[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries
s
.

VFA: How do your contracts measure up?

Steven Morrison explores the use of stochastic scenarios for assessing whether insurance contracts meet eligibility criteria for the variable fee approach under IFRS 17

05 DECEMBER 2019 | STEVEN MORRISON

© iStock
© iStock


IFRS 17 provides a specific measurement model 
for insurance contracts with direct participation features, known as the variable fee approach (VFA). This refers to the fact that such contracts are characterised by a variable fee that the entity charges in exchange for investment-related services. The variable fee is treated differently under the VFA than under the general measurement model, resulting in different attribution between insurance service and finance results, profit timing and volatility. Understanding whether existing contracts meet the eligibility criteria for the VFA is therefore of great importance to companies implementing IFRS 17.

As with many other decisions in the standard, the criteria for VFA eligibility leaves room for interpretation by individual entities. What are the key criteria, and how could they be assessed using stochastic scenarios?

Scenario-based modelling

IFRS 17 Paragraph B101 provides a definition of insurance contracts with direct participation features:

a. The contractual terms specify that the policyholder participates in a share of a clearly identified pool of underlying items

b. The entity expects to pay the policyholder an amount equal to a substantial share of the fair value returns on the underlying items and 

c. The entity expects a substantial proportion of any changes in the amount to be paid to the policyholder to vary with the change in fair value of the underlying items. 

Paragraph 107 provides further guidance as to the interpretation of the characteristics set out in Paragraph 101(b) and 101(c):

An entity shall assess the variability in the amounts in paragraphs B101(b) and B101(c): 

i. Over the duration of the group of insurance contracts; and 

ii. On a present value probability-weighted average basis, not a best or worst outcome basis. Paragraph B108 goes on to discuss the outcomes that might arise for products containing guarantees in the particular scenario.


Paragraphs B107(ii) and B108 acknowledge that different scenarios can have different outcomes in terms of the returns on underlying items and the policyholders’ share in these returns, and that assessment of the criteria set out in Paragraphs 101(b) and 101(c) should allow for this. How can stochastic scenarios be used to quantify policyholder share and variability?

The use of stochastic scenarios might appear daunting in terms of operational and computational effort. However, although they are not prescribed in the standard, stochastic modelling techniques are expected to be widely used in the calculation of fulfilment cash flows for contracts with participation features, which makes the effort of using stochastic scenarios to assess VFA eligibility more manageable. 

Description of example contract 

The illustrative contract group considered here consists of 100 contracts, with a 10-year coverage period and a single premium of CU 150 per policy received at issuance. The contract provides the following benefits:

  • Death (during the coverage period): the account balance with a guaranteed minimum benefit. We assume one policyholder dies each year. 
  • Survival (until the end of the coverage period): the account balance with a guaranteed minimum maturity benefit.

At inception, the account balance equals premiums paid. It subsequently evolves annually based on returns, subject to an annual charge to reflect investment services provided as a fixed percentage of the account balance at the end of the year (prior to paying any benefits).

This contract satisfies the requirement that the policyholder participates in a clearly identified pool of items (the account balance). The relevant question is whether the level and variability of this participation meets the requirements set out in Paragraphs 101(b) and 101(c). 

Eligibility metrics

To assess the criteria in 101(b) and 101(c), we must define metrics that quantify the policyholders’ share of the fair value returns on the underlying items and its variability.

One of the challenges with interpreting Paragraph 101(b) is that it refers to the policyholders’ share of the fair value returns. However, for many contracts, including the one considered here, the payments to both policyholders and the entity are expressed in terms of the current value of the underlying items, rather than fair value returns.

Each year, payments are made to the policy holder and the entity, while part of the returns achieved are retained to be paid in future years: 


Equation 1: EntityCashFlow(t) + PolicyholderCashFlow(t) = FairValueReturn(t) + AccountBalance(t–1) – AccountBalance(t)

The account balance is initially equal to the premium and completely distributed at maturity of the contract. Over the life of the contract, the sum of fair value returns breaks down into two terms: the sum of cash flows to the entity and sum of cash flows to policyholders:                                                       

equation 1

Different scenarios will result in different outcomes in terms of the returns on underlying items and the cash flows to policyholders. Taking the expectation (E) overall economic scenarios gives:                                                                                                         

equation 1 2

 So one possible measure of the policyholders’ share of fair value returns on the underlying items is:

screenshot 3

This metric measures the share over the duration of the contact group, so appears to satisfy the requirement in Paragraph 107(b)(i). The use of probability-weighted averages would appear to at least partly satisfy the requirement in Paragraph 107(b)(ii).

