**Richard Olswang, John Jenkins, Tom Bulpitt and Darren Clay discuss the theoretical and practical challenges posed by IFRS 17 discount rates**

IFRS 17 is a principles-based standard that requires significant interpretation before it can be implemented in practice. A key consideration is the discount rate to be used in measuring liabilities, among other related financial assumptions. The IFoA’s IFRS 17: Future of Discount Rates Working Party was established to provide thought leadership in these areas. In 2021, as part of the Current Issues in Life Assurance (CILA) webinar series, the working party provided an overview of the requirements and implications of IFRS 17 relating to discount rates, in particular for annuity business. We share some of these thoughts and describe the results of several polling questions that were put to webinar participants (Figure 1).

**IFRS 17 requirements**

IFRS 17 sets out the requirement to discount future cash flows in deriving the value of the liabilities. Paragraph 36 sets out the requirement to discount and (importantly) states that the discount rates shall reflect the characteristics of the cash flows and (very importantly) the liquidity characteristics of the insurance contracts. The discount rates should be consistent with observable market prices of financial instruments with cash flow characteristics consistent with the insurance contracts in question. Boiled down, we need a market-consistent discount rate, but we are allowed to have an illiquidity premium for illiquid liabilities.

**The key paragraphs in the application guidance are:**

- B72 to B79 – these set out the various uses of discount rates throughout the IFRS 17 calculation and disclosure processes, and cover a number of practical aspects
- B80 and B81 – these respectively refer to the bottom-up and top-down approaches for arriving at the discount rates.

Under the bottom-up approach, we add on an illiquidity premium to the risk-free rate (RFR) to reflect the illiquidity characteristics of the insurance contracts. Under the top-down approach, we start with the full market value yield on the relevant assets and make a deduction for the credit risk component of the yield, but make no deduction in respect of illiquidity. In theory, both methods give rise to the same result – a discount rate that is risk-free but allows for the illiquidity of our cash flows.

It is possible to use a hybrid approach involving some bottom-up and some top-down features, such as the Solvency II matching adjustment and fundamental spread approach. Interestingly, IFRS 17 says nothing about the definition of the RFR, and very little about the choice of assets to use in the top-down method. For this, one has to identify a ‘reference portfolio’, but this does not need to be the actual assets held. This introduces a further complexity – or, if you prefer, a further flexibility.

In other words, IFRS 17 sets out some good principles and guidance, but does not really help us with the hard part, which is to divide up the spread on (say) corporate bond investments into the component that is due to their relative illiquidity and the component that is due to genuine credit risk. This issue has been the subject of much actuarial and financial analysis, which will no doubt continue. Top-down is probably the more common approach to date, and various studies are available to assess the necessary credit risk deductions. Bottom-up is currently less common, but approaches for this do exist – for example by considering the yield on a collateralised covered bond, which can in practice be regarded as being risk-free but benefiting from an illiquidity premium.

A particularly interesting point is that in recent years (and particularly since Solvency II came in on 1 January 2016), it has been regulators who have made the running in this area. For Solvency II, the European Insurance and Occupational Pensions Authority defines the parameters for the volatility adjustment and the matching adjustment. The Prudential Regulation Authority will do so in future now that we are post Brexit (and has announced its plans to review UK Solvency II), and a new version of the UK RFR came in from 31 July 2021. The International Association of Insurance Supervisors has now set out and is testing its Insurance Capital Standard regime, which defines the illiquidity and credit risk parameters under a three-bucket approach. Other than for internal purposes, and for transactions, insurers have not really had to devise their own approach for these critical issues – they have just followed the regulatory requirements.

All this changes under IFRS 17. Each insurer has to determine its own approach, make the relevant judgments and get it all agreed with its auditors – in the knowledge that analysts will ask probing questions and compare companies’ approaches, and that the approach will drive key published and audited results. Not easy! In deciding how to proceed, the key questions facing insurers are:

- Whether to use bottom-up, top-down, or vary by product class
- Whether to work from a regulatory approach or use a completely separate approach.

Simplicity versus complexity will figure highly in the equation, and it will inevitably be necessary to bridge from the regulatory to the IFRS 17 position. Even if the approach is based on the regulatory position, the insurer will still need to demonstrate that it complies with IFRS 17.

**Polling at the CILA webinar indicated that:**

- For annuities, the most common planned approach is top-down, probably reflecting the fact that annuity business is particularly sensitive to the discount rate, and insurers need to use the optimal discount rate that can be justified.
- For other business, the most common approach is bottom-up, likely reflecting the fact that, generally, non-annuity business is less sensitive to discount rates and a simpler approach is desirable.

**Key practical challenges**

Many approaches to deriving discount rates have been taken in the past, but one thing that they all seem to have in common is a link between the discount rate and the assets backing the liabilities. It is usually assumed that the insurer expects to earn all or part of the return on the actual assets backing the liabilities, or that the backing asset portfolio is sufficiently similar to some defined reference portfolio (as is the case with the volatility adjustment under Solvency II).

Of course, this is the purpose of discounting liabilities in the first place. Since the insurer expects to earn asset returns, it does not need to hold the full amount of the liability cash flows – they can be discounted. The question then becomes: how much can one reasonably expect to earn?

