlays, and state statutes. (Ch.37)

Churning — refinancing a borrower repeatedly, at intervals too short for each transaction's costs to be recovered, principally for the benefit of the party originating the loans. (Ch.37)

Serial refinancing — the pattern of repeated refinances on one borrower whose cumulative effect is compounding financed costs, repeated amortization resets, a repeatedly restarted mortgage insurance clock, and progressive loss of equity. (Ch.37)

Home equity extraction — converting accumulated home equity into cash through a cash-out refinance or a junior lien, exchanging an illiquid asset for liquidity and a monthly obligation. (Ch.37)

Application volume — the count of new loan applications taken in a period; the leading indicator of a shop's future closings and the first number to move when rates change. (Ch.37)

Pivot — the deliberate reweighting of a loan officer's time, marketing, and partner development between the purchase and refinance channels; sustainable only where both channels already exist. (Ch.37)

Spaced Review

1. A borrower is offered a refinance that lowers the payment by \$250 a month at a cost of \$5,500, financed. Their principal reduction falls by \$95 a month and the term goes from 26 years remaining to 30. Without computing anything precisely, state which of the three errors in §37.5 this transaction commits and which single number you would ask for next.

2. (Chapter 13) Chapter 13 chose conventional over FHA for the Linden Street borrowers partly because the conventional loan's mortgage insurance terminates at payment 137 while FHA's would never terminate at that loan-to-value. Explain how §37.10 vindicates that decision, and what would have happened to this refinance analysis if the original loan had been the FHA option instead.

3. (Chapter 30) The Linden Street lock was taken on day 12 for thirty days against a contract naming a day-45 closing — three days short before anyone did anything wrong. A refinance has no contract and therefore no externally imposed closing date. Does that make the lock decision easier or harder? Name the specific failure mode that replaces the missed contract date.

4. (Chapters 13 and 30) A borrower asks you to compare a refinance at 5.125% with no points against 4.875% with one point, on a \$361,066.50 balance. Which chapter's method computes the break-even on the points, and why is that break-even a different calculation from the refinance break-even in §37.5? Name both denominators.

5. A colleague says: "I did four hundred units last year and I don't need real estate agents." Using §37.1 and §37.8, write the two-sentence reply that is neither smug nor false, and name the number they should be watching that they almost certainly are not.