Paragraph 107(b)(ii) refers to a present value probability-weighted average basis, while the metric proposed in Equation 2 is based on undiscounted values. The problem with taking present values is that, as noted above, each year the fair value return does not naturally decompose into payments to the policyholder and the entity. If we discount Equation 1 using a cash account, before summing during the coverage period, we find:

screenshot 4

So the sum of discounted fair value returns can’t be fully attributed to the policyholder and the entity, unless CashRate(t-1) = 0 at all times. Here we choose to use metrics based on undiscounted values.

Paragraph 101(c) further requires measuring to what extent changes in the amount to be paid to the policyholder varies with the change in fair value of the underlying items. For the contract considered here, the presence of guaranteed benefits means the total cash flow paid to policyholders is bounded below by some positive amount. In such scenarios, the policyholders’ cash flow can be considered fixed and independent of the return on underlying items. In all other scenarios the cash flows vary with the returns on underlying items.

Therefore, a measure of variability is the probability that the sum of policyholders’ cash flows exceeds the minimum:


screenshot 5

VFA eligibility assessment of example contract 

Here, we estimate the two VFA eligibility metrics for the example contract group above under different variations of the contract features. Metrics are estimated using 1,000 stochastic scenarios for the return on underlying items.


Contract 1: Guarantee death benefit = 170; Guaranteed maturity benefit = 0; Charges = 0.5% p.a. 


First, we consider a contract that has a minimum death benefit of CU 170 but no guarantee at maturity. With relatively small guaranteed benefits, the contract is profitable with a modest annual charge (0.5% p.a.). Figure 1 (below) shows the sum of all cash flows to policyholders plotted against the sum of all fair value returns on underlying items, in all 1,000 stochastic scenarios. The dashed lines show averages of the sum of cash flows to policyholders (numerator in Equation 2) and the sum of fair value returns (denominator in Equation 2), with the solid brown line indicating y=x. The closeness of the intersection of the dashed lines to the solid line indicates the level of the policyholders’ share.

In this example, the average sum of policyholders’ cash flow is estimated as CU 7,464, compared to an average sum of investment returns of CU 8,172. The policyholders’ share is therefore 91%.

The minimum sum of policyholders’ cash flows, estimated over all 1,000 scenarios, is CU -10,682. This value is measured in just one scenario (ie the sum of policyholder cash flows in 999 scenarios is strictly greater than this), so the variability metric (Equation 3) is estimated as 99.9%.

The high variability in this case is due to the fact that guarantees are only provided on death. With only one death per year, a significant component of the total policyholders’ cash flow is the maturity benefit, which is equal to the outstanding account balance with no minimum guarantee. The theoretical minimum sum of policyholders’ cash flows (CU -13,300) is never achieved as it would require the account balance to have been completely wiped out by year 10.


Contract 2: Guaranteed death benefit = 170; Guaranteed maturity benefit = 150; Charges = 4% p.a. 


We now consider a contract with minimum maturity benefit of 
CU 150 and a guaranteed death benefit of CU 170. A significantly larger annual charge is assumed (4% p.a.) to cover the additional cost of providing the maturity guarantee.

Figure 2 (below) shows the sum of all cash flows to policyholders plotted against the sum of all fair value returns on underlying items in all 1,000 stochastic scenarios.

In this example, the average sum of policyholders’ cash flow is estimated as CU 3,488, compared to an average sum of investment returns of CU 6,863. The policyholders’ share is thus estimated as 51%.

The minimum sum of policyholders’ cash flows, estimated over all 1,000 scenarios, is CU 200. This is the theoretical minimum amount corresponding to all 10 death guarantees biting, as well as the maturity guarantee. This minimum sum of cash flows is measured in 329 of the 1,000 stochastic scenarios used, resulting in a variability metric of 67%.


Conclusions

We have described the use of stochastic scenarios to assess whether an insurance contract qualifies as a contract with direct participation features, according to criteria set out in IFRS 17. We have also defined metrics that can be used to quantify the degree of participation as a share of the returns on underlying items (the policyholders’ share) and the extent to which variation in underlying items results in variation in policyholder cash flows (the variability). 

The metrics presented here, and alternative ones, are sensitive to assumptions embedded in the scenarios used to calculate them, in particular the assumed growth rates on underlying items. Different entities may conclude that similar contracts have different degrees of policyholder share and variability simply because they have used different metrics and assumptions.

Another item of potential controversy is whether the resulting numbers are ‘substantial’, as required by paragraph 101. Although it is a judgment of individual entities, consensus as to appropriate thresholds might emerge as insurers approach the IFRS 17 implementation date.

figure 1
Figure 1: Sum of policyholder cash flows vs sum of fair value returns (Contract 1)
figure 2
Figure 2: Sum of policyholder cash flows vs sum of fair value returns (Contract 2)