Introduction

Expense assumptions are sometimes viewed as the least “actuarial” components of insurance liability measurement. They do reach boards and audit committees, particularly where expense overruns are visible or solvency coverage is tight. However, what may not be fully appreciated is how wide the permissible range of interpretation can be within the existing standards, and how differently peers may be applying them.

For insurers operating below efficient scale, the treatment of expenses can materially affect:

• contractual Service Margin (CSM),
• onerous contract identification,
• reported earnings volatility, and
• solvency coverage ratios.

This paper sets out observations on some approaches to expenses and scale under IFRS 17 and under solvency frameworks (Solvency II and Solvency Assessment and Management (SAM)).

My aim is not to define a single “correct” approach, but to help stakeholders understand the range of choices being made in practice, and how these compare to what they may assume is standard.

This is also not a comprehensive treatment of all expense issues, but a focused discussion on scale and sub-scale treatment.

The Problem

Consider an insurer with operational infrastructure built to service a book materially larger than its current portfolio. Its per-policy (or “unit”) expenses are high, not because of operational inefficiency, but because fixed costs are spread across a small base. This already begs meaningful questions here – is the choice to build for future scale an economically rational one, or a mis-estimation of likely future scale? Or was this scale appropriate at some point in the past, but persistently declining volumes have rendered it suboptimal?

The practical assumption-setting question is straightforward: when projecting future expenses for liability measurement, what unit cost do you use? The most appropriate answer may depend on which framework applies, what the entity’s business plan says, and which regulatory or accounting objective is being served. The answers under IFRS 17 and under solvency frameworks can legitimately diverge, because the two regimes answer different questions. Divergence within these standards might appropriately reflect relevant business model and outlook differences, but could also be a function of different internal philosophies and auditor tolerance.

IFRS 17: A Range of Defensible Approaches?

IFRS 17 requires fulfilment cash flows to include entity specific (paragraph 33), directly attributable expenses (B65), while excluding costs not directly attributable to the portfolio (B66(d)) and “abnormal amounts of wasted labour or other resources” (B66(e)). The term “directly attributable” is not defined in the standard. The Basis for Conclusions (BC113) points to IAS 2 and IFRS 15 as reference points for systematic allocation. This leaves material room for judgment, and in practice three broad approaches have emerged.

Approach 1: Project current unit costs

The most problematic position I’ve seen – which can be too conservative for a growing book, but potentially too optimistic for a declining one. Current per-policy expenses are projected forward with inflation but no assumed change from future growth or decline. This anchors itself in B65’s requirement for current estimates. For entities with a stable book, or low-confidence growth trajectory, it may be appropriate. For a genuinely scaling business or one that is running off, it risks materially misstating a best estimate of future expenses.

For a rapidly growing business, this has at least three consequences:

1. that most or all contracts may appear onerous at inception.
2. with limited or no Contractual Service Margin (CSM) recognised initially, future earnings volatility is amplified with the capitalised effect of changes in assumptions directly hitting P&L.
3. insurers may be tempted to structure reinsurance arrangements, not for optimal risk sharing or technical skills or upfront expense funding, but to recognise offsetting profits on initial recognition of the reinsurance.

For a declining business or one in explicit run-off, this has the opposite effect. Expenses may be understated relative to future actual expenses. Total expenses should decrease as the book declines, but the extent of fixed overheads that can be reduced only slowly or in a lumpy fashion can distort the picture.

Approach 2: Allow for expected economies of scale

Unit costs are projected to decline over time as the business scales. This has explicit support in the Canadian Institute of Actuaries’ Explanatory Report on IFRS 17 Expenses (revised December 2024), which states that an actuary should consider “expected future economies (or diseconomies) of scale, consistent with the likelihood of these scenarios and unbiased mean.”

The IAA’s IAN 100 reinforces this. Under this approach, all relevant current expenses remain directly attributable, but the projection reflects a declining trajectory aligned with expected business growth. Unit expenses assumptions will still be high in the short term, but the liability reflects a probability-weighted best estimate over the contract lifetime.

This is usually my starting point, but I am sympathetic to the view that the immediate creation of a large liability for ongoing expenses within fulfilment cash flows the minute a single policy is sold (and a corresponding massive income statement loss) is perhaps unhelpful.

For declining portfolios, the same dynamics apply in reverse: unit costs rise as fixed costs are spread across fewer policies. A probability-weighted best estimate should not mechanically project current fixed costs across an ever-shrinking base indefinitely. Some allowance for real unit expense increases is appropriate, but capping these at a reasonable level recognises that the book will not be administered in its current form forever. Where that cap sits, and when it applies, is a key area of judgment.

Approach 3: Exclude excess capacity costs

This goes further, treating the cost of operating below efficient scale as either not directly attributable (B66(d)) or as abnormal waste (B66(e)), drawing on the IAS 2 analogy where fixed production overheads are allocated based on “normal capacity” and unabsorbed overheads are expensed to P&L.

In practice, the proportion of expenses classified as non-attributable varies widely, from almost nothing to 40% or more, depending on business model and the mix of insurance and investment business within the entity. But there is an important distinction between expenses that are genuinely not related to insurance contracts, such as general corporate marketing, shareholder costs, or costs of managing investment contracts outside IFRS 17, and expenses that are clearly related to insurance contracts but excluded because they relate to capacity that is not yet (or any longer) utilised. The former is an allocation question. The latter is an assumption about what "directly attributable" means for a sub-scale insurer, and it is the more consequential judgment.

