Skip to the content

The file

File opened · Sep 2, 2026

One file, three real Loan Estimates, and the rule behind every number.

A pre-approval is not an approval, and the rate printed on it is not locked. The only comparable document in this country is the Loan Estimate: it is free, a lender owes you one within three business days, and nobody may charge you for it except the credit report. Everybody else builds on advertised rates. This page builds on that document.

01Block

Her situation

Who she is, what she is looking at, and how much house each rulebook says she can buy

The same person, the same income, the same house — and rulebooks that disagree by $173,000 on how much she can buy. The pre-approval letter she is carrying is for the one that reaches the least.

Who she is

María

Age

Forty is the median age of a first-time buyer in the United States, and an all-time high. The person shopping for a mortgage is not a twenty-something.

40 years
Where

Tarrant County · FIPS 48439

Fort Worth, TX
Household

people, two earners plus a rented room

4
Credit score

Classic FICO · A 691 sits inside the 680–699 band of both the loan-level price adjustment grid and the mortgage insurance grid, nine points below the next band. Nine points of score are worth more than several tenths of a rate, and no comparison site tells her that.

691
Saved

The 3.5% down payment comes out of this. What is left stays in the account — and post-closing reserves are one of the compensating factors that raise her FHA cap. That is block 4.

$43,000
First-time buyer
Yes
Plans to stay
7 years

The house she is looking at

$434,100

Property
Single-family, 1 unit, primary residence
County conforming limit

The county FIPS is a price field, not a courtesy: the conforming limit, the FHA limit, the property tax, the insurance and the title premium all come out of it.

$832,750
County class
baseline — not a high-cost county

Source

NAR, Existing-Home Sales for July 2026 (national median $434,100, up 2.0% year over year from $425,700). NAR’s own statistics page shows $431,400 for that same month.

What she is carrying

The pre-approval letter

Not verified

It says

6.5%

FHA

What the public daily benchmark says for her segment

6.766%

OBMMIC30YF

A pre-approval is not an approval, its rate is not locked, and it creates no obligation on anybody. It is not a market figure either: it is what one lender told her.

Assumption

This is the general 30-year conforming index, not her LTV × FICO segment — that series has no observation we could verify. For a high LTV with a mid score, the general index understates: the observed gap between the general index and the LTV>80 / FICO<680 cross was 16.5 basis points.

Her number, if she buys this house with that pre-approval

Market convention
Principal and interest
$2,769.10
Mortgage insurance (MIP)
$191.08
Property tax
$450.38
Homeowner’s insurance
$187.58
Monthly payment
$3,598.14

$418,907

Base loan amount

$434,100 − $15,194

+ $7,331

Upfront MIP added to the balance on day one

$426,237

What she owes the day she signs

96.50%

LTV

97% rounded for the grids

Income $8,025.00/mo · other debts $0

How much house each rulebook allows

  • FHA, manual underwriting, no documented compensating factors

    Housing ratio / total debt cap: 31% / 43%

    $292,000Does not reach it

    HUD 4000.1 II.A.5.d.viii, manual underwriting with 0 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.

  • FHA, manual underwriting, two documented compensating factors

    Housing ratio / total debt cap: 40% / 50%

    $384,000Does not reach it

    HUD 4000.1 II.A.5.d.viii, manual underwriting with 2 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.

  • Conventional 97%, through Desktop Underwriter

    Housing ratio / total debt cap: — / 50%

    $465,000Reaches it

    Fannie B3-6-02: Desktop Underwriter casefiles reach 50%. It is the highest ceiling in the catalog.

This house costs$434,100

What stops her is not the total debt ratio — it is the housing ratio. FHA carries two caps and the one that breaks first is the front-end. Conventional through DU carries no housing ratio at all: only the 50% on the back. That single difference is the whole gap.

Assumption

This is the manual matrix, which is the only public rulebook. With an Accept from the TOTAL Scorecard the AUS decides instead, and that verdict is not published anywhere: there is no public sandbox for DU, LPA, GUS or TOTAL. So the honest reading is a floor, not a ceiling — and the wall we cannot see past is exactly where the industry keeps its advantage.

What is not on the table, and why

Nothing here is hidden or paginated away. When our engine knows the rule it cites it; when it does not, it says so instead of pretending.

  • Jumbo

    Below the conforming limit: she does not need a jumbo, and conforming loans have public rules and comparable pricing.

    FHFA Conforming Loan Limit FAQs 2026, question 1

  • VAOur engine does not decide this

    Nobody in this household has entitlement on file. VA eligibility comes from a Certificate of Eligibility, not from anything a calculator can compute — so this is a document question, not an engine verdict, and we mark it as one.

  • USDAOur engine does not decide this

    Her household income clears the USDA limit, so the program is not out on income. What decides it is the rural map, and we do not have it yet. Fort Worth is almost certainly not eligible, but "almost certainly" is not a rule we are willing to print as one.

02Block

The packet

The packet that makes a lender owe her a Loan Estimate

Six pieces of information turn a conversation into an "application" under Regulation Z. Once a lender has all six, it owes her a Loan Estimate within three business days, it may not charge her anything except the credit report, and it may not demand verification documents first. We prepare five of them. The sixth is hers to hand over — to whom she chooses, when she chooses.

The six pieces

  1. 1
    NameIn the packet
  2. 2
    Gross household income (annual)In the packet
  3. 3
    Property addressIn the packet
  4. 4
    Estimated property valueIn the packet
  5. 5
    Loan amount requestedIn the packet
  6. 6
    Social Security numberNot in the packet

    Without the Social Security number there is no "application" under 12 CFR 1026.2(a)(3)(ii): no Loan Estimate obligation, no hard credit pull, and no trigger lead. That is a legal decision dressed up as a user-experience one, and it is the reason nobody starts calling her the day after she uses this page.

