Prudent versus unbiased: IFRS 17 insurance liabilities

A hidden conservative bias in the form of ‘prudent’ reserving has previously been a common feature of insurance accounting. This practice has made analysing the performance of insurance companies extremely difficult for investors.

Hidden prudence is eliminated under the new IFRS 17 and the allowance for insurance risk in measuring liabilities should be fully transparent. However, considering some recent company presentations, we wonder whether this benefit for investors will be fully realised.

It has been many years in the making, but the long-awaited reform of IFRS insurance accounting is nearly here. The first financial statements under the new IFRS 17 will be the first half year (for some, the first quarter) in 2023. Results for 2022 will initially be published under the old IFRS 4 rules but restated to IFRS 17 when presented as comparatives to the 2023 financial statements.

Over the past few months many insurers have given investor presentations about IFRS 17. These have been mainly educational in nature, aimed at helping analysts and investors to get to grips with the new standard. Few detailed numbers had been provided by the time we wrote this article, but companies have given some indications of the effects of adopting IFRS 17 and explained some of their choices in its application.

Recent investor presentations have highlighted choices available under IFRS 17

Two presentations interested us in particular – those by Zurich and Hannover Re. We learned that both companies will use the OCI option¹, and both will apply a confidence level approach to determine the risk adjustment. In contrast, Hannover Re will apply the general measurement model (commonly referred to as the building block approach)² to all its insurance liabilities, whereas Zurich will apply the simplified premium allocation approach to those contracts where it is permitted to do so – essentially short-term non-life policies.

Differences in the application of IFRS 17 are to be expected considering that the standard provides several explicit choices. Some of these will create only limited difficulties for investors, such as the option to use the simplified premium allocation approach. We think other choices will potentially cause confusion and will necessitate adjustments by investors to ensure comparability. We believe the use of OCI falls into this category.

Is describing liabilities as ‘prudent’ merely an innocuous choice language or is it something more?

However, there was one difference between the Zurich and Hannover Re presentations that is not related to an explicit choice in the standard – how they described their approach to measuring the present value of insurance liability cash flows, often referred to as the best estimate liability (BEL). Hannover Re used the word ‘prudent’ multiple times in their presentation (both in the written slides and the discussion), whereas Zurich did not use this expression at all, apart from in the context of removing the effects of “prudency” from their pre-IFRS 17 measurement of insurance liabilities. Of course, this could simply be a different choice of language and nothing of substance to concern investors, but it did attract our attention and prompt us to consider the effect of prudence or conservatism in insurance accounting and specifically the measurement requirements of IFRS 17.

Measuring uncertain insurance liabilities

Insurance claims are uncertain. For a portfolio of insurance contracts, claims are more predictable given the law of large numbers, but, even for a portfolio, expectations represent a distribution of potential outcomes rather than a precise forecast. However, it is clearly not possible to represent an insurance liability as a distribution in the balance sheet – a single number is required.

The question is which number should be chosen? Should it be one that reflects the so-called unbiased best estimate, in other words the probability weighted expected claims and, therefore, the mid-point of the distribution (assuming the distribution is symmetrical)? Or should it be at a higher (more prudent) point in the distribution than the probability weighted average?

Under current insurance accounting, companies commonly measure insurance liabilities at an amount that exceeds the expected cash flows that will actually be paid to settle the obligations. This ‘padding’ of the liability, or added conservatism or prudence³ (as many would describe it) in measurement, is often evident in the inevitable reversals of liabilities that most likely occur when claims are finally settled. Conservative measurement (or ‘reserving’) results in more experience gains than losses when liabilities are paid. Lower profit initially is offset by higher profit later.

Hidden prudence is replaced by an explicit risk adjustment in IFRS 17

Analysing such ‘gains’ due to prior conservatism and trying to establish the extent of the conservatism applied, and hence the potential for further gains in the future, has been an important component of analysing insurance companies. The problem is that the degree of conservatism applied in measuring insurance liabilities was not transparent under the old IFRS 4, which made this analysis, and therefore the forecasting of results, difficult.

In our recent article ‘IFRS 17 insurance – more comparability and new insights’, we explained some of the benefits to investors of the new approach to insurance accounting. One such benefit is that the prior approach of ‘prudent’ reserving, where the degree of prudence is impossible to identify, is replaced by an explicit risk adjustment which forms one of the ‘building block’ components of insurance liabilities.

Under the IFRS 17 building block approach an insurance liability is reported as the sum of:

A best estimate liability equal to the present value of expected fulfilment cash flows (future premiums less claims, benefits, and expenses).

A risk adjustment equal to the amount the insurer charges for accepting insurance risk.