In a purist, risk-neutral framework, the assumption is that one can only expect, on a best estimate basis, to earn risk-free rates of return.

If we look at some well-known approaches to discount rates:

- Solvency I in the UK allowed for a discount rate equal to 97.5% of the risk-adjusted return on the actual backing asset portfolio
- Solvency II adopts a risk-free plus approach where the ‘plus’ is either a matching adjustment (based on a subset of the actual portfolio, separately managed) or volatility adjustment (based on a defined reference portfolio)
- Market Consistent Embedded Values/QIS5 adopted a risk-free plus approach where the ‘plus’ was the spread on the actual assets held, minus a fixed proportion for credit risk.

**What’s so hard about IFRS 17?**

Under IFRS 17, the requirement is that the discount rates reflect the characteristics of the liabilities. There is no reference to assets or the actual portfolio held, no reference to returns on assets or the amount of that return one may expect to earn. The key question is: how does one determine the characteristics of the liabilities and, in particular, the liquidity characteristics?

The following five-step process could provide a possible approach.

Note that this process is necessarily subjective, and expert judgment plays a key role. Nevertheless, setting out an overarching framework enables consistent and repeatable application of these judgments.

- Derive a set of liability characteristics and assign a score to each one based on the expert view of how indicative it is of the liquidity of a contract. Some possible characteristics are:
- Ability of the policyholder to lapse their contract (in full or in part)and/or existence of any penalties
- Valuable features of the insurance contract(for example any attractive options or guarantees)
- Contract term
- Tax incentives
- Biometric risks (for example the ability to lapse and repurchase being economically prohibitive).

- Define a set of liquidity buckets and, for each bucket, set a score range.
- For each homogeneous risk group or group of similar contracts, calculate the score based on the characteristics of those contracts.
- Allocate each group to a bucket using its score.
- Either:
- Derive an illiquidity premium based on the portfolio of assets actually held and then apply a proportion of it to each bucket (for example 0%, 25%, 50%, 75%, 100% or 100%+ of liquidity premium), or
- For each bucket, define a reference portfolio of hypothetical assets and take 100% of that portfolio’s illiquidity premium.

Clearly, one area of expert judgment here is how much of the illiquidity premium to take for each bucket. One aspect to consider is the relative liquidity of the assets and the liabilities. Specifically, it could be argued that certain insurance contracts (such as annuities) are more liquid than many assets, and so taking more than 100% of the asset-derived illiquidity premium could be a defendable position. Although this could be considered imprudent, and may lead to some undesirable dynamics, it may provide justification for not taking any less than 100%.

**Implications for annuity business**

Taking a key product line such as annuities as an example, what does this mean in practice? Including an illiquidity premium in the discount rate leads to a lower best-estimate liability, but this is offset by a higher contractual service margin (CSM). Although the discount rate does not have to be related to own assets, a mismatch between the two will lead to ongoing profits or losses in finance income and expenses. This profit or loss may not be matched by the profile of the release of the additional CSM, thereby leading to volatility in the income statement.

Of the three modelling approaches under IFRS 17, annuities are likely to be accounted for under the general measurement model, where the discount rates used to calculate the CSM are locked in at inception. Assuming assets are valued at fair value through profit and loss (P&L), the movements in assets and liabilities are presented as in Figure 2.

A mismatch between locked-in rates and actual asset returns results in an impact on the P&L of (A) – (B), which can introduce volatility for business that is sensitive to discount rates.

A related issue is the treatment of the broader set of financial assumptions in addition to the discount rate. IFRS 17 is clear that the effect of financial risk and changes in financial risk on fulfilment cash flows should not adjust the CSM (paragraph B97). However, for changes in non-financial risk that do adjust the CSM, it only specifies that they should be measured using locked-in discount rates (paragraph B96c), and does not explicitly refer to other financial assumptions. This gives rise to differing views – for example on the treatment of inflation when adjusting the CSM of index-linked annuities for non-financial assumption changes (such as longevity).

The three possible views are:

- Use current inflation rates
- Use locked-in inflation rates both retrospectively and prospectively
- Lock in inflation prospectively but using the actual past inflation (in other words, base the adjustment to CSM on the actual annuity amount at the reporting date).

Polling at the CILA webinar demonstrated that a variety of approaches are being considered. Option 2 is the most complex in practice as it would require maintaining a ‘shadow’ value of benefits as if the past had emerged in line with the original locked-in assumptions. The validity of each of these options is subject to wide discussion within the insurance industry and with the major audit firms.

**Conclusion**

The approach to determining discount rates for IFRS 17 is complex and subject to significant judgment. This demonstrates the importance of how these matters are communicated between actuaries and their stakeholders, both internal and external. The IFRS 17: Future of Discount Rates Working Party has produced a variety of papers discussing these and other topics in more detail to support actuaries in this process. These can be accessed at bit.ly/FutureDisRates. Further papers under development will consider determination of the reference portfolio and discount rates for reinsurance contracts.

**Richard Olswang** is IFRS 17 technical lead at Prudential, and chair of the IFoA’s IFRS 17: Future of Discount Rates Working Party

**John Jenkins** is a principal with Milliman

**Tom Bulpitt **is director of capital strategy at Athora

**Darren Clay** works on the technical application of IFRS 17 for the Phoenix Group