Mortgage Stress Test Pack

Why this Data Set exists

Mortgage origination tooling is well-developed for the modal case: prime credit, stable W-2 income, conforming loan, 30-year fixed. The tooling falls over interestingly outside that lane: the gig-worker borrower whose income is structured as 1099 receipts and deposit history; the small-business-owner borrower with K-1 distributions and depreciation add-backs; the post-divorce borrower whose income just changed and DTI calculation depends on which look-back window you apply; the underwater homeowner evaluating modification eligibility under FNMA flex modification or VA partial claim.

These non-modal cases dominate the actual loan-modification, default-risk, and fair-lending workstreams that consume servicer resources. They're also the cases most poorly represented in fixture data. Production loan-tape data captures the originated loans but not the rejected applications or the modification candidates; servicing data captures ongoing performance but not the structural details of the original underwriting decision; and the rare integrated views are heavily customer-controlled.

This Data Set provides 90 households where the full mortgage lifecycle is structured as JSON fields: the application surface (income, assets, employment history, credit profile); current LTV computed from actual property valuation and outstanding balance; DTI calculation pre-computed with the realistic income-source breakdown; forbearance and modification flags where applicable; and the mortgage-origination edge cases (DPA programs, FHA, VA, USDA loans, ITIN mortgages) that test the breadth of underwriting logic.

Use Cases

Who uses this Data Set

What's inside

The 90 households span four mortgage-relevant archetypes: young dual-income couples with first-mortgage applications (F-03), young families with new mortgages already in place (A-01), first-time homebuyers with DTI stretching (MB-01), and distressed-mortgage households evaluating modification (MB-02). The mix is intentional — about 35% are at the application stage (testing origination logic), 30% are recent originations within the first 24 months (testing servicing-onboarding logic), 25% are distressed or in forbearance (testing modification logic), and 10% are underwater (testing strategic-default and short-sale workflows).

Every household carries the structured fields a mortgage application requires: W-2 income, 1099 income, deposit/seasoning balances, paystub-equivalent year-to-date earnings, employment history (at least 2 years of continuous tenure or a structured gap explanation), credit tradeline records, and property appraisal data. Current state includes outstanding balance, current LTV (computed from current appraisal value rather than original purchase price), DTI ratio with the income-source breakdown, escrow account balance, and any forbearance / modification flags. Forbearance / modification scenarios include the structured eligibility-determination data needed for FNMA flex, VA partial claim, FHA-HAMP, and other modification programs.

The Data Set ships as JSON and CSV, accompanied by the WealthSynth Methodology PDF. The methodology covers the application-field schema, the DTI calculation methodology (with the realistic income-source-specific add-back rules for K-1 distributions, depreciation, capital gains, and other non-W-2 income), the modification-program eligibility decision trees, and the calibration source for typical loan-program distributions (FHA, VA, USDA, conforming, jumbo, ITIN). No rendered documents (W-2 PDFs, paystubs, bank statements, etc.) are shipped — every value is in the JSON, ready for buyers to render into their own document templates if needed.

Preview a sample household

A redacted summary of one household from this Data Set — names, employers, exact balances, and metro area are stripped. Ages are bucketed, income and net worth are reported as bands. The full record (and all 90 like it) ships in the ZIP.

Technical Highlights

Sample Schema Fields

{
  "liabilities.mortgage.balance": <value>,
  "liabilities.mortgage.original_amount": <value>,
  "real_estate.current_appraisal": <value>,
  "credit.dti_ratio": <value>,
  "real_estate.ltv_pct": <value>
}

Sample queries

households.filter(h =>
  h.events.life_events.some(e =>
    e.type === 'mortgage_application' &&
    e.status === 'pending') &&
  h.credit.dti_ratio >= 0.43 &&
  h.credit.dti_ratio < 0.50
)

households.filter(h =>
  h.liabilities.mortgage.payments_past_due >= 1 &&
  h.liabilities.mortgage.modification_eligibility_flags
    .some(f => f === 'FNMA_flex' || f === 'FHA_HAMP' ||
               f === 'VA_partial_claim')
)

households.filter(h =>
  h.real_estate.ltv_pct > 1.0 &&
  h.real_estate.ownership_status === 'primary'
)

households.filter(h =>
  h.events.life_events.some(e =>
    e.type === 'mortgage_application') &&
  ['1099', 'self_employed', 'K-1', 'royalty']
    .includes(h.income.primary_source_type)
)

Methodology

Each household's mortgage profile is generated against archetype-specific patterns. Young dual-income couples are weighted toward conforming conventional loans with 15-20% down. First-time homebuyers (MB-01) carry realistic FHA, VA, USDA, or DPA-program structures with the higher LTV and DTI typical of first-time-buyer programs. Distressed mortgages (MB-02) have realistic delinquency patterns and modification-eligibility profiles. Underwater scenarios are seeded at realistic frequencies for declining housing-market regimes. DTI calculations use the actual income-source-specific add-back rules: depreciation add-back for self-employed Schedule C income; K-1 distribution treatment that varies by guarantor type; income-volatility-adjusted documentation for 1099 contractors. The corpus passes the WealthSynth consistency validator (DTI/LTV math is correct; modification-eligibility flags fire when underlying conditions are met; income-documentation completeness is internally consistent) and the LLM-as-judge gate. Annual refresh tracks FNMA / FHLMC eligibility changes, FHA / VA / USDA loan-limit updates, and any major-modification-program statutory changes.

Included Archetypes (4)

Frequently asked questions