A bank funds 60% of its loans from household deposits and 40% from wholesale markets that roll every three months. Wholesale rates have just jumped.
Management Overlay: a Worked Example
The model knows the past. The overlay is what the bank knows now.
BY MICHAEL SHANG · SIXTEEN YEARS IN COMMERCIAL BANKING · UPDATED
// THE SHORT ANSWER
A management overlay is the part of a bank's loan loss provision that comes from judgement rather than from its models. Expected credit loss models estimate losses from the history they were built on. When something is happening that the history does not contain, the bank adds an amount on top of the modelled figure, or occasionally takes one off. A reader of the accounts asks three questions about it: how large it is against the modelled figure, what specific risk it is for, and what would have to happen for it to come out. An overlay that cannot answer the third is the one to watch.
BASIS Practitioner judgment from sixteen years in commercial banking, set out in How Bankers Think (Highbank Press, 2026). Every borrower on this site is a composite constructed for teaching, not a client. The market is mid-market commercial lending as practised in New Zealand and Australia; where a convention differs elsewhere, the page says so.
// MAKE THE CALL
Issue 09 — The Tuesday List. Eight dashboards. Two intensive-care slots. Portfolio triage, Tuesday 8:30am. Your call first, then the senior banker's.
// 01 — WHY A MODEL NEEDS AN ADJUSTMENT
A provisioning model turns a borrower's risk grade, the security held and a set of economic scenarios into an expected loss, using relationships estimated from past defaults and recoveries. It is only as good as that past. Interest rates that rise quickly after a decade of low ones, a wave of failures in one industry, or a model already known to react slowly are all cases where the history and the present part company, and the modelled figure is likely to be too low or too high.
Supervisors watch this closely. The UK's Prudential Regulation Authority said in its September 2025 letter to the chief financial officers of the major UK banks and building societies that model risk remains elevated, because today's combination of credit risk factors probably differs from those the models were built to capture, and it asked firms to challenge the completeness of their adjustments. The vocabulary varies: overlay, post-model adjustment, management adjustment, temporary adjustment. The PRA uses post-model adjustment to cover all of them, including overrides of model output.
// 02 — HOW AN OVERLAY IS BUILT AND GOVERNED
An overlay starts with a named risk that the model misses, and a reason it misses it. It is then sized, usually by running the affected loans through the model again under a harsher assumption: moving part of a segment to stage 2, raising the default rate for a sector, or cutting the value of a type of security. Top-down estimates are used when there is no better data, and are the hardest to defend.
Good practice gives each overlay a written rationale, a sizing method, an owner, approval at a provisioning committee, a review every reporting period and a stated condition for release. It is also checked against the model, so that the same risk is not counted twice once the model starts to see it. In the accounts, the disclosure standard IFRS 7 and its New Zealand and Australian equivalents require banks to explain the inputs, assumptions and estimation techniques behind expected losses, and many banks show overlays as separate lines in the credit risk note.
// 03 — THE EXAMPLE
A lender, constructed for this page, has US$1.8bn of loans to trade subcontractors: electricians, plumbers, roofers. Several large residential builders fail within months, leaving subcontractors with unpaid progress claims. The lender's model, estimated over years in which few builders failed, has not yet downgraded a single subcontractor: the modelled figure includes older problem loans, but the fifth of the book most exposed to the failed builders sits in stage 1 at a twelve-month rate of 0.35%. Every figure below is the example's.
A year later there are three ways the overlay can leave.
In A, US$300m of loans have been downgraded and two borrowers have defaulted, so the model now carries most of the risk itself; the overlay falls to the part it still misses, and the total does not move. In B, builder failures have stopped, the subcontractors have been paid or have written off their claims, and no downgrades came: the reason has gone, and so does the overlay. C has the same numbers as B and nothing behind them. The book is unchanged, the model still sees nothing, and the US$6m comes back into profit because the year needed it. That is the overlay as a smoothing tool, and the only way to tell it from B is the evidence the bank gives for the release.
