Why a Mortgage File Has Fifty Documents to Establish About Twelve Facts
A completed mortgage file runs to several hundred pages across dozens of document types. Set against what it actually establishes, that is a striking ratio: the file exists to determine roughly a dozen things.
Who the borrower is. What they earn. What they owe. What they have. What the property is worth. Whether it is insurable and lienable. What the loan's terms are. Whether the borrower can be expected to repay on those terms. Whether the required disclosures were made, correctly and on time.
Fifty documents, twelve facts. Understanding why that ratio exists explains most of what is difficult about reviewing one.
The file is built out of corroboration
The ratio exists because a mortgage is underwritten on facts that cannot be observed directly and must be established through evidence — and because the party supplying most of that evidence has an interest in the outcome.
So income is not asked, it is triangulated. The borrower states it on the 1003. A paystub shows current earnings. A W-2 shows what was reported to the IRS. A tax return shows what was filed. A verification of employment obtains it from the employer directly. Each of these is a different party's account of the same underlying fact, with different incentives and different failure modes.
This is the file's actual design principle, and it is a sound one. It also produces a characteristic failure mode of mortgage quality control: many defects are not in a document at all. They are in the relationship between documents.
One qualification before going further, because the naive version of this idea causes real problems. Corroborating documents are not expected to match literally. A paystub reports a current pay period, a W-2 reports a completed tax year, a tax return may include income from sources the employer never sees, and a verification of employment may state a base figure that excludes variable compensation. Different periods and different definitions produce different numbers legitimately, and a system that flags every numerical difference as a discrepancy is generating noise, not findings.
Qualifying income is a calculation, and more than one method can be defensible for the same borrower — a 24-month average, a year-to-date annualisation, a declining-income treatment. Guidelines permit judgement here. [1] The requirement is not that a single correct figure exists and every document points to it. It is that the method chosen was appropriate for the borrower's circumstances, applied to the right documents, and written down — so that a reviewer two years later can follow the derivation rather than guess at it. The defect is an unexplained difference, not a difference.
Where the file's structure creates blind spots
If the file establishes each fact through several independent documents, then a review organised around documents is organised against the grain of the evidence.
The single-document blind spot. A reviewer examining a paystub can confirm it is legitimate, current, and belongs to the borrower. They cannot tell from it whether its year-to-date figure is consistent with the W-2, or whether the annualised income it implies matches what underwriting used. Each document passes. The file is wrong.
The sequencing blind spot. Many requirements are about when, not what. The Loan Estimate must reach the borrower within a period of application. The Closing Disclosure must be received a required interval before consummation. Intent to proceed governs when fees may be collected. Credit and appraisal have validity windows. A file where every document is correct and the sequence is wrong is defective, and nothing in the documents themselves announces it.
The absence blind spot. A document that should be in the file given the loan's programme and characteristics, and is not, cannot be found by examining what is there. Detecting it requires knowing what the file should contain — which is a function of the loan's programme, property type, occupancy, borrower type and jurisdiction, all of which are themselves facts in the file.
The third is the hardest, and it is the one where the difference between "a check passed" and "no check ran" stops being philosophical.
The twelve facts, and the documents that contest them
Identity. Established by the application, credit report and SSA-89 verification. Contested by mismatches in name, date of birth or SSN across documents that were completed at different times.
Income. The most heavily corroborated fact in the file and the most frequent defect category. Multiple documents, several legitimate calculation methods, and a qualifying figure that must match what underwriting used.
Employment. Established by paystub, W-2 and verification of employment; contested by gaps, recent changes, or a verification dated inconsistently with the employment it describes.
Liabilities. Established by the credit report; contested by the 1003's declarations and by debts that appear on a tax return or bank statement but not on credit.
Assets. Established by statements and verification of deposit; contested by unsourced large deposits, and by funds to close that do not reconcile against the settlement statement.
Property value. Established by appraisal; contested by comparables, adjustments, and by the property characteristics stated elsewhere in the file.
Property eligibility. Established by appraisal, condominium project documents and occupancy declarations; contested by project-level facts that have nothing to do with the borrower.
Insurability. Established by the hazard declaration and flood certification; contested by coverage amounts, deductibles and effective dates that must align with the loan amount and closing.
Title. Established by the title commitment; contested by liens, easements and outstanding requirements.
Loan terms. Established by the note; contested by every other document that restates them — the Closing Disclosure, the 1008, the security instrument.
Ability to repay. Established by the underwriting summary's ratios; contested by every input that feeds them. This is where an error in any earlier fact ultimately expresses itself.
Disclosure compliance. Established by the disclosures and their timing records; contested by dates rather than content.
What this implies about reviewing a file
Three consequences follow from the file's structure, and they hold regardless of who or what performs the review.
Reviewing document by document reviews the file against its own design. The evidence is organised as corroboration, so a method that examines documents in isolation cannot see the property that matters most.
Reading well is necessary and not sufficient. Extracting every field from every document with perfect accuracy leaves the actual question untouched: are the differences between these values explained, and do the results satisfy the requirements that apply. Reading is a prerequisite for that question, not an answer to it — and an inaccurate reading makes the question unanswerable, so extraction quality is not a lesser concern.
Coverage of the file matters as much as depth on any document. Because defects live in relationships, the number of documents examined determines how many relationships can be checked at all. A review of the six most important documents cannot find a contradiction involving the seventh.
The fifty documents are not bureaucratic excess. They are what it takes to establish twelve facts from interested parties. The difficulty of reviewing them is not the page count — it is that the information the review needs is distributed across the gaps between them.
Sources
- Fannie Mae Selling Guide, B3-3.1-01, General Income Information. Qualifying income is a calculation with more than one defensible method; the requirement is that the method used be documented.