Case Study 8.2 — The File That Was Right: A Household Approved at the Top of Its Ratio

⚠️ This is a clearly labeled composite. It is not a real household, and no figure in it is a measurement of anything. It is assembled from patterns that are ordinary in residential lending — a high but approvable back-end ratio, a large increase over prior housing expense, thin reserves, and an income disruption in the first year — and every number in it is constructed to compute. No statistic here should be cited as data. The regulatory history at the end is real and public.

The file, as underwritten

A two-earner household, first-time buyers, buying a single-family home at \$305,000.

Every number in that table is defensible. The income was fully documented. The debts came off the credit report. The loan was a fully amortizing fixed-rate mortgage with no teaser rate, no negative amortization, no interest-only period, and no prepayment penalty. The file received an automated approval and closed without incident.

Nobody in this transaction did anything wrong, and nothing in it was a close call. A 49.00% back-end ratio is inside what an automated conventional approval can return. That is the entire point of the case.

The number nobody said out loud

$$\frac{\$2{,}142.00}{\$1{,}040.00} = 2.06 \quad \text{— an increase of } 106.0\%$$

This household was proposing to pay more than twice what it had been paying for housing, after four years at the same rent. That figure appears nowhere in the qualifying arithmetic. It is not a ratio, it has no limit attached to it on a conventional file, and no automated system reported it. The originator had every fact needed to compute it — the rental history was in the file, verified — and there is no evidence anyone ever did.

What the ratio could not see

Now run the household's own arithmetic, the way §8.2 runs it.

(1% of \$305,000 is \$3,050 a year, or \$254.17 a month — a planning rule of thumb, not a guideline.)

From the first payment, this household was running about \$126 a month short of its own costs, and the shortfall was funded the way such shortfalls are always funded: revolving credit. Over the first eight months that is roughly \$1,009 of new balances — which begins adding minimum payments to a ratio that was already at 49.00%.

Nothing in the loan file reflects any of this, because nothing in a loan file is designed to.

The counterfactual: \$35,000 of house

Hold everything else constant and put the same household in a \$270,000 home. On this file's cost structure, the housing payment scales to roughly $\$2{,}142.00 \div 305 \times 270 = \$1{,}896.20$.

| | At \$305,000 | At \$270,000 | |---|---|---| | Housing payment | \$2,142.00 | \$1,896.20 | | Back-end ratio | 49.00% | 45.39% | | Net, less payment and debts | \$1,818.00 | \$2,063.80 | | Less living expenses | −\$1,690.00 | −\$1,690.00 | | Margin | \$128.00 | \$373.80 | | Less maintenance at 1%/yr | −\$254.17 | −\$225.00 | | What is actually left | −\$126.17** | **+\$148.80 |

Same household, same income, same debts, both files approvable, and \$35,000 of purchase price is the difference between running a small deficit every month and running a small surplus every month. After eight months, the surplus household has accumulated about \$1,190; the deficit household has borrowed about \$1,009. The gap between the two, by month eight, is roughly **\$2,200** — and it is entirely invisible to a back-end ratio computed on gross income.

What happened

In month eight the primary earner's hours were reduced by roughly ten percent. Gross income fell from \$6,800.00 to about **\$6,120.00**; net deposits fell to roughly \$4,700.00.

Recompute the ratio on the same obligations, which have not changed:

$$\frac{\$3{,}332.00}{\$6{,}120.00} = 54.44\%$$

The household's monthly position moves from about −\$126 to about **−\$576**. The revolving balances, already growing, accelerate. The first thirty-day delinquency is recorded in month eleven.

At \$270,000 the same reduction would have left them at roughly −\$301 a month before maintenance choices — still difficult, still requiring cuts, and materially more survivable, with eight months of accumulated buffer instead of eight months of accumulated debt.

Why the regulatory answer is only a partial answer

The post-crisis rulebook addressed a real set of failures with real precision. The Dodd-Frank Act established the ability-to-repay requirement: a creditor must make a reasonable and good-faith determination that the consumer has a reasonable ability to repay, based on verified information, and the Qualified Mortgage framework built a category of loans presumed to meet it. Those rules eliminated the product features that made the crisis-era failures so violent — unverified income, negative amortization, interest-only periods, and payment resets that doubled an obligation on a schedule the borrower had not understood.

The general QM category originally incorporated a specific debt-to-income threshold; the Bureau later replaced that with a price-based approach. Verify the current rule text and thresholds — this framework has been amended, and it will be again.

Note what none of it reaches. This loan was fully documented, fully amortizing, fixed for the full term, and had no reset of any kind. The ability-to-repay determination is made on the ratio, and the ratio is computed on gross income. A household can satisfy every requirement of the rule and still have negative margin in month one, because margin is not what the rule measures.

There is one significant exception in mainstream American lending, and it is worth knowing precisely because it proves the point: VA loans require a residual income analysis — a calculation of the dollars remaining after the mortgage, other obligations, taxes, and estimated maintenance and utilities, tested against a figure that varies by region and family size — in addition to the ratio. It is a structurally different question, and it is the closest thing in the guidelines to the arithmetic in §8.2. Chapter 17 covers it. (Verify the current residual income tables and requirements with the Department of Veterans Affairs.)

The lesson

The file was right and the outcome was bad, and both of those sentences are true at the same time.

That is the hardest thing to hold in this business, and it is why §8.2 exists. If you believe the ratio answers the borrower's question, then every approvable file is a good file, and a household like this one is simply unlucky. If you understand that the ratio answers a different question — a narrower one, asked by an investor who wants to know whether the loan will perform on average across a pool — then the gap becomes visible, and it becomes yours, because you are the only person in the transaction who ever speaks to the household.

Three specific things a loan officer could have done, none of which requires any authority you do not already have:

1. Compute the multiple. \$2,142.00 against \$1,040.00 is 2.06 times. Sixty seconds, using a number already verified in the file. Say it out loud on day 1.
2. Ask for the net deposit figure and the living-expense total, and subtract. Not to decide — to show. The household would have seen −\$126.17 before they made an offer.
3. Put the second price on the page. Not instead of the first: alongside it. \$305,000 and \$270,000, with the margin figure under each.

And one thing to be equally clear about: this household is not a cautionary tale about people who overreach. They were shown a number they qualified for and they believed it, which is a reasonable thing to do when a licensed professional says it. The failure in this composite belongs to a process that answers a question nobody asked and stays silent on the one everybody is actually asking.

Discussion questions

1. The 2.06× multiple was computable from documents already verified in the file. Why do you think it was not computed? Name the structural reasons, not the personal ones.

2. A 49.00% back-end ratio is approvable. A 45.39% back-end ratio is approvable. If both files pass, on what basis does a loan officer raise the difference at all — and what is the line between raising it and substituting your judgment for the borrower's?

3. This household's living-expense figure was \$1,690.00. You did not verify it and could not. Does that make the exercise worthless? Argue both sides, then state what you would actually do.

4. VA's residual income test asks a version of §8.2's question inside the guidelines. Why do you think the conventional and FHA frameworks have not adopted something similar? What would be gained and what would be lost?

5. Rewrite the day-1 conversation for this file. Write the four sentences that would have put −\$126.17 in front of this household before they made an offer, and do it without telling them what to do.

6. Suppose you have this conversation and the household buys the \$305,000 house anyway. Have you accomplished anything? Be specific about what, and about what your file should contain afterward.