A contractual service margin which is the (risk adjusted) unearned profit.

Summary of the building block components of an insurance liability

IFRS 17 and The Footnotes Analyst

It is in the measurement of the best estimate liability (BEL) that the issue of prudence applies.

Click here to read a more detailed explanation of the IFRS 17 building block approach (and the PAA and VFA variations).

Prudent versus unbiased

The word ‘prudence’ has a troubled history in accounting⁴. At various times it has been included in, and then excluded from, the conceptual framework for IFRS reporting. The problem is that the word is commonly used in practice to depict a conservative bias in measurement. However, the actual use in IFRS reporting is that prudence means being careful and considered when applying accounting standards and measuring assets and liabilities in the way the standard intends, but not biased. Indeed, the concept of prudence is included in the current IFRS framework alongside a requirement to be ‘neutral’, in other words unbiased, when applying standards.

Neutrality in financial reporting does not rule out a conservative outcome when specific accounting standards are applied. Different recognition and measurement approaches can be applied to almost any transaction, and a specific accounting standard may well require the more conservative approach. Indeed IFRS 17 itself can be regarded as more conservative than the previous IFRS 4 accounting because it does not permit the recognition of day 1 gains.

Unbiased estimates make unearned profit and value of new business measures more meaningful

IFRS 17 requires that the fulfilment cash flows included in the best estimate liability must be unbiased probability weighted expected values. This component of the total insurance liability should not depend on management policies and should therefore, in theory, be fully comparable between companies. In particular, the amount should not be affected by the insurer’s policies regarding prudence or conservatism, sometimes previously referred to as the ‘strength’ of reserving.

The reason IFRS 17 requires that the best estimate liability should be an unbiased estimate is to ensure that data is comparable and understandable, and to facilitate the analysis by investors of the value of new business. An unbiased best estimate liability results in an unbiased unearned profit (CSM) and the correct recognition of that profit in the periods the insurance services are provided,

Under prior accounting the strength of reserving (the upward bias) was often seen as a sign of the financial strength of insurance companies. Under IFRS 17, a biased present value of fulfilment cash flows will be viewed as a potential compliance issue that will draw the attention of auditors and regulators.

IFRS 17 emphasises that the best estimate liability should be an unbiased probability weighted expected value

We think the unbiased requirement for the BEL is very clear in IFRS 17: the word ‘unbiased’ is used multiple times in the standard, whereas the word ‘prudent’ is not used at all. It is for this reason that we were surprised to hear the word prudence used so often in the Hannover Re IFRS 17 presentation.

Unbiased insurance cash flows

The Footnotes Analyst

While it is possible to be both prudent and unbiased at the same time (if prudence is defined in the manner explained in the IFRS framework), we think that, in practice, when prudence is mentioned by companies, in all likelihood what is actually meant is a bias. This is certainly how we interpreted the Hannover Re comments. Indeed, in one of the presentation slides there is a reference to an additional ‘layer of prudence’ in measuring the BEL in addition to the risk adjustment.

Of course, in the specific case of Hannover Re, we may be over-interpreting their choice of language. However, we do think that the prior practice of hidden prudence may be difficult for some insurers to give up. Hopefully the auditors and regulators will ensure the faithful implementation of IFRS 17, and investors will receive the full benefit of more transparent information about insurance liabilities and performance.

There is an argument that, because insurance liabilities are uncertain, there may be a range of equally plausible estimates of a best estimate liability and that consequently selecting an amount at the ‘prudent’ end of the range is acceptable. Indeed IFRS 17 in paragraph B54⁵ specifically refers to such a range. However, in our view, if insurers systematically select an amount at the conservative end of the range, the outcome would not comply with the requirement to be unbiased.

(Nearly) everything eventually pulls to cash

Remember that conservatism in measurement only affects the timing of profit recognition. Almost all gains and losses reported in profit and loss eventually ‘pull to cash’. Over the full life of a transaction the total profit recognised is the difference between cash in and cash out, and this cannot be affected by the intervening accounting that is used to derive profit in quarterly or annual reporting periods. However, for many insurance contracts the final pull to cash can be many years or even decades after the insurance contract was written. It is for this reason that measurement, including the question of prudence, is so important in analysing insurance companies.

Insights for investors

Recent company presentations on the adoption of IFRS 17 in 2023 show some of the likely effects on financial statements and illustrate how companies may make different choices when presented with options under the standard.

The insurance liability is measured using a building block approach that comprises a best estimate liability, a risk adjustment and a balancing contractual service margin set to ensure no day 1 profit is recognised. This leads to a more conservative pattern of profit recognition compared with previous common practice.

