Bank Jerusalem 2025: Why Provision Expense Fell While Problem Credit Rose

The main article argued that Bank Jerusalem's profit looked stronger than the core economics underneath it, which is exactly why Export's discount cannot be judged only through the bottom line. This follow-up isolates the sharpest version of that gap: the provision-expense ratio fell to 0.32% from 0.51%, but non-accrual or 90-day-past-due credit rose to 1.96% from 1.38%.

This is not a technical contradiction. It is an earnings-quality contradiction. A reader who stays with the income statement sees provision expense of NIS 52.4 million instead of NIS 79.7 million, and therefore attributable net profit of NIS 194.6 million instead of NIS 154.6 million. A reader who goes into credit quality sees something much less comfortable: the allowance balance as a share of public credit fell to 1.33% from 1.43%, even as stress inside the book increased.

The year-end exit rate already hinted at it. In the fourth quarter, credit-loss expense rose to NIS 19.1 million from NIS 16.9 million. So the annual headline looks cleaner, but the run-rate exiting the year was already moving the other way. That is exactly the kind of gap the market tends to miss when it reads earnings before it reads the book.

The Contradiction Sits At The Center Of The Filing

The core metrics table compresses the contradiction into a few brutal lines. On one side, the provision-expense ratio fell by 19 basis points, from 0.51% to 0.32%. On the other side, non-accrual or 90-day-past-due credit jumped by 58 basis points, from 1.38% to 1.96%. At the same time, the allowance ratio fell by 10 basis points, from 1.43% to 1.33%.

That chart says something simple: 2025 was not a clean-up year for the book. It was a year in which the loss estimate recognized through the P&L fell while the stock of stress already sitting in the portfolio rose. That distinction matters. Provision expense is a flow measure. Problem credit is a stock measure. When the two move in opposite directions, the key question is which one is leading and which one is lagging.

The issue is not only the direction. It is also the earnings contribution. The drop in provision expense removed NIS 27.3 million of pressure from the bottom line versus 2024. Without that move, the jump in 2025 profit would have looked materially less impressive.

Where The Book Actually Deteriorated

The deterioration is not buried in a side pocket. Total non-performing public assets rose to NIS 290.8 million from NIS 196.1 million. That is about a 48% jump. Over the same period, total public credit rose to NIS 19.10 billion from NIS 18.29 billion, only about 4.4% higher. In other words, the stress layer grew far faster than the book itself.

The split matters even more. Non-accrual housing loans rose to NIS 191.4 million from NIS 146.2 million. Non-accrual commercial credit jumped to NIS 95.3 million from NIS 44.1 million. This is no longer a narrow mortgage-only story. Commercial credit now carries much more weight inside the problem bucket.

The movement during the year is not reassuring either. Credit newly classified as non-accrual during 2025 rose to NIS 219.4 million from NIS 166.5 million. At the same time, credit that returned to accruing interest income fell to NIS 83.8 million from NIS 103.7 million. So not only did the ending stock rise, the inflow into the bucket accelerated and the outflow weakened.

Housing stress is already visible in a way that is hard to wave away. Housing loans more than 90 days past due reached a recorded balance of NIS 182.0 million versus NIS 114.8 million at the end of 2024. That is already 1.9% of total housing credit. Anyone who focuses only on the comforting message around real-estate collateral misses that the delinquency stock itself is no longer small.

There is another subtle but important detail here. Within non-accrual credit, the portion that is not yet 90 days past due rose to NIS 65.8 million from NIS 33.7 million, driven mainly by commercial credit, NIS 49.5 million versus NIS 3.2 million a year earlier. That means some of the deterioration is already being captured not only through technical delinquency, but through proactive credit classification. That is usually a sign that the bank itself is identifying weakness before it fully rolls through the payment schedule.