// 04 — HOW IT DIFFERS FROM A SPECIFIC PROVISION
The nearest neighbour is the individually assessed provision on a loan that is already credit-impaired. That provision belongs to one named borrower, and it is calculated from that borrower's expected cash flows and the value of its security. An overlay belongs to no one in particular. It covers losses the bank expects to arise somewhere in a segment, before it knows which borrowers will produce them. When the risk does arrive, it moves from the overlay into the model, or into specific provisions, which is why case A leaves the total unchanged.
An overlay also differs from a model override, where a credit officer changes one borrower's grade because the model's grade is wrong for that file. Both are judgement; one acts on a portfolio, the other on a borrower.
DEFINED IN ONE PARAGRAPH → · COST OF RISK, WORKED → · LOAN LOSS PROVISION →
// 05 — WHERE IT DECIDES A FILE
For a borrower in a sector under an overlay, the bank has already put a number on the sector's risk, and the credit officer's question at review is whether this borrower is the one the number is for. In the example that means the debtor ledger: how much is owed by the builders that failed, how much by those still trading, and how much retention money is at risk. A subcontractor that can show it has no exposure to the failed builders gives the bank evidence to keep its grade where it is. One that cannot may be downgraded, which is the overlay arriving in the loans.
For a reader of bank results, the overlay note repays attention: its size against modelled losses, its movement in the year, and the release condition the bank has written for it.
HOW TO READ BANK RESULTS → · THE EXAM HALL · BANK FINANCIAL ACCOUNTANT →
// SOURCES
The documents the statements on this page rest on. Each was opened and checked when cited.
- Bank of England, Prudential Regulation Authority, Thematic feedback on accounting for IFRS 9 expected credit losses (Dear CFO letter), 30 September 2025 — covering letter (model risk), annex paragraphs 6 and 13, footnote 5 — Model risk remains elevated because current credit risk factors likely differ from those models were built to capture; post-model adjustments cover model overlays, management overlays and overrides, and firms rely on them for sector risks models miss — sections 01 and 04.
// QUESTIONS PEOPLE ASK
- What is a post-model adjustment?
- It is any change a bank makes to the output of its expected credit loss models because the models do not reflect a risk or an uncertainty properly. The term covers management overlays applied to a portfolio or sector, overrides of a model's result for particular loans, and corrections for known model weaknesses. Supervisors such as the UK's Prudential Regulation Authority use it as the umbrella term and review how complete and how well governed these adjustments are.
- Why do banks release management overlays?
- For one of two good reasons. Either the risk the overlay was made for has arrived in the loans themselves, so the model and the specific provisions now carry it and the overlay would count it twice, or the risk has passed without arriving. A release without either, made because profit was short, is the poor reason. The accounts usually say which applies, and the movement in modelled losses and downgrades shows whether the explanation fits.
- Are management overlays allowed under IFRS 9?
- IFRS 9 sets no separate rules for overlays. It requires expected credit losses to reflect reasonable and supportable information about past events, current conditions and forecasts. When a model cannot reflect some of that information, an adjustment is how the bank meets the requirement, so overlays are a common part of compliance rather than an exception to it. They do have to be supportable, and auditors and supervisors test that.
- Do management overlays affect a bank's capital?
- Indirectly, yes. An overlay is a charge against profit, so it reduces retained earnings and with them common equity tier 1 capital. Releasing it does the reverse. The net effect can be smaller than the charge: a bank using its own models compares provisions with its regulatory expected loss, and a higher provision can shrink a shortfall that would otherwise be deducted from capital. How provisions count is set by each regulator's rules, which differ between the standardised and model-based approaches and between New Zealand, Australia and the UK.
THE IN-TRAY · ONE MINUTE
Make the call on one file.
A short file on the ground this page covers. Every option opens with how far it sits from a senior banker’s read.
1 FILES · PICK A CALL ON ANY OF THEM
3 MORE ON THE BANK · FUNDING AND LIQUIDITY →WHERE DO YOU READ FROM? THE PLACEMENT →
// SUBSCRIBE
The Lab.
One case file and one question, in your inbox every Tuesday morning, New Zealand time. The senior banker's read opens only after you answer.
FREE WITH IT · ANNOTATED PDF
A complete credit paper on a published file, with the approver's notes on what each section has to answer.
No spam. Unsubscribe anytime. Educational content only.