On trigger leads

The law shut off the sale of trigger leads by the bureaus on Mar 4, 2026 — but left an exception where the consumer consents. We do not ask for that consent.

The fear that stops people from asking twice

Almost nobody collects more than one quote, and the reason they give is always the same: they are afraid several lenders pulling their credit will wreck their score. The regulator answers that question in one sentence, and it is worth reading in the original:

“Within a 45-day window, multiple credit checks from mortgage lenders are recorded on your credit report as a single inquiry.”

— Consumer Financial Protection Bureau, Ask CFPB

So the whole search fits inside one window. Ask three to five lenders inside 45 days and the credit report sees one inquiry, not five. Removing that fear is the product.

45 days

One window, one inquiry

3 business days

A lender owes the Loan Estimate within

Nothing except the credit report

What it may charge you for it

Source

12 CFR 1026.2(a)(3)(ii) (the six pieces of information); 12 CFR 1026.19(e)(1)(iii)(A) and (e)(2) (the three-business-day deadline and the ban on charging for it); CFPB, Ask CFPB, “What happens when a mortgage lender checks my credit?” (the 45-day window)

What to ask each lender, word for word

Seven asks. Six of them are things the regulation already gives her and almost nobody claims; the seventh is ours, and it is the one that makes three documents comparable at all.

  1. 1

    "Here are the six items of the application. Please send me the Loan Estimate."

    The rule

    12 CFR 1026.19(e)(1)(iii)(A): deliver or place in the mail no later than the third business day after receiving the application. 12 CFR 1026.2(a)(3)(ii) defines the six items.

  2. 2

    "Do not charge me anything before I have the Loan Estimate and tell you I want to proceed."

    The rule

    12 CFR 1026.19(e)(2)(i): no fee before the consumer receives the Loan Estimate and indicates intent to proceed. The single exception is a bona fide and reasonable credit report fee.

  3. 3

    "Do not ask me for verification documents before you give me the Loan Estimate."

    The rule

    12 CFR 1026.19(e)(2)(iii): the creditor may not require verification documents as a condition of issuing the Loan Estimate.

  4. 4

    "Quote me without discount points. If you want to sell me points, send both versions."

    The rule

    12 CFR 1026.37(f)(1)(i): points paid to reduce the rate must be itemized separately as a percentage and as a dollar amount, under the literal label "____% of Loan Amount (Points)" — and if no points are charged the field must be left blank. Comparing a rate with two points against a rate with none is the exact lie this product exists to break.

  5. 5

    "Tell me both expiry times: the rate lock and the estimated closing costs."

    The rule

    12 CFR 1026.37(a)(13)(ii): the form must state whether the rate is locked and, if so, the date and time with time zone — and separately, the date and time the estimated closing costs expire. They are two different clocks and they almost never fall on the same day.

  6. 6

    "Send me the written list of providers for the services I am allowed to shop for."

    The rule

    12 CFR 1026.19(e)(1)(vi)(C): when the consumer may shop for a settlement service, the creditor must provide a written list of providers. Block C is the one she can move.

  7. 7

    "Use the same estimated closing date on all of them: October 15."

    The rule

    This one is not in the regulation — it is ours, and it is the one that makes the papers comparable. Prepaid interest in block F runs from closing to the end of the month, so two Loan Estimates with different closing dates differ in F for a reason that has nothing to do with the lender. Fix the date and the difference that survives is real.

Copy it and read it out on the phone, or print it and take it in.

The packet, ready to hand over

MORTGAGE APPLICATION — ITEMS PROVIDED
───────────────────────────────────────────────────────
Name                             María
Gross household income (annual)  $96,300
Property address                 Fort Worth, TX — Tarrant County
Estimated property value         $434,100
Loan amount requested            $418,907
Social Security number           WITHHELD — provided by the borrower directly, at her choice

CONTEXT FOR THE QUOTE
───────────────────────────────────────────────────────
Program                          FHA
Term                             30 years, fixed
Down payment                     $15,194 (3.5%)
Occupancy                        Primary residence, 1 unit
Estimated closing date           Oct 15, 2026
Expected time in the home        7 years

Please send the Loan Estimate within three business days (12 CFR 1026.19(e)(1)(iii)(A)). Quote without discount points; if you offer points, send both versions. Include both expiry times: rate lock and estimated closing costs. Please include the written list of providers for services I may shop for.

Sent to 3–5 lenders inside the 45-day window, all with the same estimated closing date of Oct 15, 2026.

03Block

Three Loan Estimates

Three Loan Estimates for the same loan, normalized and put side by side

Same house, same price, same down payment, same term, same closing date. Everything that differs below differs because a lender chose to make it differ — or because a rule fixed it and no lender can touch it. Telling those two apart is the whole job.

Why these carry no lender name

A Loan Estimate is a firm price from one named lender to one named person. In her own file those three papers carry names, because they are hers. On a public page they do not: lender names are public, lender pricing is not, and printing "this lender quoted her 6.5%" is the one thing this product refuses to do.

Demonstration figures

Not verified

These three are demonstration documents: the fee lines were typed in so the mechanics are visible. Everything around them is real — the upfront MIP comes from the 1.75% rule, the lender’s title premium from the rate the Texas Department of Insurance promulgates, the appraisal from the VA fee table, the escrow deposit from the aggregate-accounting rule, and the prepaid interest from each document’s own rate.

