Asset Dependency Discounting
In this article, I would discuss about the "Asset dependency discounting" issue in the IFRS 17 final standard.
According to paragraph B74:
Estimates of discount rates shall be consistent with other estimates used to measure insurance contracts to avoid double counting or omissions; for example:
(a) Cash flows that do not vary based on the returns on any underlying items shall be discounted at rates that do not reflect any such variability;
In this article, I would discuss about the "Asset dependency discounting" issue in the IFRS 17 final standard.
According to paragraph B74:
Estimates of discount rates shall be consistent with other estimates used to measure insurance contracts to avoid double counting or omissions; for example:
(a) Cash flows that do not vary based on the returns on any underlying items shall be discounted at rates that do not reflect any such variability;
(b) Cash flows that vary based on the returns on any financial underlying items shall be:
i. Discounted using rates that reflect that variability; or
ii. Adjusted for the effect of that variability and discounted at a rate that reflects the adjustment made.
A first grant in the standard seems to be obvious. Discount asset dependent cash flows (ADCF) by risk discount rate, and discount non-asset dependent cash flows (NADCF) by risk free rate. In the example of a Traditional Life (TL) Participating (PAR) product, dividend related cash flows will be asset dependent, whist, other guarantee cash flows, including premium and expense, will be non-asset dependent.
The problem arises when considers the following hypothetical product:
Consider a Universal-Life contract, 3-pay-5. The promised return is exactly equal to the investment return, 5%. Benefit is only payable upon maturity, regardless of the policyholder died or not. Hence no policyholder will choose to lapse. For simplicity, assume no expenses / charges / deductions. Assume risk-free rate to be 3%.
The account value roll-forward is as below:
And the split cash flow discounting is as below:
One can see that, if we use risk-free rate to discount the premium (which does not vary based on the returns on any underlying items), but use risk-discount rate to discount the maturity benefit, this will result in day-one negative Best-Estimated-Liabilities (BEL). This negative BEL will become the Contractual Service Margin (CSM) at initial recognition, and amortized in future years.
However, there are no gain / loss from the insurer side for this contract. Since the insurer is crediting exactly what it earns to the policyholder, even if we considered there is an arbitrage spread (5% investment return vs 3% interest expense), this gain is solely attributed to the policyholder, not the insurer. The day-one CSM built-up from the insurer side is double counting the economic value from this contract.
Therefore, we believe the correct statement for B74 should be "two-way" instead of "one-way":
Estimates of discount rates shall be consistent with other estimates used to measure insurance contracts to avoid double counting or omissions; for example:
(a) Cash flows that do not vary based on, and do not affect the returns on any underlying items shall be discounted at rates that do not reflect any such variability;
(b) Cash flows that vary based on, or may affect the returns on any financial underlying items shall be:
i. Discounted using rates that reflect that variability; or
ii. Adjusted for the effect of that variability and discounted at a rate that reflects the adjustment made.
Under the corrected statement, there are 2 ways for splitting the cash flow:
1. Consider all cash flows as ADCF. Say, for a Unit-Linked contract, even the "guarantee" part of cash flows like Guarantee Minimum Death Benefit (GMDB), will not be paid since the contract is lapsed if the account value drops to 0. By nature it is like a deep out-the-money (OTM) call contract hence is an embedded derivative, thereby asset dependent.
2. Split the cash flows with the premiums backing it. Say, for a TL PAR contract, we can consider the guarantee cash flows to be non-asset dependent but the non-guarantee cash flows to be asset dependent. It is theoretically possible to separate the portion of premium backing both part separately. For example, calculate the NADCF BEL and ADCF BEL, determining their ratios and use the same ratio to split the premium.
One may argue that, according to Paragraph B77, the splitting of cash flows is not required:
IFRS 17 does not require an entity to divide estimated cash flows into those that vary based on the returns on underlying items and those that do not. If an entity does not divide the estimated cash flows in this way, the entity shall apply discount rates appropriate for the estimated cash flows as a whole, for example, using stochastic modelling technique or risk-neutral measurement techniques.
I believe this statement is pointing to B74 (b) (ii), such that using risk-neutral measurement techniques may avoid the split of cash flow problem. (Whether the use of risk-neutral technique is reasonable in IFRS 17 is another topic, I will discuss this in another article)
However, the above logic still does not stand. In VFA model, the subsequent measurement needs to be unlocked by change in NAD BEL and change in variable fees from the underlying item. The splitting of premium is still required to calculate the asset share backing ADCF, but not NADCFs.