The abnormal waste approach has a basis in BC113’s reference to IAS 2, but requires serious justification if it is permitted at all. It involves asserting that a significant portion of actual cash expenditure on people, systems and operations is irrelevant to the measurement of insurance liabilities, and requires an objective definition of “normal capacity” for an insurer, which is considerably harder to establish than for a factory.

This approach results in regular expense overruns in the income statement (whether or not allocated as actual insurance expenses or further down in the income statement).

How to square this range of approaches

Each approach has at least some support through specific interpretations of the standard. All require significant judgment and interpretation of the principles based standard. The range between them can be uncomfortably large, particularly for the CSM, for onerous contract identification, and for earnings quality overall.

In my experience in South Africa, external auditors have appeared somewhat open to some version of Approach 3 (excluding excess capacity costs), particularly for entities with clear growth trajectories or planned transfers of business.

Anecdotally, this does not appear to be as widely accepted in other markets. Whether this reflects sample error in my observations, a genuine difference in interpretation, or simply different facts and circumstances is difficult to say. It is a potential source of cross-jurisdictional inconsistency that users of financial statements should be aware of.

Solvency II: A Clearer Picture?

The solvency world approaches this from a different starting point. The primary objective is not faithful representation of expected economics but ensuring that the entity holds enough to meet its obligations. (The exit value approach of Solvency II and SAM muddies the waters here by aiming to measure solvency through a hypothetical transfer of obligations to a reference entity.)

Article 31 of the Solvency II Delegated Regulation requires the best estimate to include all administrative, investment management, claims management and acquisition expenses, including overhead allocated “in a realistic and objective manner.” Notably, Article 31 requires all of these expense categories to be included. It provides no carve-out for costs related to excess capacity, and no equivalent of IFRS 17’s distinction between directly attributable and non-directly-attributable expenses.

The conceptual underpinning reinforces this. Article 77(3) of the Solvency II Directive defines technical provisions as the amount an insurer “would be expected to require in order to take over and meet the insurance and reinsurance obligations.” The hypothetical reference undertaking that takes over the obligations is, by construction, an empty shell with no existing book over which to spread fixed costs. It would face the same scale diseconomies as the original insurer, and arguably worse ones. Stripping out overhead costs because the original insurer considers itself temporarily below efficient scale would understate this transfer value.

The going-concern assumption in Article 31, which requires expenses to be projected on the basis that the undertaking will write new business, provides some relief by allowing overhead to be shared with expected future business. But this is a projection assumption about how costs evolve, not an invitation to exclude costs that currently exist. Amendments to the Delegated Regulation effective from January 2027 tighten this further, requiring expense projections to align with Board decisions about the future of the business rather than reflexively assuming growth. The regulatory direction of travel is towards more complete expense recognition.

SAM: Following Solvency II principles with explicit guidance

In South Africa, the SAM framework follows Solvency II principles with local legislation. Prudential Standard FSI 2.2 paragraph 4.1 requires technical provisions to correspond to the value of obligations “in the event that such obligations were to be transferred immediately to another insurer”, the same transfer value concept as Solvency II Article 77(3), inviting the same conclusion.

The Guidance Note on FSI 2.2 requires that technical provisions include “all on-going maintenance expenses” and that any decision to exclude expenses must be justified against the overarching objective of providing benefits to policyholders. It says nothing about excluding expenses for lack of scale.

APN 113, issued by the Actuarial Society of South Africa, clarifies this further. It advises that expense assumptions should target zero expense variances in future years, and states explicitly that where expense overruns should be borne by in-force business, liabilities must include those overruns until they are expected to become zero. The direction is unambiguous: expense overruns from a lack of scale must be incorporated into solvency technical provisions. (Note, APN 113 is advisory and is scheduled for review in 2026.)

Despite this apparent clarity, I sometimes hear views outlining the solvency and funding impact, and deterrent to start-up insurers, of having to provide for expected future expense cash flows for a not-yet-at-scale business. This does not appear to be a settled question in practice.

Governance questions to ask

Insurers in broadly similar positions could report materially different expense outcomes under IFRS and solvency.

Does your board and audit committee fully understand:

• the range of defensible IFRS and solvency approaches, and how near the edge of your auditor’s comfort levels you are,
• the degree of judgment embedded in their chosen interpretation,
• the expectation or likelihood of negative expense variances over the coming year, or even an expense basis strengthening in the coming year,
• and how differently peers may be applying similar standards?

Even where you are comfortable with the principles and interpretations applied, how comfortable are you with the rigour of application and the clear-eyed views of future volume expectations? Does the valuation report you see consider sensitivities to crucial assumptions?

When managing a book in run-off, how have fixed expenses been treated, and are the assumptions around future sale or outsourced administration of the business been determined?

Does your Own Risk and Solvency Assessment report talk to meaningful scenarios of differences in volumes due to persistency, new business sales, and potentially mix of business?

Are the variances between prior projections adequately analysed and explained, and does it seem that there is a feedback loop to improve these estimates in future?

Conclusion

Expense assumptions embed subjective views about uncertain growth, operational scale and strategic success. They are not mechanical inputs.

Observed practice across IFRS 17 is varied, including in the treatment of sub-scale operations and excess capacity. Under solvency regimes, the direction is more constrained, but optimism can still enter through growth assumptions.

The standards permit judgment. They do not remove the need for scepticism.

For boards and audit committees, the key question is not only whether the chosen approach is technically defensible, but also whether it has been:

• stress-tested,
• monitored through robust actual versus expected analysis,
• documented against the relevant standards,
• and debated with full awareness of alternative interpretations.

Only where these disciplines are present can the divergence between and within frameworks be understood and managed.