Arrears are the last number to move in a mortgage renewal wave

Every quarter since the renewal wave began, I have watched directors take comfort from a mortgage arrears rate that barely moves. I understand why. Canadian residential mortgages have historically had very low arrears, and a ratio measured in hundredths of a per cent looks like evidence that households are coping. My position is that during a wave of renewals at much higher rates, the arrears rate is the last number to move and the least useful one to watch. A bank should build a register of payment strain from its own loan and deposit data, use it to reach borrowers months before they renew, and let it inform loan-loss provisions. It should also refuse to use that register to set prices.

Why the arrears rate stays calm

A household facing a renewal payment several hundred dollars a month higher rarely misses a mortgage payment first. It stops saving. Then it draws down savings, carries a balance on a credit card it used to clear, postpones a car repair and, sometimes, borrows from relatives. The mortgage tends to be the last obligation a Canadian household lets slip, partly because the consequences are severe and partly because many borrowers still have equity to draw on. By the time a mortgage is ninety days in arrears, the household has usually been under strain for a year or more, and the bank's options for helping have narrowed.

At Aubreuil, about 51,200 mortgages worth C$10.8 billion renew in fiscal 2025 and 2026. Most were five-year fixed terms signed in 2020 and 2021. At contractual amortisation, the average monthly payment rises by about 19% at renewal, and for one borrower in eight by more than 30%. We also hold a smaller book of variable-rate mortgages with fixed payments, whose amortisation lengthened as rates rose and which return to their contractual schedule at renewal. None of this appears in an arrears report until it is too late to be useful.

What the register reads instead

In 2024 my office built what we call the renewal register. For every mortgage renewing within eighteen months, it calculates the payment at current rates and that payment as a share of the income we can observe. For the 58% of mortgage holders whose main chequing account is with us, it adds three behavioural signals: whether regular payroll deposits have fallen or stopped, whether balances have trended down for six months, and whether card balances that used to be cleared are now carried. No signal is decisive alone. Together they identified about 6,800 households under significant strain, long before any of them fell behind.

We used the register for two purposes. The first was outreach. Our mortgage advisers called 4,900 of those households between six and nine months before maturity to discuss term choices, amortisation within our policy limits, payment dates and, where appropriate, hardship arrangements. About 1,620 made a change. The second was provisioning. Under IFRS 9, a lender recognises lifetime expected credit losses on a loan whose credit risk has increased significantly since it was made. Our credit risk team concluded that strain identified by the register was relevant evidence of such an increase, and our model validation team reviewed the method before it was used. Stage 2 balances rose by C$1.9 billion and our allowance by C$17 million.

Our chief financial officer opposed the provisioning use at first, and his concern was reasonable. The register was new, it had not been through a credit cycle, and it would raise provisions on signals that had not yet produced losses. We agreed to apply it as a reviewed adjustment for two quarters while validation finished, and to report its performance to the audit committee. In June 2026, 90-day arrears in the renewing cohort were 0.29%, against a forecast of 0.55% made before the outreach began. I cannot tell the board how much of that difference came from our calls rather than from employment holding up better than expected. I can tell them that the households we called later needed hardship arrangements about three times as often as other renewing borrowers, which suggests the register was reading something real.

Where the data stops and the conversation starts

The register has an obvious limit. For the 42% of mortgage holders who keep their main account elsewhere, we see only the mortgage. We can estimate their payment shock but not their behaviour. My first design accepted that gap, and I would not accept it now. A renewal is a natural moment to ask a borrower plainly whether their income has changed, and to explain why we are asking. We began doing so in mid-2025 and found that borrowers outside our deposit base were as likely to be under strain as those within it. We had simply been unable to see them.

The more contested limit concerns use. When the register reached our executive committee, our head of personal banking asked whether it could also guide renewal pricing, offering sharper rates to customers likely to move lenders and less to those whose strain made moving difficult. I understand the commercial logic. A borrower under strain is less likely to qualify elsewhere and less likely to shop around. Our privacy office and I opposed it, and our chief executive agreed. Customers gave us their deposit data so that we could run their accounts, not so that we could work out how little they would be able to bargain. Pricing on signs of hardship would turn a tool built to help people into a charge on the same people, and I do not believe any board could defend that to the customers concerned.

The decision cost something. Renewal retention in the cohort ran about two points below the plan the personal banking team had built with risk-based pricing in it. I would make the same decision again.

I would put three questions to any Canadian lender's board about its renewal book. How many borrowers renewing in the next eighteen months face a payment increase above a threshold the board chooses, and what does the bank know about their income now rather than at origination? Which signals does the bank read in its own deposit data, and who approved their use? And has the strain the bank can already see reached its provisions, or is it waiting for arrears to make the case? A board that hears only the arrears rate is reading the final chapter of the story first.