To conclude, I believe the IASB should change the wording in B74 accordingly. And even after the change, the classification of asset dependent cash flows, and splitting of premium will still be another debatable topic, regardless of whether the company use risk-neutral measurement techniques or not.
What do you think?
i. Discounted using rates that reflect that variability; or
ii. Adjusted for the effect of that variability and discounted at a rate that reflects the adjustment made.
A first grant in the standard seems to be obvious. Discount asset dependent cash flows (ADCF) by risk discount rate, and discount non-asset dependent cash flows (NADCF) by risk free rate. In the example of a Traditional Life (TL) Participating (PAR) product, dividend related cash flows will be asset dependent, whist, other guarantee cash flows, including premium and expense, will be non-asset dependent.
The problem arises when considers the following hypothetical product:
Consider a Universal-Life contract, 3-pay-5. The promised return is exactly equal to the investment return, 5%. Benefit is only payable upon maturity, regardless of the policyholder died or not. Hence no policyholder will choose to lapse. For simplicity, assume no expenses / charges / deductions. Assume risk-free rate to be 3%.
The account value roll-forward is as below:
And the split cash flow discounting is as below:
One can see that, if we use risk-free rate to discount the premium (which does not vary based on the returns on any underlying items), but use risk-discount rate to discount the maturity benefit, this will result in day-one negative Best-Estimated-Liabilities (BEL). This negative BEL will become the Contractual Service Margin (CSM) at initial recognition, and amortized in future years.
However, there are no gain / loss from the insurer side for this contract. Since the insurer is crediting exactly what it earns to the policyholder, even if we considered there is an arbitrage spread (5% investment return vs 3% interest expense), this gain is solely attributed to the policyholder, not the insurer. The day-one CSM built-up from the insurer side is double counting the economic value from this contract.
Therefore, we believe the correct statement for B74 should be "two-way" instead of "one-way":
Estimates of discount rates shall be consistent with other estimates used to measure insurance contracts to avoid double counting or omissions; for example:
(a) Cash flows that do not vary based on, and do not affect the returns on any underlying items shall be discounted at rates that do not reflect any such variability;
(b) Cash flows that vary based on, or may affect the returns on any financial underlying items shall be:
i. Discounted using rates that reflect that variability; or
ii. Adjusted for the effect of that variability and discounted at a rate that reflects the adjustment made.
Under the corrected statement, there are 2 ways for splitting the cash flow:
1. Consider all cash flows as ADCF. Say, for a Unit-Linked contract, even the "guarantee" part of cash flows like Guarantee Minimum Death Benefit (GMDB), will not be paid since the contract is lapsed if the account value drops to 0. By nature it is like a deep out-the-money (OTM) call contract hence is an embedded derivative, thereby asset dependent.
2. Split the cash flows with the premiums backing it. Say, for a TL PAR contract, we can consider the guarantee cash flows to be non-asset dependent but the non-guarantee cash flows to be asset dependent. It is theoretically possible to separate the portion of premium backing both part separately. For example, calculate the NADCF BEL and ADCF BEL, determining their ratios and use the same ratio to split the premium.
One may argue that, according to Paragraph B77, the splitting of cash flows is not required:
IFRS 17 does not require an entity to divide estimated cash flows into those that vary based on the returns on underlying items and those that do not. If an entity does not divide the estimated cash flows in this way, the entity shall apply discount rates appropriate for the estimated cash flows as a whole, for example, using stochastic modelling technique or risk-neutral measurement techniques.
I believe this statement is pointing to B74 (b) (ii), such that using risk-neutral measurement techniques may avoid the split of cash flow problem. (Whether the use of risk-neutral technique is reasonable in IFRS 17 is another topic, I will discuss this in another article)
However, the above logic still does not stand. In VFA model, the subsequent measurement needs to be unlocked by change in NAD BEL and change in variable fees from the underlying item. The splitting of premium is still required to calculate the asset share backing ADCF, but not NADCFs.
To conclude, I believe the IASB should change the wording in B74 accordingly. And even after the change, the classification of asset dependent cash flows, and splitting of premium will still be another debatable topic, regardless of whether the company use risk-neutral measurement techniques or not.
What do you think?
Comments
Post a Comment