Same loan, same closing date · Oct 15, 2026

Three Loan Estimates for the same loan, normalized and put side by side

LineLoan Estimate 3Received Sep 8, 2026Loan Estimate 1Received Sep 4, 2026Loan Estimate 2Received Sep 5, 2026Spread
Rate lock expiresOct 8, 2026Oct 4, 2026Not locked—
Estimated costs expireSep 22, 2026Sep 18, 2026Sep 19, 2026—
A · Origination chargesThe lender’s own fees, discount points included. This is where the lender lives.
Underwriting fee$1,395—$1,400$5
1.000% of Loan Amount (Points)$4,262———
Origination fee—$1,095$995$100
A · Origination charges$5,657$1,095$2,395$4,562
B · Services you cannot shop forAppraisal, credit report, flood certification — and in an FHA loan the upfront mortgage insurance premium, which is 1.75% of the base loan by rule and identical on all three.
Appraisal fee$775$775$775—
Credit report fee$75$65$85$20
Flood certification$20$19$25$6
FHA upfront mortgage insurance premium$7,331$7,331$7,331fixed
B · Services you cannot shop for$8,201$8,190$8,216$26
C · Services you can shop forIn Texas the lender’s title premium is promulgated by the state, so it is the same on all three. What actually moves here is the settlement fee, the endorsements and the survey.
Lender's title insurance policy$2,392$2,392$2,392fixed
Settlement / escrow fee$525$395$650$255
Title endorsements$100$75$150$75
Survey$625$525$525$100
C · Services you can shop for$3,642$3,387$3,717$330
Total loan costs (A + B + C)These three blocks — and only these three — are what the federal mortgage data set calls total_loan_costs, and what every article about closing costs is really measuring.$17,500$12,671$14,327$4,828
Other costs — outside total_loan_costs
E · Taxes and government feesCounty recording. Texas is cited as a state with no transfer tax, and we have not verified that in a primary state source.$132$132$132—
Prepaid interest · 17 days$1,290$1,365$1,388$97
Homeowner’s insurance · 12 months$2,251$2,251$2,251fixed
F · PrepaidsPrepaid interest from closing to the end of the month, plus twelve months of homeowner’s insurance. Prepaid interest is the only line here that changes with the rate.$3,541$3,616$3,639$97
G · Initial escrow deposit at closingCalculated, not guessed: the lowest projected month-end balance must be zero plus the allowed cushion, and the cushion cannot exceed one sixth of the annual disbursements.$5,405$5,405$5,405fixed
Cash she brings to the tableThe upfront MIP is inside block B but it is NOT cash: it is financed on top of the loan. Counting it as cash at closing is the error that inflates every calculator on the internet.$34,440$29,686$31,365$4,754
What is left in her account afterwardsPost-closing reserves are a documented compensating factor under FHA manual underwriting. Spending them on a discount point can cost her the factor that lets her qualify at all.$8,560$13,314$11,635$4,754
What she pays every month
Note rate6.5%6.875%6.99%—
Discount points1 pts———
Principal and interest$2,694.11$2,800.08$2,832.91$139
Mortgage insuranceIdentical on all three: FHA sets the premium by rule, and it does not depend on the lender or on her score.$191.03$191.09$191.11fixed
Monthly payment with taxes and insurance$3,523.10$3,629.13$3,661.98$139
APR7.654%7.916%8.078%—
Total cost over her horizonEverything she pays and puts in over the years she says she will stay: payments, mortgage insurance for as long as it lasts, taxes, insurance, closing costs and the down payment.$328,010$332,126$336,553$8,543

Order of this tableOrdered by total cost over the 7-year horizon she declared, ascending. That is the criterion HUD blessed in writing as a neutral display in 1996 ("APR calculated to include all charges and to account for the expected tenure of the buyer"). No document is highlighted, promoted, repeated or hidden, and none of them pays us anything.

This is our own comparison, not a copy of the official form. Reproducing the headings, content and format of the H-24 Loan Estimate in a written estimate given before the real one is prohibited by 12 CFR 1026.19(e)(2)(ii), and we do not do it.

What the three papers actually say

Where the money is

Total loan costs differ by $4,828 across the three. But block B barely moves, because the upfront MIP fills it and that is fixed by rule; and the title premium inside block C is fixed by the state of Texas. Strip the discount point out and the lender-controlled part of block A still ranges 2.19× between the cheapest and the dearest — on the same loan, for the same person, in the same week.

That is the same phenomenon the federal data shows at scale: in a controlled 2025 cohort of 531 conventional purchases in one county — same program, same term, same occupancy, LTV between 79 and 81, DTI between 36 and 43, loan between $300,000 and $500,000, that is to say borrowers indistinguishable to an underwriter — the 90th percentile paid 3.71 times what the 10th percentile paid.

Observed

The one with the lowest rate is selling her a point

The cheapest rate on the table costs $4,262 up front, and it buys $105.97 a month against the cheapest rate that comes without a point. It pays for itself in 41 months and she says she is staying 84 months, so on those numbers it earns its keep — but it is paid out of the reserves that FHA counts as a compensating factor, and that trade is not on any Loan Estimate.

Break-even against her horizon

41 / 84

Cash now against cost later

And the paper that wins over her horizon is not the one that asks for the least cash. The gap between the cheapest cash to close and the cheapest total cost is $4,754, pointing in opposite directions. Which one matters depends entirely on how long she stays — which is why that is the question this product asks first and everyone else asks last.

Cash she brings to the table

$29,686 lowest$34,440 highest

Total cost over her horizon

$328,010 lowest$336,553 highest

APR

The legal APR leaves out exactly the most variable costs — title, appraisal, credit report, survey — so two loans with the same APR can cost thousands of dollars apart at the table. And the law only requires it to be accurate within one eighth of a point: below that, a difference is not information. These three are further apart than one eighth of a point, so the difference is real.