IFRS 17 requires that the best estimate liability is a probability weighted expected value and should not include a conservative (‘prudent’) bias. This represents a significant change from prior common practice.

The issue of prudence also applies where liabilities are measured using the simplified premium allocation approach or the variable fee approach (where a different approach to CSM accretion is used). But both are based on the building block approach.

Investors should look out for evidence that the best estimate liability includes a conservative bias. If this is the case, the analysis and comparison of performance may be more challenging.

Appendix: Insurance liabilities – the IFRS 17 building block approach

The foundation of IFRS 17 is the general measurement model, also commonly referred to as the building block approach (BBA). Although you will also see insurance liabilities measured under the premium allocation and variable fee approaches, these are a simplification and a slight variation, respectively, of the general approach (more on this later). Our comments about prudent versus unbiased estimates apply to all three approaches.

In the building block approach, insurance liabilities are the sum of a current value measurement of expected fulfilment cash flows plus the unearned profit which comprises a risk adjustment and a contractual service margin.

Present value of expected fulfilment cash flows

The first component in the building block approach is the present value of all expected cash flows related to an insurance contract, including forecast premium receipts, contract acquisition costs, claim and benefit payments, and expenses arising from processing claims and managing the liabilities. These cash flows are collectively called ‘fulfilment cash flows’ and their present value is commonly called the ‘best estimate liability’ (BEL).

Cash flows must be unbiased probability weighted expected values

IFRS 17 requires that the BEL cash flows must be unbiased expected values. This means that cash flows are probability weighted to allow for potential variation in claims and that there should be no upward bias for prudence or conservatism.

The discount rate applied in measuring the present value of expected cash flows must reflect the characteristics of the liability. Only if the cash flows are themselves dependent on asset returns should the discount rate include asset risk. For example, if a liability is simply to pass on the returns earned on a portfolio of risky assets, the discount rate applied to the estimated cash flows would be the expected return on those assets. In fact, in this case no explicit discounting is necessary at all because this present value must equal the value of the linked assets.⁶

In practice, asset linkage for insurance contracts is never complete due to the insurance related cash flows and the sharing of asset returns between policyholder and insurer. This can result in complex calculations and stochastic modelling to obtain the required present value.

If there is no asset linkage the discount rate should not include a premium for asset risk. However, IFRS 17 requires that the liquidity characteristics of insurance contracts should be considered. This means that for most non-participating insurance contracts the discount rate is an ‘illiquid’ risk-free rate. The problem is that this rate cannot be directly observed because illiquid, yet risk-free, investments are very hard to come by. As a result, the (il)liquidity premium must be estimated and there will no doubt be variations between companies. However, the requirement to disclose not just discount rates but the full yield-curve used should, in time, promote comparability.

Risk adjustment

The risk adjustment is the amount the company charges the policyholder to accept insurance risk. There is no theoretically correct way to calculate a risk adjustment. A variety of techniques could be used and, for each, a variety of assumptions. Therefore, risk adjustments are entity specific and not designed to be comparable between companies.

Risk adjustment is not comparable between companies, but it is transparent

The sum of the present value of expected cash flows and the risk adjustment is called the fulfilment value of a contract. In effect, fulfilment value is a conservative risk adjusted measure of the present value of expected fulfilment cash flows, much like the prudent reserving that may have been previously used. However, the conservative bias inherent in the fulfilment value is explicit and visible to investors. The risk adjustment and details of its calculation, including a summary of what confidence level the adjustment represents, are separately disclosed.

Because the best estimate liability includes all future cash flows measured on a probability weighted basis, the risk adjustment represents part of the expected unearned profit from an insurance contract. As risk declines over the life of a contract so the risk adjustment falls and is released to profit and loss.

You will likely hear a risk adjustment described in one of two ways:

An adjustment to the best estimate liability to reflect risk, and therefore not a part of unearned profit. The contractual service margin (see below) is therefore a risk-adjusted measure of unearned profit.

A component of unearned profit, albeit one that is run off as the entity is released from risk, rather than when insurance related services are provided.

The different view will affect how the ‘value of new business is calculated’ and performance analysed (but this is not something we will consider further in this article).

Contractual service margin (CSM)

The CSM is the expected profit from an insurance contract in excess of that represented by the risk adjustment. It represents the difference between the initial cash flow arising at contract inception, such as the initial premium received, and the fulfilment value calculated as described above.

For example, if the initial premium received is 100 and the risk adjusted present value of expected fulfilment cash flows is a net outflow of 60, the expected profit (CSM) is 40. The CSM is a component of the liability, which is therefore measured initially at the initial premium received of 100.

