IAS 37 Explained: Provisions and Contingent Liabilities

Every provision begins with a contradiction: the amount or timing is uncertain, but the financial statements still need an answer at the reporting date.

Wait until payment is certain, and liabilities may be understated. Recognise too freely, and a provision becomes a reserve that management can build and release to reshape profit.

IAS 37 is built around that tension. Its categories determine whether uncertainty is recognised, disclosed or left outside the financial statements. This article explains where that structure came from, how it works and where judgment still matters.

Two metaphors worth knowing

Two recurring terms make the risk easier to see:

• Big bath: The expression draws on the English idiom “take a bath,” meaning to suffer a large financial loss. In accounting, it describes concentrating losses and charges in an already poor year—absorbing the damage at once and making later performance easier to improve. “Taking a bath” appeared in Paul Healy’s 1985 study of bonus schemes and accounting decisions.

• Cookie-jar reserves: The image is of putting reserves into a jar in good years and taking them out when earnings need support. In 1998, SEC Chairman Arthur Levitt placed cookie-jar reserves alongside big-bath charges in his influential speech, “The Numbers Game.”

The two techniques move profit in different directions: a big bath pulls losses into the present, while a cookie-jar reserve stores profit for later.

Historical background

Prudence and creditor protection

The idea behind provisions predates IAS 37 by more than a century.

As limited liability companies spread in nineteenth-century Britain, shareholders gained protection from claims against their personal assets. Accounting and company law then had to determine how much profit could safely be distributed without weakening the assets available to creditors.

Traditional prudence responded asymmetrically: foreseeable losses were reflected before final settlement, while gains were generally deferred until realised. Provisions became one way to charge uncertain obligations against profit before payment.

Modern IFRS uses prudence differently. The Conceptual Framework describes it as caution under uncertainty in support of neutrality—not a general rule to recognise liabilities earlier than assets. IAS 37 nevertheless retains an asymmetry: a probable outflow may support a provision, whereas a probable inflow generally leads only to disclosure. The asset is recognised when the inflow becomes virtually certain.

That discretion also created risk: uncertain losses could be overstated, building reserves that management might release in later periods. Against this background, the IASC moved the accounting for contingencies from IAS 10, issued in 1978, into a dedicated standard. IAS 37 was issued in September 1998, with a more explicit sequence for identifying, recognising, measuring and reassessing uncertain obligations.

How IAS 37 is structured

The table below maps the complete standard. Its purpose is to demonstrate coverage, not to reproduce every requirement.

Scope and definitions: what stays inside IAS 37?

IAS 37 is a residual standard. It applies only after items addressed more specifically elsewhere have been excluded. Examples include IAS 32 and IFRS 9 for financial instruments, IAS 12 for income taxes, IFRS 16 for leases, IAS 19 for employee benefits, IFRS 17 for insurance contracts, IFRS 3 for an acquirer’s contingent consideration and IFRS 15 for contracts with customers. IAS 37 still supplies the onerous-contract model that IFRS 15 does not contain.

Ordinary liabilities are also separated from provisions. Trade payables relate to goods or services already received and invoiced or formally agreed. Accruals also relate to goods or services already received, but payment, invoicing or formal agreement remains outstanding. Some estimation may be required, but the uncertainty is generally lower than for a provision. These items do not enter the IAS 37 provision model.

The scope analysis therefore works as a filter:

• Payable or accrual: recognise an ordinary liability under the applicable requirements.

• Provision: recognise an uncertain present obligation when the IAS 37 criteria are met.

• Contingent liability: do not recognise it; disclose it unless the possibility of an outflow is remote.

• No present or possible obligation: recognise and disclose nothing.

Expenditure that the entity can avoid through its future conduct is normally a future business cost rather than a present obligation.

Probability language: a hierarchy, not a percentage scale

Across IFRS Accounting Standards, probability terms broadly form the following order. Only “probable” has a general numerical boundary.

This is a qualitative hierarchy, not an IFRS percentage conversion table. “Possible obligation” also concerns whether an obligation exists, whereas “probable” and “remote” describe the likelihood of an outflow. More detailed percentage bands used in practice are internal conventions rather than IFRS definitions.

After these exclusions, the article focuses on how the remaining boundary is applied, with recognised provisions as the main subject.

Recognition: probable is not enough

The Scope section has already shown what happens when an outflow is not probable. But a probable outflow is only one of three conditions required to recognise a provision:

• a present legal or constructive obligation arising from a past event;

• a probable outflow of resources; and

• a reliable estimate of the amount.

The first condition usually requires the most judgement. A contractual or statutory obligation may be relatively clear. A constructive obligation depends on the entity’s own conduct: established practice, published policies or a sufficiently specific statement may create a reasonable expectation that the entity will accept responsibility. An internal plan alone is not enough. The question is whether a past event has left the entity with no realistic alternative to settlement.

The final condition is rarely the deciding barrier. Uncertainty does not prevent recognition if the amount can still be estimated with sufficient reliability. Only in exceptional cases where no reliable estimate can be made does the item remain on the contingent-liability path.

Probability therefore does not determine the answer by itself. A provision is recognised only when all three conditions are present. IAS 37.14 provides the reference point.

Measurement: estimating the obligation

Once the recognition conditions are met, the question changes from whether to recognise the obligation to how much to recognise. The best estimate is the amount an entity would rationally pay to settle the obligation at the reporting date, or to transfer it to a third party at that time.