What the data set measures

Inside total_loan_costs

  • A. Origination Charges (lender fees, discount points included)
  • B. Services You Cannot Shop For (appraisal, credit report, flood certification, and FHA’s upfront MIP when the borrower pays it)
  • C. Services You Can Shop For (settlement agent, lender’s title insurance)

Outside it — and outside every comparison site

  • E. Taxes and Other Government Fees (recording, transfer taxes)
  • F. Prepaids (prepaid interest, premiums, taxes)
  • G. Initial Escrow Payment at Closing
  • H. Other (owner’s title policy)

Source

CFPB, “2023 Mortgage Market Activity and Trends” (December 2024), p. 24 and notes 21 and 23 · Our own computation on primary HMDA 2025 microdata (FFIEC/CFPB), county 48201, pulled through /v2/data-browser-api/view/csv. Figures reproduced and corrected by the report’s verifier.

04Block

The documents

Every document, with the reason it exists and what it costs when it is missing

Lists of required documents are everywhere and they are useless, because they never say what the paper decides. That column is half the product. Every number in the red line below was recomputed by running the engine again without that document.

Pay stubs and W-2s for both earners

Field 11 of the 20 that move the price

What it decidesTwo earners, one household. Both incomes count for the debt-to-income ratio, and the documentation each one needs depends on how it is earned — W-2, 1099, self-employment, rent, pension, military. The type does not change how much it is worth; it changes which paper proves it.

What it costs if it is missing

With only one of the two documented, the maximum price the FHA manual matrix supports falls from $292,000 to $134,000.

First-time buyer declaration

Field 15 of the 20 that move the price

What it decidesBetween 95.01% and 97% loan-to-value, conventional requires at least one first-time homebuyer on the note. And the 97% conventional is the only program on this page that reaches the house she is looking at.

What it costs if it is missing

The conventional at 97% is excluded outright — and with it goes the only rulebook that reaches this house.Fannie Selling Guide B2-1.3-01 (5-nov-2025)

Household income measured against the area median income

Field 14 of the 20 that move the price

What it decidesA first-time buyer at or below 100% of the area median income (120% in a high-cost area) gets every loan-level price adjustment waived. Household income and county FIPS are price fields, not courtesy fields.

What it costs if it is missing

Her conventional carries 1.125 points of loan-level price adjustment. Documented, the waiver would take that to zero: $4,737.

Bank statements showing reserves after closing

Field 18 of the 20 that move the price

What it decidesDocumented post-closing reserves are a compensating factor in FHA manual underwriting, and compensating factors are what move the caps from 31/43 to 40/50.

What it costs if it is missing

Without two documented factors her FHA cap stays at 31/43 instead of 40/50, and the maximum price falls from $384,000 to $292,000.

Credit score for each borrower separately

Field 10 of the 20 that move the price

What it decidesThe rule differs by program and almost nobody knows it: conventional uses the average of the borrowers’ median scores, FHA uses the lowest among borrowers. Two people at 640 and 691 give 665 conventional and 640 FHA — two different loans out of the same couple.

What it costs if it is missing

If the second borrower comes in at 640: FHA barely moves, because its premium does not depend on score at all ($0.03 a month). The conventional moves three ways at once — $115.80 a month in premium, $526 in price adjustments, and the rate anchor itself drops out of its band from 6.766% to 6.931%, which alone is $3,895 of extra interest inside her horizon. The three together: $14,149.

The appraisal

Field 4 of the 20 that move the price

What it decidesOn a purchase, loan-to-value runs on the LOWER of contract price and appraised value. A low appraisal does not lower the price she agreed to pay: it raises the ratio the lender is underwriting.

What it costs if it is missing

It is not that it goes missing — it is that it can come in low. An appraisal at $428,000 pushes her ratio to 97.88%, above the 96.5% ceiling for an FHA purchase. She would have to bring $5,887 more in cash or renegotiate the price.

Source of the down payment funds

Field 5 of the 20 that move the price

What it decidesNot just how much, but where it came from and when it landed. And here the amount matters to the dollar, because the mortgage insurance cliff is not at "10% down" — it is at 90.00% loan-to-value, measured on the base loan excluding the upfront premium.

What it costs if it is missing

With all $43,000 of her savings the ratio lands at 90.09% and the premium runs the full 360 months. With $410 more it lands at exactly 90.00% and the premium stops at 132 — $410 at the table against $18,771 in premium.

Documented history of the rented room

Field 11 of the 20 that move the price

What it decidesThe room is real income and it is paying the mortgage every month. Whether an underwriter may count it is a separate question from whether it exists, and it depends on documented history — a rule we have not been able to verify in a primary source, so we count it as zero.

What it costs if it is missing

Counted at zero today. Every $100 a month that can be documented is worth about $5,000 of maximum purchase price.

Notice what is not on this list: the Social Security number. It is not a document she owes us, and we never ask for it.

Every figure in the red lines above was produced by running the engine again with that input removed, on the same 7-year horizon as the rest of the file.

05Block

Her income

Her income, and what each source is actually worth for qualifying

Three sources of money come into this household every month, and an underwriter counts them three different ways. This is the question the self-employed ask first and get answered last.

$96,300 a year · $8,025.00/mo

The three sources

Projected
SourceWhat it is worth for qualifying todayWhat proves it
First earnerCounted in fullPay stubs and W-2s, or the equivalent for the way each one earns
Second earnerCounted in fullPay stubs and W-2s, or the equivalent for the way each one earns
The rented roomZeroDocumented rental history

Source

The household total is anchored to the median income of a Hispanic homebuyer, which is a segment anchor and not her paycheck. How the total splits between the two earners is not in the file, and the split does not change the ratio — it changes which paper is needed.