CSM is set to ensure no initial profit is recognised

When an insurance liability is first recognised the CSM is set to an amount that ensures there is no immediate gain or loss. This is one of the key differences from prior accounting, where some insurance products were commonly measured in a way that produced day 1 gains. When IFRS 17 is first applied you will likely see a reduction in shareholders’ equity for many companies. A key contributing factor to this will be the reversal of past day 1 gains due to retrospective application of the new standard.

The CSM can never be negative and if the fulfilment liability exceeds the initial premium received then a contract is ‘onerous’, and the loss is recognised immediately. Onerous contract liabilities are determined for groups of contracts⁷ and not individually and should be relatively rare at initial recognition. However, subsequently, changes in estimates may erode the CSM and result in an onerous contract loss.

IFRS 17 building block approach – an illustration

Assume a policyholder pays a premium of 1,000 for a 10-year life policy that will pay out 50,000 in the event of death during the coverage period. The insurer estimates the probability of death in the first year of coverage to be 0.2% (after allowing for policyholder age, health, etc.) which means that the expected cash outflow in that period is 100. Similar calculations are done for each subsequent period and the resulting expected cash outflows discounted to a present value that equals (let’s assume) 920. This calculation will generally be done on a portfolio basis; however, the probability weighted approach means that the sum of the individual contract cash flows must equal those for a portfolio.

The liability for the present value of expected future cash outflows is increased by the risk adjustment. This takes account of the extent of insurance risk – the degree to which the outcome could vary – after allowing for the diversification effects of writing multiple policies. However, the magnitude is also determined by the insurer’s attitude to risk and how it prices that risk. Let’s assume the risk adjustment in this example is 30. The sum of the present value of expected cash flows and the risk adjustment (920 + 30 = 950) is called the fulfilment value.

The contractual service margin is now the balancing figure. The insurer has received a premium of 1,000 and recognises a fulfilment value liability of 950. The CSM is therefore 50 and the total insurance contract liability including the CSM balance is 1,000. At initial recognition IFRS 17 effectively applies a ‘cost’ approach with the liability (or sometimes asset) equal to the initial cash flow – the receipt of 1,000 in this case.

BBA versus PAA and VFA

The building block approach is called the general measurement model in IFRS 17 because it is the basis for the whole standard. However, you may find that the pure BBA is only applied to a minority of insurance contracts; expect to see extensive use of the premium allocation approach (PAA) and variable fee approach (VFA).

Premium allocation approach (PAA)

The premium allocation approach is a simplified version of the building block approach that may be applied to certain (usually Property & Casualty) short duration contracts. The insurance liability is initially measured at the same amount as for the BBA (no gain at initial recognition) but this amount is not separately analysed into the three BBA components. Consequently, there is no remeasurement of these components and the premiums received are simply allocated to revenue over the coverage period.

Once claims are recognised and insurance coverage has ended, the PAA reverts to being the same as the BBA, with the liability for incurred claims measured and presented as the sum of the best estimate liability and a risk adjustment (there is no CSM at this point).

Our comments about prudent versus unbiased measurement apply to the best estimate liability for incurred claims in the PAA in exactly the same manner as for the BBA.

Variable fee approach (VFA)

The variable fee approach is a modification of the BBA that must be used for qualifying participating contracts. The modification affects how the CSM is remeasured after initial recognition and does not affect the other components of the liability. Under the BBA the CSM is accreted for the time value of money at the discount rate applicable at the initial recognition of the contract. However, under the VFA the CSM is remeasured to allow for the effect of changes in the value of the insurance company’s interest in the underlying assets of the contract. This more comprehensive remeasurement smoothes out the effect of changes in the value of the insurance company’s interest in those assets.

The issue of prudence applies to the best estimate liability of the VFA in the same manner as for the BBA. However, prudence is perhaps more relevant in the measurement of non-participating contracts (particularly P&C) than to the participating contracts that may be allocated to the VFA.

Click here to return to the main article

Don’t miss new articles published by The Footnotes Analyst – subscribe here:

No advertising, no spam, just new articles. Unsubscribe anytime.

Send us a question or comment about this article:

For more articles select subject tags:

Or see suggested related articles below:

Cookie	Duration	Description
cookielawinfo-checkbox-analytics	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Analytics".
cookielawinfo-checkbox-functional	11 months	The cookie is set by GDPR cookie consent to record the user consent for the cookies in the category "Functional".
cookielawinfo-checkbox-necessary	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookies is used to store the user consent for the cookies in the category "Necessary".
cookielawinfo-checkbox-performance	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Performance".
viewed_cookie_policy	11 months	The cookie is set by the GDPR Cookie Consent plugin and is used to store whether or not user has consented to the use of cookies. It does not store any personal data.