Amlok
The idea in brief.
A mortgage is assessed at a moment in time and repaid over many years. Income, obligations, household spending, property value, and interest-rate terms can change on different schedules. A useful risk view preserves the original assumptions and distinguishes new evidence from a new interpretation. [1]
- Track affordability and collateral separately; neither one replaces the other.
- Match the definition and timing of income, housing costs, debts, and property values.
- Keep scenario results conditional on their assumptions and preserve the lender’s policy and human decision.
The Canadian macro indicators below describe groups of borrowers, not an individual’s mortgage risk. The household budget is hypothetical and is not a qualification test. No mortgage repayment model has been validated by this guide.
Build two connected views
The affordability view asks whether the household can meet its obligations under a specified income and expense path. The collateral view asks what supports the exposure if repayment fails. A borrower can have substantial equity and still face immediate cash-flow pressure. Another can maintain payments while a falling property value reduces collateral protection.
Preserve income source, period, gross/net convention, stability, and verification method. Housing costs should identify which items are included, such as principal and interest, property taxes, heating, or relevant fees. Other obligations need a consistent payment basis. Ratios computed from mismatched periods or definitions can look precise while answering the wrong question. [1][2]
Property value also needs a date and method. An appraisal, an automated estimate, and a current sale price are not interchangeable. Record prior liens and the exposure included in a loan-to-value calculation. Monitoring should update the evidence without overwriting the original underwriting record.
Two views, one review history
- 01Household cash flow
Verified income, timing, essential costs, and debts.
- 02Property & liens
Value method, valuation date, and secured exposure.
- 03Terms & scenarios
Renewal date, rate structure, and payment assumptions.
- 04Institutional review
Policy interpretation, exceptions, and ownership.
Interpret household stress with its definition
The Bank of Canada’s 2026 Financial Stability Report describes roughly 2.5% of borrowers without a mortgage and 1.3% of mortgage holders as more than 60 days late on at least one account. This is a borrower-level measure across accounts. It is not a 90-day mortgage-arrears rate and does not show that taking a mortgage causes lower risk. [3]
The comparison illustrates why population definitions travel with every statistic. Mortgage holders can differ from other borrowers in income, assets, age, credit access, and selection. Applying an aggregate group rate to a particular applicant would discard that context. Use macro evidence to frame review priorities and scenarios, then assess the household using its authorized records.
When monitoring changes over time, keep the delinquency definition, account coverage, and population stable. If one period uses any-account arrears and another uses mortgage-only arrears, the trend is not comparable. A dashboard should expose that discontinuity rather than connect the numbers with a smooth line.
The population changes the interpretation
View figure data
| Measure | Value |
|---|---|
| Borrowers without a mortgage | ≈2.5% |
| Mortgage holders | ≈1.3% |
Stress the residual, not just the ratio
Consider a hypothetical household with monthly take-home income of CAD 7,000. Mortgage payments are CAD 2,400, other housing costs CAD 600, essential non-housing costs CAD 2,400, and other debt payments CAD 600. The simplified residual is CAD 1,000 before unmodelled discretionary spending and saving.
Assume the mortgage payment increases by CAD 500 at renewal, with every other input unchanged. The residual falls to CAD 500. If take-home income also falls by 10%, the residual becomes negative CAD 200. This is arithmetic under specified assumptions, not a forecast of the household’s behaviour or a substitute for a lender’s qualification rules.
The example shows why a combined scenario matters: two manageable changes in isolation may interact. A useful review considers available savings, the timing and duration of the shock, realistic spending flexibility, and any relevant assistance process. It should not infer distress solely because a stylized scenario is negative.
| Measure | Baseline | Higher payment | Higher payment + 10% less income |
|---|---|---|---|
| Take-home income | 7,000 | 7,000 | 6,300 |
| Mortgage payment | −2,400 | −2,900 | −2,900 |
| Other housing costs | −600 | −600 | −600 |
| Essential non-housing costs | −2,400 | −2,400 | −2,400 |
| Other debt payments | −600 | −600 | −600 |
| Residual | 1,000 | 500 | −200 |
Make renewal a scheduled evidence review
A review window should be anchored to the actual contract and renewal date, not a generic calendar alert. Record the remaining amortization, current payment structure, and assumptions used in any alternative payment calculation. Distinguish a known contractual change from a market-rate scenario.
Where new financial records are authorized, compare them with the origination baseline on a consistent basis. Changes in payroll timing, account coverage, or household composition may require interpretation. A missing account should reduce confidence in the view; it should not be treated as a disappearance of the associated obligations.
The review packet should show what changed, the source of the change, the periods compared, and unresolved questions. Keep the borrower communication and institutional response separate from the analytical signal. This makes it possible to correct evidence or provide context before a consequential decision. [4][1]
Know what the research data can test
Amlok’s published inventory includes 208,818 New York mortgage origination records from the 2024 HMDA slice. Those data describe originations and contain privacy-modified fields. They do not supply subsequent repayment histories, so they cannot establish a mortgage-default prediction result on their own. [5]
A repayment study would need an authorized outcome source, a documented linkage method, a precise event definition, sufficient follow-up, and a realistic point-in-time feature view. Calibration and errors should be inspected across relevant cohorts and data-coverage conditions. Any institution-specific rule must be evaluated in the population where it will be used.
Until those requirements are met, the defensible contribution is better evidence organization, transparent calculations, and a reproducible review process. Those are meaningful operational goals, but they are different from claiming lower mortgage losses. [6]
Sources & methodology.
Figures identify their population, period, and units. Scenarios marked illustrative are Amlok calculations using stated assumptions, not observed customer outcomes.
- 1Mortgage lending workflowsAmlok research and product documentation
Housing affordability, property, collateral, and ongoing monitoring context.
- 2Consumer underwriting researchAmlok research and product documentation
Affordability, temporal evidence, and consumer-specific evaluation requirements.
- 3Financial Stability Report 2026: HouseholdsBank of Canada · 2026
Borrower-level share more than 60 days late on at least one account, grouped by mortgage ownership. Not the 90-day mortgage-delinquency rate.
- 4Monitoring and change detectionAmlok research and product documentation
Proposed monitoring design, observation windows, baselines, and evaluation requirements.
- 5Dataset inventory and acquisition scopeAmlok research and product documentation
Inventory preserves source-specific units: loan-program rows, disclosures, originations, and financial facts. A row count is not a closed-deal count.
- 6Evaluation designAmlok research and product documentation
Evaluation protocols and research objectives. A specified protocol is not evidence that an experiment has been completed.