“Best estimate” does not prescribe one calculation method. For a large population, such as warranty claims, the estimate may reflect a range of weighted outcomes. For a single obligation, such as a lawsuit, the most likely outcome may be more informative, although other credible outcomes cannot simply be ignored. The method should follow the nature of the uncertainty.

Long-term provisions may also require discounting. Cash-flow assumptions, risk adjustments and the discount rate must be internally consistent so that the same uncertainty is not reflected twice. Discounting is not only an initial measurement step: as the settlement date approaches, the unwinding of the discount is recognised as a finance cost in each subsequent period.

Any expected recovery from an insurer or another party is assessed separately. If the entity remains responsible for settlement, the expected recovery does not remove or reduce the underlying obligation. A reimbursement is recognised as a separate asset only when recovery is virtually certain, and the amount recognised does not exceed the provision itself.

The estimate is updated as evidence changes, and the provision is used only for the expenditure for which it was recognised. This is also where earnings management may become visible: repeated additions without new evidence, use for unrelated costs or releases that consistently support profit may indicate that the provision is being used as a cookie-jar reserve.

Application: where the boundary becomes difficult

Onerous contracts

A contract becomes onerous when the economic benefits expected from it no longer cover the unavoidable cost of fulfilling or exiting it. A common example is a fixed-price customer project whose estimated remaining costs have risen above the revenue expected from the contract.

Once the loss becomes unavoidable, IAS 37 requires a provision measured at the lower net cost of completing the contract or exiting it. Any impairment of assets used to fulfil the contract is recognised first. This ordering prevents double counting; it does not replace the contract-by-contract assessment.

The 2020 amendment clarified the cost base used in that assessment. It includes not only incremental costs, such as direct labour and materials, but also an allocation of other costs directly related to fulfilling the contract. For entities that previously considered only incremental costs, the broader cost base may cause contracts to be identified as onerous that would not previously have crossed the threshold.

A practitioner’s perspective

The author previously served as both a project manager and a business unit manager at a systems integrator. Each quarter, he estimated the profitability of projects under his responsibility and reported whether a loss provision might be required.

For a troubled project, the arithmetic was rarely the hardest part. The greater difficulty was judging the assumptions behind it: the remaining work, staffing and subcontracting costs, delays, scope changes and amounts recoverable from the customer. A credible recovery plan had to be distinguished from the expectation that the project would somehow improve.

The important point is that the person preparing the underlying estimate is not an independent observer. I knew the project best, but recognising a forecast loss could affect not only the project’s results but also my own performance evaluation. There was no inherent assurance that the information reported upward was neutral, complete or based on a sufficiently critical outlook.

To be candid, when the business unit had room in its earnings, there was a natural incentive to recognise the loss early. When earnings were already under pressure, the instinct was the opposite: avoid recognition if possible. We would look for recovery measures, consider potential additional claims against the customer and examine whether more favourable assumptions could be supported. Accounting judgement cannot be completely separated from the position and incentives of the person making the estimate.

The judgement involved in provisions should therefore not be treated solely as a senior-management issue. Bias may enter the numbers at the project level, before they reach finance, management or the auditor. Reviewing the calculation is not enough; the underlying operational assumptions must be challenged against the actual condition of the project.

Restructuring

Restructuring is particularly exposed to big-bath incentives because weak performance, management change and strategic reset often occur together.

A plan alone is still something management can change. The key question is when implementation or communication makes reversal unrealistic for those affected. Even then, the estimate must separate the direct cost of the restructuring from investment, retraining and other costs of running the business after the change. That boundary prevents a strategic reset from becoming a container for future expenditure.

Future operating losses

Expected losses from continuing a store, factory or business line are not recognised as provisions. Unlike a loss on an existing onerous contract, these losses are not unavoidable consequences of a present obligation: the entity can change, reduce or discontinue its future operations.

Such forecasts may indicate that related assets are impaired, but the expected operating loss itself is not recognised as a liability.

IFRS and US GAAP: why the outcomes can differ

US GAAP addresses loss contingencies through several Codification Topics rather than a single standard. Differences in threshold and measurement can change both timing and amount.

Although both frameworks use the term “probable,” they do not apply the same recognition threshold. Under IAS 37, probable means “more likely than not”—that is, greater than 50%. Under US GAAP, probable is defined as “likely to occur,” without a numerical threshold. Practice literature often associates it with a likelihood of around 70–80%, but this range is not specified in authoritative US GAAP. The same word therefore represents different recognition thresholds under the two frameworks.

Conclusion

IAS 37 deals with estimates. It is not a Standard that produces an answer simply by following detailed rules. Yet this is precisely why provisions require significant judgment—and particular attention in an audit.

Documents and calculations alone are not enough to determine whether an estimate faithfully reflects the entity’s position. That assessment requires an understanding of the environment in which the entity operates and of the events that created the uncertainty.

IAS 37 links recognition, measurement, reassessment and disclosure to evidence about the obligation rather than to management’s preferred timing of profit. Applying it well therefore requires more than technical knowledge: it requires the ability to understand the business reality behind the numbers.

Practice Questions

Practice questions based on this article are available on IFRS-OneQ. Reinforce what you've read — one question at a time.

Disclaimer: This article is for educational purposes only and reflects the author's interpretation of accounting standards and their history. It does not constitute professional accounting, tax, or financial advice. Readers should consult qualified professionals on specific transactions.