An arrears dashboard can look reassuring if you read only one column. This week's Canadian data are a reason to read the next one.
Statistics Canada published its second-quarter 2026 non-bank mortgage survey on September 28, covering a market with $420.2 billion in outstanding residential mortgages. Canadian Mortgage Trends reports that balances on mortgages with overdue payments rose to $12.1 billion from $10.8 billion a year earlier, while the count of affected mortgages changed little.
The executive takeaway is not that every overdue mortgage is headed for loss. It is that account counts, and dollar exposure answer different questions, and your operating decisions should reflect both.
The survey defines mortgages in arrears as loans with payments overdue at quarter-end. Canadian Mortgage Trends explicitly clarifies that the reported balance is what borrowers still owe on those loans, not the amount of missed payments, and that any overdue duration qualifies.
That prevents two bad readings. The $12.1 billion is not a bill for unpaid instalments, and it is not a pool composed exclusively of loans more than 90 days past due.
The survey includes local credit unions, mortgage finance companies, trusts, insurers, mortgage investment corporations, and private lenders. “Non-bank” should not be used as shorthand for one uniform lender type or risk profile.
Consider two hypothetical portfolios with the same number of overdue loans. If one carries larger balances, a count-based report will miss an important difference in exposure.
The national release should prompt that question inside the institution, not answer it on the institution's behalf. Your mix of borrowers, collateral, insured loans, and delinquency ages may differ from the aggregate.
The reported uninsured overdue-loan balance increased to $8.7 billion from $7.7 billion, accounting for most of the overall dollar increase. That is a reason to examine your segmentation, not evidence that every uninsured loan needs a tighter approval threshold.
The discipline is to ask what changed, where it changed, and whether the movement reflects balance size, persistence, concentration or something else. A bigger number is an investigation prompt, not a diagnosis.
For servicing, define what happens after a payment becomes overdue. Who verifies the account status, confirms the borrower's explanation, assesses the appropriate response, and owns the next contact?
Differentiate temporary payment friction from recurring delinquency without assuming either from the first alert. Keep the decision history visible when a loan returns to its current status; a cured account and an account that was never overdue should not become indistinguishable in the analysis.
For underwriting, review whether recent performance reveals gaps in the original assessment or simply a change in circumstances. Avoid turning a servicing observation into a blanket origination restriction before the evidence supports it.
Policy is the moat when thresholds, exceptions, and escalation paths produce consistent decisions. More alerts without that structure simply give the backlog better graphics.
Start with the existing loan, payment, and servicing systems. Confirm that balances, days past due, and account status use compatible definitions and reporting times.
An attractive new risk layer will not resolve a disagreement about the underlying record. Build on existing infrastructure where it works; make the connections and ownership explicit where it does not.
Then use a controlled representation of the workflow to test representative cases. A first late payment, repeated delinquency, a disputed status, and a return to current should each produce an explainable path.
This is digital twinning of a process, not data pooling. Use synthetic or appropriately protected records where possible, and test whether the rules behave as intended before changing production decisions.
Ask for a view that combines affected-loan counts, balances, delinquency ages, and cure behaviour. Require someone to explain what changed and which operational action follows.
The newly released survey covers Q2, not current-day delinquencies. Use it as a benchmark and a challenge to your reporting, not a substitute for your own current portfolio evidence.
A count can tell you how many accounts need attention. It cannot tell you everything about how much is at stake.
No. The reported amount is the remaining balance on mortgages with overdue payments, not the sum of unpaid instalments. It should not be described as realized losses either.
No. The survey counts a payment overdue at quarter-end regardless of duration. Separate delinquency-age buckets before comparing this measure with a 90-day arrears statistic.
No. Statistics Canada's coverage includes credit unions and several other non-bank lender types, including mortgage finance companies and insurers. The aggregate should not be attributed solely to private lenders.
Check the definitions and reconcile the reporting before changing credit rules. Then examine which exposures persist, which cure, and whether every escalation has an accountable owner.