Why Provision Expense Still Fell

This is where the contradiction becomes economically useful. The P&L explanation is explicit: the drop in credit-loss expense during 2025 came mainly from roughly NIS 36.7 million lower group provision, primarily because balances to private individuals were smaller. Against that, specific provision increased by NIS 18.7 million and charge-off expense declined by NIS 9.5 million versus 2024.

In other words, the lower headline provision number does not mean stress fell. It means three forces worked together: a smaller retail credit base pulled down the group provision, charge-offs were more favorable, and the increase in specific provision still was not large enough to offset the full benefit of those two items.

The bank's qualitative risk review completes the picture. Management explicitly says that mortgage delinquencies and failures are trending upward, mainly because of high interest rates and the end of the payment-freeze framework. At the same time, it says write-offs remain very low because the loans are secured by residential property. That is the key sentence. It explains how the book can show more delinquency and more problem credit without immediately generating the same intensity in the income statement.

So 2025 was not a year in which risk disappeared. It was a year in which collateral protection and the accounting structure of provisioning allowed the P&L charge to stay relatively low even as stress in the book rose. That is why reading the lower provision number as proof of a cleaner portfolio turns the filing upside down.

There is another layer here. The bank also says that in 2025 it tightened classification for groups of consumer and housing borrowers where negative indicators were identified, including rejected direct debits, delinquency, or long deferral periods relative to reasonable repayment capacity, and that it expanded disclosure around troubled borrowers whose terms were modified. That matters because it shows the bank itself identified more pressure rather than managing 2025 as if the risk backdrop had become easier.

What Note 29(b) Adds To The Story

Note 29(b) matters here not because it adds more numbers, but because it sharpens the type of pressure. The total balance of loans to borrowers in financial difficulty that underwent a modification in terms rose to NIS 73.6 million from NIS 67.3 million. But the mix changed: in 2024 the story was almost entirely mortgages, while in 2025 a NIS 26.8 million commercial bucket appeared.

That matters because it shows the pressure is no longer only a leftover from mortgage deferrals and restructurings. It is also moving into a commercial layer that required modified terms. At the same time, loans that failed during the reporting year after a modification in terms actually fell to NIS 2.7 million from NIS 10.1 million. That is the intelligent counter-thesis: not every quality indicator is breaking at once, and the deterioration does not yet look like a system-wide collapse of the loans that had been reworked.

But that nuance should not blur the main point. If post-modification failures fell while the non-accrual stock and deep delinquency still rose, then the main deterioration in 2025 did not come only from one narrow path of modified loans later breaking. It came from a broader stock of problem credit, especially in housing and commercial exposures, that migrated deeper into problem status.

You can see that in the aging of the non-accrual stock as well. Total non-accrual balances older than one year rose to NIS 84.8 million from NIS 56.2 million. In housing alone, that amount rose to NIS 60.5 million from NIS 44.3 million. This is not just a one-quarter tremor.

What Has To Happen Next For The Gap To Close

The first 2026 checkpoint is straightforward: does the provision-expense ratio begin to move back toward what the credit-quality data is already signaling. If the non-accrual stock and deep-delinquency stock stay elevated, it will be hard to hold a 0.32% expense ratio and a falling allowance ratio for very long.

The second checkpoint is housing. After the move to NIS 182.0 million of housing loans more than 90 days past due, the market will need to see stabilization or at least moderation. Otherwise, the argument that collateral is enough to cushion the damage will face a tougher test.

The third checkpoint is commercial credit. The jump to NIS 95.3 million of non-accrual commercial exposure and the appearance of commercial loans modified in terms mean the story is no longer local to mortgages. If commercial stress keeps worsening, it will be much harder to argue that the issue is only a temporary residue of the end of the payment-freeze regime.

This matters especially for Export. The main article argued that the market gives too much weight to the bank's profit and not enough weight to the quality of that profit. This follow-up says the gap is not theoretical. In 2025 it is already visible in the numbers. Profit rose, but it rose in a year when stress indicators in the book climbed faster than the expense recognized against them. That is exactly the earnings-quality issue the market will have to unpack in the next few reports.