Insight for insurance firms
The new business volumes written in the UK pension buyout/ buy-in markets in recent years include a notable growth in the volume of insured deferred member pension benefits. While insurance firms writing pension risk transfer business note the additional risks and complexity that can arise from deferred liabilities and price the business accordingly, the pricing differential observed between pensioner and deferred liabilities has narrowed over the last five years or so. These trends in market pricing and volumes indicate that insurers are increasingly confident of managing the risks associated with deferred liabilities. However, some of the key features of deferred pensions can create complications for the treatment of these liabilities under insurers’ solvency regulations, particularly the Matching Adjustment (MA). This blog discusses these challenges and their potential solutions.
Surrender or die
Pension annuities in payment are exposed to only one form of demographic risk – the time until the annuitant’s death. The cashflow timings on deferred pensions have an additional source of uncertainty. Deferred pension scheme members retain forms of surrender options when their benefits are insured by a Bulk Purchase Annuity (BPA) firm (either in buyout or buy-in).
Prior to taking their pension, deferred members can opt to receive a cash value that they can transfer into a defined contribution scheme (Transfer Value, TV); or they can opt to take up to 25% of the value of their pension as a tax-free Pension Commencement Lump Sum (PCLS) when they start receiving their pension.
These surrender options do not naturally fit into the MA framework, which is conceptually predicated on the illiquid and fixed nature of MA-eligible liabilities. However, deferred liabilities have been MA-eligible since Day 1 of Solvency II:
- The PRA Rulebook (Matching Adjustment 2.2 (4)(b)) states that a liability with a surrender option where the surrender value does not exceed the value of the backing assets may be MA-eligible.
- And paragraph 3.11 of Prudential Regulation Authority (PRA) Supervisory Statement 7/18 confirms deferred liabilities can be MA-eligible providing the contractual terms of the surrender will not result in the surrender causing a loss and that firms manage the related liquidity risks.
Firms have not found it problematic to meet the eligibility requirements for deferred liabilities in the PRA Rulebook (Matching Adjustment 2.2 (4)(b)) and SS7/18, and, as noted above, deferred liabilities are no longer novel elements of MA portfolios. However, the surrender optionality of deferred liabilities can create some complications when implementing the wider MA regulatory requirements around cashflow matching and the management of interest rate and liquidity risks for deferred liabilities. As the scale of the portfolios of MA deferred liabilities increases, so too does the size of the actuarial headaches caused by trying to fit the square peg of surrenderable liabilities into the round hole of MA cashflow matching.
Deferred cashflows and the Matching Adjustment
The surrender options that apply to deferred liabilities do not create fundamental economic risks for the insurer. The surrender value basis will be set so that the surrender will not create a loss for the insurer when the liability cashflows that arise in the absence of surrender are well matched (the surrender value will typically be close to, but slightly less than, the best estimates of liabilities (BEL)). So, from an economic perspective, the main issue that arises from these surrender options is not cashflow matching risk but liquidity risk – the insurer must ensure it has sufficient asset liquidity to meet plausible levels of surrender by deferred annuitants.
Nonetheless, the presence of deferred liabilities creates interesting questions for the firm’s MA implementation – for example, how should the firm calculate its MA benefit in the presence of the surrender options; and how does the firm demonstrate that their cashflow matching strategy for deferred liabilities is appropriate and aligned to wider MA requirements?
The answers to these MA questions develop from how the best estimate MA liability cashflows are adjusted to reflect the presence of the surrender option. There are two basic alternatives:
The surrender option is ignored for the purposes of defining the BEL cashflows, i.e. the best estimate liability cashflows are defined as those payable in the event the surrender option is not exercised (even where the best estimates of the take-up of TVs and PCLSs are non-zero).
The probabilities of the surrender being exercised at various times over the deferral period are estimated and used to determine the best estimate liability cashflows (inclusive of the surrender values that are expected to be payable and with projected annuity payments that are correspondingly reduced).
Under approach A, the cashflow matching strategy for the deferred liability is entirely unaffected by the presence of the surrender options. From an economic perspective, this is intuitive as the surrender value basis ensures no change in matching strategy is required.
However, from an MA perspective, some issues quickly emerge. If the firm expects some surrenders, but has matched the cashflows on the assumption of no surrenders, then it must also expect to make future asset sales to provide the necessary liquidity to fund these surrenders. This presents a problem as the MA rules and associated PRA expectations appear to rule out assuming expected MA liability cashflows can be met by future asset sales. In particular:
- Paragraph 6.3 of SS7/18 suggests the PRA does not consider that the selling of assets from an MA portfolio to generate liquidity is consistent with the conditions of regulation 4(5) of the IRPR regulations.
- Paragraph 4.10 of SS7/18 states that firms are not expected to assume any management actions for the purpose of conducting Test 1.
- Planned regular sale of assets is inconsistent with the buy-to-maturity philosophy underlying the matching adjustment framework. Paragraph 2.30 of SS7/18 sets out that matching assets may change only in limited circumstances which are out of the control of the firm, and where expected liability cashflows have materially changed.
These statements appear to have been written without TVs and PCLSs in mind. For example, Paragraph 2.30 explicitly mentions longevity risk as a driver of change in liability cashflows but makes no mention of the surrender options associated with deferred liabilities. There is therefore some ambiguity, and perhaps some scope to argue that the matching strategy of approach A could indeed meet the above requirements.