Not verified

Which is the whole point for anyone self-employed: the dollar value of the income does not change because you earn it on a 1099. What changes is the documentation, and how many years of it a program demands. That rule is one of the first we still have to close: it is not verified in our sources and we will not print it as if it were.

Not verified

Boarder income does not count without documented history, and we could not verify in a primary source exactly what history each program demands. Counting it at zero is the conservative direction: counting it would inflate her buying power, which is the dangerous way to be wrong.

Every $100 a month of documented room rent is worth about $5,000 of maximum purchase price. The room is not a small thing: it is a down payment’s worth of buying power sitting in a spare bedroom.

Not against a generic 43%. Four rulebooks, four different pairs of numbers, and the order of generosity inverts the popular intuition that FHA forgives more.

Her ratio against the caps that actually apply to her

44.8%

Her housing ratio

$3,598.14 /mo ÷ $8,025.00

44.8%

Her total debt ratio

+ $0 in other monthly debts

  • FHA manual · no compensating factors
    Housing ratio44.8% / 31% · over
    Total debt44.8% / 43% · over

    HUD 4000.1 II.A.5.d.viii, manual underwriting with 0 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.

  • FHA manual · one documented factor
    Housing ratio44.8% / 37% · over
    Total debt44.8% / 47% · fits

    HUD 4000.1 II.A.5.d.viii, manual underwriting with 1 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.

  • FHA manual · two documented factors
    Housing ratio44.8% / 40% · over
    Total debt44.8% / 50% · fits

    HUD 4000.1 II.A.5.d.viii, manual underwriting with 2 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.

  • Conventional · through Desktop Underwriter
    Housing rationo housing cap
    Total debt46.8% / 50% · fits

    Fannie B3-6-02: Desktop Underwriter casefiles reach 50%. It is the highest ceiling in the catalog.

  • VA · residual income decides
    For context — she is not eligible for this one
    Housing rationo housing cap
    Total debt44.8% / 41% · over

    38 CFR 36.4340(c): 41% is a DOCUMENTATION TRIGGER, not a cap. What decides a VA loan is residual income.

  • USDA · the strictest pair in the catalogue
    hard capFor context — she is not eligible for this one
    Housing ratio44.8% / 29% · over
    Total debt44.8% / 41% · over

    7 CFR 3555.151(h)(1)(i): 29/41, the strictest pair in the catalog. Child care and transportation do NOT count as debt.

VA and USDA are shown so the order of generosity is visible, and it is worth seeing: the popular intuition that FHA forgives more is wrong in both directions. Her own ratios are computed on the FHA payment, because that is the loan she is carrying.

Her housing payment alone eats 44.8% of gross income. Under FHA’s manual matrix the housing cap is 31%, or 40% with two documented compensating factors — so the housing ratio breaks before the debt ratio does. Conventional through Desktop Underwriter has no housing cap at all: only 50% on total debt, and she is at 46.8%. That is why the conventional reaches this house and the FHA she was pre-approved for does not.

The question everybody actually asks — "can I buy the car before we close?" — solved by clearing the cap: room = cap × income − full housing payment.

What she can still borrow for something else

  • FHA manual · no compensating factorsNone. She is already over it with no debt at all.
  • FHA manual · one documented factor$173.61/mo
  • FHA manual · two documented factors$414.36/mo
  • Conventional · through Desktop Underwriter$259.99/mo

And this is the optimistic version. The file does not record her recurring debts, so everything above is computed at zero — the assumption that gives her the most buying power. Any real car payment, student loan or card minimum makes every number on this page worse, never better.

06Block

The three hits

The three hits that are already scheduled, with dates and amounts

None of these is a surprise to the industry. All three are arithmetic, all three are known before she signs, and none of them appears on a rate comparison site. Seeing them in advance is literally what this product is for.

Hit 1 · Day one

She starts out owing more than she asked for

1

The FHA upfront mortgage insurance premium is 1.75% of the base loan, and it is financed: it does not get paid at the table, it gets added to the balance. She asks for $418,907 and signs for $426,237. On the Closing Disclosure it lands in "services you cannot shop for" — the section she is not allowed to negotiate.

$418,907

What she asks for

+ $7,331

1.75% upfront premium, financed

$426,237

What she signs for

What that financed premium costs in interest over the full term: $9,814

The rule

24 CFR 203.284 — the premium is excluded from the loan-to-value ratio and excluded from the base of the annual premium: "excluding the portion of the remaining balance attributable to the premium". Almost every calculator on the internet adds it to both, which raises the rate band and the base it is charged on — the error gets paid twice.

And it is why FHA closing costs look the way they do at national scale

Observed
  • Median closing costs, FHA purchase$11,368
  • Median for her demographic group$9,080
  • Median closing costs, all purchases$6,684

Careful with a label that circulates badly: the $5,911 figure often quoted as "the conventional median" is not that. It is the median for the non-Hispanic white group, from a breakdown by group, not by product. Mixing the two turns a racial gap into a product comparison.

Source

CFPB, “2023 Mortgage Market Activity and Trends” (December 2024), pp. 24–27, Tables 3D and 3H · Originations from calendar year 2023

Hit 2 · Every month for thirty years

Her mortgage insurance never dies

2

With 3.5% down the ratio lands at 96.50%, which falls in the "above 95%" band, and the premium runs 55 basis points for the FULL term. It does not cancel. To get rid of it she has to refinance. With 10% down — a ratio at or below 90% — the same premium would stop after eleven years.

The rule

The duration of FHA’s annual premium depends only on loan-to-value and term. Not on her score, not on her payment history, not on what the house is worth later.