You may wonder why we are even discussing approach A. If the firm expects some levels of surrender, then surely approach B is the obvious and more intuitive approach? Perhaps. But approach B comes with its own complications. Most fundamentally, under approach B some portion of the best estimate MA liability cashflows will not be fixed – the portion of the projected liability cashflows generated by the assumed payment of surrender values will materially vary with the level of interest rates.
This is both conceptually and practically problematic. It is conceptually problematic because the PRA Rulebook (Matching Adjustment 2.2 (2)) does not include interest rate risk as one of the risks that is permitted to influence MA-eligible liability cashflows; and it is practically problematic because if these liability cashflows are treated as fixed for the purposes of matching and passing PRA Test 1, this could result in a material interest rate mis-match that is economically undesirable and that may result in failing to pass PRA Test 2.
The firm could manage the economics of this problem by matching the cashflows as though they were fixed and then holding some interest rate-sensitive assets outside of the MA asset portfolio such as interest rate swaps. But this seems at odds with the principle that the MA asset portfolio is managed in such a way that there is low risk that it generates insufficient cashflows to meet the MA liability cashflows as they call due.
An illustrative example
We will use a highly simplified example to illustrate the points discussed above. Suppose we write an annuity certain with a deferral period of 15 years and annual payments of £1.00m for the following 35 years. This deferred annuity policy comes with a surrender option – the policyholder can surrender all or part of the policy for cash at any time during the deferral period (but not after). We suppose it is known that the policyholder will take 25% of their pension as a lump sum when the pension commences and will not take a transfer value. The surrender value (SV) is determined as the present value of the annuity cashflows foregone, as implied by the MA discount rate, which we suppose is 4.0% (comprised of a risk-free rate of 3.0% and a 1.0% MA benefit). Moreover, we suppose zero-coupon bonds (ZCBs) exist at all terms.
What are the implications of each of the approaches in this simplified example?
In this case, we construct the MA asset portfolio to match the cashflows in the event of no surrenders, and we sell bonds to meet surrender values as surrenders occur. The PRA matching tests will pass in a straightforward way; but only by assuming we will liquidate assets to fund surrender payments as and when they arise. The liability value is £10.92m and the MA benefit is £3.51m (24% of the non-MA liability value). The asset portfolio and deferred liability cashflows have durations of 28.4 years under the zero-surrender assumption.
In this case, the MA liability cashflows from years 15 to 50 reduce from £1.00m to £0.75m. But we now have an additional liability cashflow of £4.92m in year 15 that arises from the 25% pension lump sum. The liability value and MA benefit is unchanged by moving from approach A to approach B. But, if we treat the £4.92m expected surrender value as a fixed cashflow (rather than one that varies with interest rates) this will result in the duration of the MA liability cashflows being materially understated – instead of having the duration of 28.4 years that is representative of their actual interest rate sensitivity, the liability duration is assessed to be 25 years.
This asset portfolio will ensure the PRA Matching Test 1 is passed (if the surrender value is treated counterfactually as a fixed cashflow) but the duration mismatch will result in a poor performance in Test 2, as is highlighted in the table below.
We could hold interest rate swaps outside of the MA asset portfolio to ensure the overall Solvency II (SII) balance sheet does not have an interest rate mismatch. But this will not be captured in PRA Test 2. This approach will result in (material) MA surpluses and deficits being produced over time as interest rates move up and down, which would require re-balancing of assets between the MA and non-MA asset portfolios.
Other approaches?
The two most obvious approaches to treating liability surrender optionality within the MA framework both have drawbacks: one approach relies on the assumption of future asset sales to fund liability cashflows; and the other introduces non-fixed cashflows into the MA liability cashflow profile. Are there other, more creative ways of fitting the square peg of surrender options into the round hole of the MA?
We noted above that the surrender value is not a fixed cashflow, and that the interest rate swaps that would be held outside the MA portfolio under approach B would not be MA-eligible. However, the PRA’s implementation of the MA framework contains the concept of a ‘paired asset’, where eligibility can be assessed for two (or more) assets when considered jointly. Could we use a similar concept to pair an interest rate-sensitive asset with an interest rate-sensitive liability cashflow (the interest rate variability of the surrender value) to create a paired MA-eligible asset/ liability with low interest rate sensitivity and near-fixed cashflows?
The recent Solvency UK reforms may point to another potential approach. These reforms operationalise a quantitative interpretation of what it means to have high confidence in the MA being earned – the MA benefit should not exceed the 85th percentile adverse outcome. In the context of surrender options this could point towards an approach where only the liability cashflows that arise under the 85th percentile surrender outcome are deemed MA-eligible (with the remaining portion of the liability held outside the MA portfolio)1. However, this creates a couple of challenges. First, unbundling contracts in this way is at odds with the PRA Rulebook (Matching Adjustment 2.3). But this may not be as large as an obstacle as it may initially appear, as the Solvency UK reforms allow some elements of policy unbundling for with profits annuities and claims in payment on income protection policies. This approach does imply that a portion of currently eligible deferred liabilities are (initially) deemed ineligible, which would go against the current policy grain of widening asset and liability MA eligibility.
In a nutshell, the treatment of surrender options within the MA framework is inherently messy. It is a topic that is likely to continue to become more material as record volumes of deferred liabilities are written in new buyouts/buy-in transactions. We expect actuarial best practices in this area to continue to evolve and emerge in response.
1 We are grateful to Alan Reed of APR for introducing this idea when discussing this topic with him.
Press and media enquiries
For more information, please contact our press team on +44 149 478 8813 or via email.
Email usLatest insurance insights
Explore further commentary and insights from Barnett Waddingham's expert insurance consulting and longevity team.
Find out more