It does shrink, though — and that matters

FHA charges the annual premium on the average outstanding balance of each year, so the monthly amount falls every year: $191.08 in year one, $165.89 in year ten, $7.90 in year thirty. Conventional PMI is the opposite — a fixed amount on the ORIGINAL loan until it terminates. They are two different engines and showing them as one falsifies the comparison.

1102030$379$191
FHA · MIP · 360 monthsThe conventional, for contrast · 148 monthsYear of the loan

FHA · MIP

Monthly, year one
$191.08
How long it runs
360 months
Total premium over the life
$45,589

The conventional, for contrast

Monthly, fixed on the original loan
$378.97
Terminates by law
month 148 · April 2039
Total premium over the life
$56,087

On a conventional the premium terminates by federal law when the scheduled balance reaches 78% of the original value, and can be cancelled on written request at 80%. On her numbers that is month 148 — about 12.3 years — and then it is gone.

Assumption

Two things almost everybody gets wrong here. First, the thresholds run on the ORIGINAL value, so the house appreciating does not speed anything up. Second, the "month 89" figure that circulates everywhere only works at a rate near 3%: at today’s rates the balance takes far longer. That does not weaken the case — twelve years of premium is much more money than seven — but the number has to be the one the engine computes, not the inherited one.

Premium inside her 7-year horizon

FHA · MIP$15,454
The conventional, for contrast$31,833

And here is the honest twist: over seven years the FHA premium costs her LESS than the conventional one, because her score puts the private premium in an expensive band. The FHA premium is blind to score; the conventional one is not. Which program wins depends on how long she stays, and the answer flips.

Hit 3 · Month 14 · January 2028

In month fourteen her payment goes up, and it goes up twice at once

3

When the escrow computation year closes, the servicer re-analyses the account. If taxes and insurance went up, two things happen on the same day: the shortage gets spread over twelve months or more, AND the base deposit for the coming year rises because the annual bill is bigger. Both increases land in the same payment.

The rule

12 CFR 1024.17 — aggregate accounting is mandatory, the cushion may not exceed one sixth of the estimated annual disbursements (two months), and a shortage of one month’s deposit or more may be collected in equal instalments over twelve months or more. It is one of the four confusions the CFPB documents in its own 2025 complaints report.

What the account costs her at closing

Annual disbursements

$5,405 + $2,251

$7,656
Monthly escrow deposit
$637.96
Cushion (two months, the legal maximum)
$1,276
Initial deposit at closing
$5,405

Projected month-end balance through the first year

123456789101112

Dec 1, 2026 → Nov 1, 2027 · month 2: tax · month 11: insurance

Lowest projected balance: $1,276

The lowest point of that curve is the cushion and nothing more. That is the whole rule, and it is why the deposit is calculable to the dollar instead of estimated.

With the only growth figure that has a source

Projected

Insurance rising at 2.4% a year — the floor of the range the insurance regulators publish — and property tax flat, because no source in our research quantifies the increase. Result: $4.50 a month. That is the floor of the hit, not a forecast.

Source

NAIC, homeowners market analysis 2018–2024 (published range of 2.4% to 5.3% real per year)

No observation

Pull the Tarrant County appraisal district’s historical series, or the ACS county figures year over year.

And the mechanism that actually produces the shock in Texas

No observation

Here the initial escrow is usually built from the SELLER’s tax bill, while the buyer’s arrives reassessed at the purchase price. Nobody has quantified that reassessment in a primary source, so what follows is a sensitivity, not a projection: what the same arithmetic produces if the tax bill comes back higher.

Tax bill in year 2Shortage spread over 12 monthsRise in the base depositTotal monthly increase
$5,405 · flat, with a sourceProjected—+ $4.50+ $4.50
$5,675 · +5%+ $22.52+ $27.02+ $49.54
$5,945 · +10%+ $45.04+ $49.54+ $94.58
$6,215 · +15%+ $67.56+ $72.06+ $139.62
$6,485 · +20%+ $90.08+ $94.58+ $184.65

The double increment is the point: at every step above, roughly half of the rise is repaying last year’s hole and the other half is funding next year’s bigger bill. They are two separate mechanisms and they arrive in the same envelope, which is exactly why people call their servicer thinking it is a mistake.

Not verified

Confirm what the Tarrant County appraisal district actually does, and what the seller’s homestead exemption does to the first bill.

07Block

Down payment help

Down payment assistance: 2,746 programs, and almost nobody uses one

The median contribution is $15,000. In Dallas it is $16,862. Around 44% of buyers in the ten largest metros are potentially eligible for something — and in 2023 only 15% of FHA borrowers used government-source assistance. Nobody does this matching inside the quoting flow, and this is what it would look like.

Counted as of Jul 1, 2026

What the catalogue looks like

Observed

2,746

Programs counted nationally

2,114

Active and funded

632 inactive, suspended or waitlisted

$16,862

Median assistance · Dallas

$15,000 nationally

  • Available to first-time buyers62%
  • With no income restriction11%
  • Structured as a second mortgage56%
  • Outright grants9%
  • Run by municipal governments39%
  • Potentially eligible in the ten largest metros44%
  • FHA borrowers who used government-source help in 202315%

Read those last two together and the gap is the whole story: about 44% of buyers could get help, 15% of FHA borrowers got it from a government source. And 56% of these programs are a second mortgage, not a gift — that is debt with its own terms, and it belongs on the debt ratio, not in the "free money" column.

Source

Down Payment Resource, “Homeownership Program Index” as of Jul 1, 2026 (program counts and the split by type and by administrator); Urban Institute, “Expanding down payment assistance awareness could help more people afford a home” (take-up, eligibility and medians for 2022–2023) · Program count as of Jul 1, 2026; take-up and medians from 2022 and 2023

We cannot name the programs. We can run her file against the restrictions the survey itself describes, which is more than any quoting flow does today.

What her own file already says about the gates

  • She is a first-time buyer

    opens the 62% of programs reserved for first-time buyers

  • Household income against the area median

    unknown — and only 11% of programs have no income restriction

  • Primary residence, one unit

    the standard configuration these programs are written for

  • Tarrant County, Texas

    municipal programs are 39% of the catalogue, and they are county by county

$16,862 · median assistance in Dallas, 2022

What the Dallas median would do to this loan

Down paymentLoan-to-valuePremiumPremium over the lifeMonthly paymentHousing ratio
The 3.5% minimum, as pre-approved$15,19496.50%55 bps · 360 months$45,589$3,598.1444.8%
The minimum plus the Dallas median assistance$32,05692.62%50 bps · 360 months$39,776$3,462.3143.1%
All of her savings$43,00090.09%50 bps · 360 months$38,693$3,385.4342.2%
All of her savings plus the Dallas median assistance$59,86286.21%50 bps · 132 months$19,083$3,266.9740.7%

The help is not really about the down payment. Stacked on top of her own savings it drops the ratio below 90%, which is the cliff where the premium stops running for the full term and stops after eleven years instead — $26,505 less premium over the life of the loan, $331.17 less every month, and a housing ratio that drops from 44.8% to 40.7% against a 40% cap. It moves her from "no" to "almost" on the only rulebook that is published.

And the smallest version of the same point: she is $410 short of the 90.00% cliff. The median Dallas assistance is 41 times that.

We deliberately do not put these scenarios into the total-cost-at-horizon column. The engine counts the whole down payment as her cash, and 56% of these programs are a second mortgage — somebody else’s money with its own repayment terms. Dropping a third party’s money into her "down payment" line would falsify the comparison in the flattering direction.

What we cannot do yet, said plainly

The catalogue of the 2,746 programs belongs to a commercial data provider, not to a public agency. We do not have it. There are two honest ways to get it — license it, or build the crawler across the 1,068 municipal administrators — and both are work, not a weekend. Until then this block shows the shape of the opportunity and her own gates, and names no program.

Aug 25, 2026 · 7 federal agencies

And the ground moved two weeks ago. On 25 August 2026 seven federal agencies rescinded the 2022 interagency statement on Special Purpose Credit Programs — the main legal vehicle for credit and assistance aimed at underserved buyers. It does not make those programs illegal under the fair lending statutes, but it withdraws the regulatory cover, and the 2,746-program landscape is exactly the landscape that assumed that cover existed.

Source

HUD, CFPB, DOJ, FDIC, NCUA, OCC and FHFA — rescission of the February 2022 “Interagency Statement on Special Purpose Credit Programs Under ECOA and Regulation B”

08Block

What this file still does not know

What this file still does not know

Every number above has a source or a label. These are the holes — what is missing, what it blocks, and what it would take to close it. A product that hides this list is a product that will be caught out by a lender in a meeting.

What is missingWhat it blocksHow it gets closed
The daily rate benchmark for her exact LTV × FICO segmentEvery payment on this page is anchored to the general 30-year conforming index, because the series for her cross has no observation we could verify. For a high LTV with a mid score the general index understates — the observed gap to the LTV>80 / FICO<680 cross was 16.5 basis points.Pull the segmented series from the public data release. And note the dependency risk: that index is licensed revocably, so a plan without a fallback is a product risk, not an attribution detail.
A verified rate differential for FHA against conventionalFHA is quoted here with the conventional anchor, with nothing subtracted. The previous version of our engine carried hand-written differentials; they came from intuition, and they are gone.Observe the program-specific series instead of postulating the spread.
Her recurring monthly debtsEvery ratio on this page is computed at zero debt, which is the assumption that gives her the most buying power. The verdicts are a ceiling, not an estimate.Ask her. It is field 12 of the 20.
The area median income for Fort WorthWhether the loan-level price adjustment waiver applies. It is worth $4,737 on the conventional, and it is the difference between 1.125 points and zero.Cross her qualifying income against the published area median for the metro.
The automated underwriting verdictThe real answer to "how much house". The manual matrices are public; the systems that decide almost all real volume are not, and there is no public sandbox for any of them. Our $292,000 FHA ceiling is the manual floor, and the AUS may well say more.It requires becoming an approved business partner. This is the wall, and it is the same wall that separates showing calculated cost from showing a firm price.
The catalogue of the 2,746 down payment assistance programsNaming a single program she could actually apply to.License the commercial database, or build the crawler across 1,068 municipal administrators.
Transfer taxes, verified state by stateBlock E of the comparison outside Texas. There is no primary national source: it is 51 jurisdictions plus counties, built by hand. And it is not small — in Philadelphia the transfer tax on a $500,000 house is $22,890, more than three times the national median total loan cost, and it appears in no comparison site because it is not in the federal data set.Build the table from each state’s department of revenue. There is no shortcut.
Real private mortgage insurance pricingThe conventional premium on this page is modelled on the one insurer of six that still publishes a rate card. The other five quote only through a proprietary engine inside the lender’s pricing system — and the borrower does not even choose the insurer.Agreements with at least two mortgage insurers. Until then it stays a declared proxy.

What the engine itself declared while computing this file

Not verified

Every result our engine returns carries its assumptions attached, with the reliability label of the constant it came from. These are the ones behind the numbers on this page. Nothing here was filtered: if it does not fit on screen, that is the screen’s problem, not the engine’s.

  • FHA_CONVENCION_PROMEDIONot verified

    Averaged using the “MEDIA_DE_SALDOS_MENSUALES” convention. HUD says “average outstanding balance” but does not publish the exact averaging convention in any source we were able to read.

  • FHA_UMBRAL_EN_DISPUTADisputed

    Pricing threshold applied: $726,200 (reading: APPENDIX_1_0). Appendix 1.0 — revised Mar 20, 2023 and still in force in Update 18 of Aug 12, 2026 — prints $726,200; Mortgagee Letter 2023-05 redefines that same threshold as “the national conforming loan limit”, which is $832,750 in 2026. Both are verified against primary source and they contradict each other.

The rest of the assumptions, all verified (15)
  • SIN_AVALUO

    No appraisal has come in, so the LTV is computed off the purchase price. On a purchase the value is the LESSER of price and appraisal, so a low appraisal would raise the LTV and could move the band.

  • LIMITE_FHA_SIN_CONDADO

    FHA’s national FLOOR was used — $541,287 for one unit — which is the limit that governs in most counties. The real limit is by county and is NOT in FHFA’s CSV: it comes out of HUD’s CHUMS files at entp.hud.gov. In a high-cost county it reaches $1,249,125, so this cut can exclude a loan that would be perfectly eligible there.

  • LLPA_NO_APLICA

    Loan-level price adjustments and Credit Fees belong to Fannie and Freddie: they do not apply to FHA, VA or USDA, which carry their own premium and fee structures.

  • FHA_UFMIP_EXCLUIDO

    The financed upfront premium is excluded from the LTV and from the base of the annual MIP, under 24 CFR 203.284 — “excluding the portion of the remaining balance attributable to the premium”.

  • FHA_MIP_DECRECIENTE

    Annual MIP of 55 basis points on the average balance for the year, which means the premium FALLS every year. The year-one figure is not a 30-year payment.

  • FHA_MIP_NO_MUERE

    At 96.5% LTV the MIP runs the FULL term — 360 months — and never cancels. The only way out is to refinance. Only at 90% LTV or below, meaning 10% down, would it drop to 11 years.

  • APR_EXCLUSIONES

    Excluded from the finance charge: 0 charge(s) under 1026.4(c)(7) — title, appraisal, credit report, survey and document preparation, where bona fide — and under 1026.4(c)(5), seller’s points. The legal APR leaves out precisely what varies most between lenders.

  • APR_TOLERANCIA

    The law treats an APR as accurate if it is off by no more than 0.125 percentage points — one eighth. Smaller differences are NOT informative, and a ranking should not fake a precision the law itself does not require.

  • BANDA_SCORE_MAYOR_O_IGUAL

    Score bands are read as “>=”. Fannie prints “>740” and Freddie prints “>=740”, so a score of exactly 740 falls into no row of Fannie’s grid at all. We follow Freddie’s reading.

  • PMI_PROXY_ENACT

    Mortgage insurance priced off Enact’s published rate card (updated Jul 17, 2025). MGIC, Radian, Essent, National MI and Arch quote only through proprietary engines inside the lender’s own pricing system, and the borrower does not even choose the insurer. This is a declared proxy.

  • HPA_SOBRE_VALOR_ORIGINAL

    All three dates run off the ORIGINAL value — the lesser of price and appraisal. Appreciation does not speed up cancellation under the Homeowners Protection Act; that is what the GSE rules are for, and those are different and are not verified in this project.

  • HPA_BUEN_HISTORIAL

    Cancelling on request requires a “good payment history”, and that is TWO separate windows: nothing 60 or more days late in months 13 through 24 back, and nothing 30 or more days late in the last 12. The intuitive version — “nothing 60+ days late in the past year” — denies cancellations the law actually grants.

  • ESCROW_AGREGADO

    Aggregate accounting is mandatory (12 CFR 1024.17): the account is analyzed as a whole. The cushion is capped at one sixth of the annual disbursements — 2 months.

  • ESCROW_DOBLE_INCREMENTO

    The year-two payment rises along TWO paths at once: the shortage spread over 12 months (§1024.17(f)(3)) PLUS the increase in the base deposit, because the annual bill itself grew. It is one of the four confusions the CFPB documents in its 2025 complaint report.

  • ESCROW_REPARTO_DEL_FALTANTE

    The shortage is smaller than one month of deposit, which means the servicer MAY demand it in a single payment. Here it is spread over 12 months, which is both the usual practice and the conservative reading for the monthly payment.

And the warnings the engine raised about this specific loan

FHA · 3.5%

  • There is no verified spread for FHA against conventional. OBMMI publishes its own series — OBMMIFHA30YF, for instance — and that is the right route: observe the spread, do not postulate it. This is quoted off the conventional anchor, with nothing invented subtracted from it.
  • No observation for segment OBMMIC30YFLVGT80FB680A699. The general 30-year conforming OBMMI is used instead, and it averages every profile: for a high LTV with a mid score it UNDERSTATES the rate. The observed gap between the general index and the LTV>80 / FICO<680 cross was 16.5 basis points.
  • She will pay FHA mortgage insurance for the full 30 years. The only way out is to refinance. With 10% down — 90% LTV or less — it would drop to 11 years.
  • Debt-to-income of 45%, above FHA’s 43%. HUD 4000.1 II.A.5.d.viii, manual underwriting with 0 documented compensating factor(s). With an Accept from the TOTAL Scorecard the automated system decides, not this matrix.
  • Housing ratio of 45%, above FHA’s 31%.

Conventional · 97%

  • No observation for segment OBMMIC30YFLVGT80FB680A699. The general 30-year conforming OBMMI is used instead, and it averages every profile: for a high LTV with a mid score it UNDERSTATES the rate. The observed gap between the general index and the LTV>80 / FICO<680 cross was 16.5 basis points.

Assumption

A calculator that returns a number and swallows its own warnings is not a calculator: it is a sales tool. Ours returns both, and the screen shows both.