IFRS 17 – Dis-count Dracula
It’s everyone’s favourite time of the month again – APR’s next IFRS 17 article has arrived! Our topic of interest this month is discounting (pun very much intended).
So why is discounting so important in IFRS 17?
Without wanting to take you back to day one of an actuarial science course, all actuarial valuations need to include some adjustment to consider the time-value of money. As a general principle, the more risk / uncertainty there is concerning the amount / timing of future cashflows, the more harshly they should be discounted when determining their present value.
Valuations are highly sensitive to changes in the discount rate used, and this is especially true for long-term contracts as a result of compounding (as shown in the very straightforward illustration below). The accounting standards need to make sure the numbers reported on financial statements are prudent and faithfully representative of an insurance company’s liabilities. They therefore have a direct interest in regulating how companies discount and in making sure excessively high discount rates are not used to understate the value of insurance liabilities.
The core principle of discounting under IFRS 17 is that the discount rates used should reflect only the time-value of money and relevant financial risks. Discount rates should not be adjusted to account for non-financial risks – under IFRS 17 adjustment for non-financial risks is catered for under the Risk Adjustment which we discussed in our previous article.
Well then, what options does IFRS 17 give me when it comes to discounting?
Flat Rate or Curve?
Either – but preferably a curve. IFRS 17 does not explicitly specify whether an entity must use a single discount rate or a discount yield curve. The standard’s text itself uses the terms “rate” and “curve” interchangeably, so technically, either is acceptable. However, in practice, a yield curve will almost always be used – especially in the business of long-term insurance – as this will be much more accurate and representative from a finance point of view. Nevertheless, in the rare case where constructing a yield curve is impossible, the option of using a single discount rate to value a contract is available, and not contrary to the standard.
Accounting Procedures
The discount rates / curves assumed will change from year to year as they are updated, creating differences in valuations over time. Under IFRS 17, the income / expenses resulting from changes in the discounting assumptions can be accounted for in either Profit and Loss, or Other Comprehensive Income. Both are acceptable under the standard so long as consistency is maintained.
We have held a specific group of assets to support a group of insurance liabilities. Can’t we just use the returns of these backing-assets as the discount rate for this group?
Absolutely not! (I’m sorry, but unfortunately it is true☹) Whilst IFRS 4 was more than happy to let you use the yield of the insurance contract’s backing-assets as the discount rate, IFRS 17 is not going to let you get off so easily.
Returns on financial assets are always greater than the “risk-free rate of return” (usually denoted by the return on Government Bonds). This excess return (or risk premium) is due to numerous reasons, the most prominent two being the following:
- Most financial assets are not perfectly liquid like Government Bonds[1] – investors demand a “illiquidity premium” to compensate for this.
- Most financial assets have default or credit risk unlike Government Bonds – investors demand a “credit risk premium” to compensate for this.
The view of IFRS 17 is that insurance contracts are illiquid for the policyholder – the contracts cannot be resold onto an open market, and usually cannot be surrendered without paying a penalty. However, they involve little-to-no credit risk. Therefore, insurance contracts should be discounted at the Risk-Free Rate plus an Illiquidity premium. Notably, the discount rate used should be exclusive of the Credit-Risk Premium which asset yields will typically have. That said, as illustrated in the chart below, the following relationship holds:
Risk Free Rate < IFRS 17 Discount Rate < Yield of Backing-Assets
Note: The risk-free yields and asset returns in the above chart (and others below) are hypothetical – these yields are not taken from any data source.
So essentially, if we are to discount using the yield of the insurance contract’s backing-assets, we would be understating the present value of the contract’s liability by discounting too harshly – and so IFRS 17 forbids this practice.
And what approaches does IFRS 17 put forward for deriving these rates?
Well the idea is that we must find some rate in between the risk-free yield and the yield of backing-assets. IFRS 17 gives us two approaches – either start with the risk-free rate and add a derived illiquidity premium (called the bottom-up approach), or start with the yield of backing-assets and “strip-down” the credit risk premium (called the top-down approach).
So in the botton-up approach shown here, the difference between the two yield curves represents the illiquidity premium:
and in the top-down approach, the difference between the curves represents the credit risk premium:
Importantly, the entity does not have to show that the rate or curve used can be derived through both approaches – only one approach needs to be used. IFRS 17 acknowledges that these approaches may result in different final rates / curves and does not require the entity to show they reconcile. The only accounting requirement is for the approach used to be stated in the accounting notes[2].
Clearly, if the top-down approach is used, a reference portfolio – that is, the portfolio of assets you start from – needs to be chosen. IFRS 17 has no specific requirements for reference portfolios, only a note that the more similar a reference portfolio is (in terms of cashflow amounts, timing, liquidity, risk etc.) to the group of insurance contracts we are valuing, the fewer adjustments need to be made when deriving the discount rate / curve. In most cases, there should be a sensible reference portfolio available to choose.
That’s all well and good but stripping out the credit risk premium from an asset’s return, or estimating the illiquidity premium itself, doesn’t sound like a trivial task. How should I go about this?
A good point! Unfortunately, IFRS 17 itself provides you with no help here. The standard just specifies that the discount rate must be the risk-free rate plus a premium for illiquidity, and then leaves you alone to figure out how to actually estimate this. Fortunately, there are some non-IFRS sources that have recommendations and ideas, but we should be clear that these are in no way part of the standard. There is no explicit requirement to use the techniques below to derive discount rates in order to sign off your statements as IFRS-compliant.
As examples, the following section consists of two recommended methods from the International Actuarial Association (IAA).
Estimating an illiquidity premium for the bottom-up approach
The IAA recommends using covered bonds. These are a kind of derivative product whereby the issuer promises a set of cashflows to the holder (as per a normal bond), except the cashflows are covered by the issuer being explicitly obliged to hold assets (such as mortgage loans) which provide cash for the issuer to pay the holder. That is, the bond payments are “covered” by other cashflows from assets the issuer holds, and hence the bonds are very safe with respect to credit risk. The market for covered bonds is not that significant (compared with corporate or government bonds) making covered bonds less liquid. The difference between a 10-year covered bond yield and a 10-year (risk-free) Treasury bond yield therefore represents an estimate of an illiquidity premium at the 10-year term (but the yield curve would still be below that of, say, a corporate bond, which includes an element of credit risk).
Estimating a credit-risk premium for the top-down approach
The IAA recommends using credit default swaps (CDSs) – these are essentially insurance products against credit default. A holder of a bond takes out a CDS with a counterparty, the bond-holder pays the CDS issuer premiums, and the CDS issuer pays out the value of the loan credit in the event of default – thereby taking on the credit risk. A portfolio consisting of a group of assets plus a corresponding group of CDSs (to remove all credit risk) should thereby have a yield whose excess return represents an illiquidity premium only. CDSs can be used in this way to strip-down credit risk from a reference portfolio.
Is there anything else I should consider for discounting?
Discounting Real vs Nominal Cashflows
As it has always been, cashflows which are linked to inflation (ie real cashflows) must be discounted at real discount rates – the risk-free rate in this context being the rates on index-linked government bonds, with the the relevant illiquidity premiums then added. Cashflows independent of inflation (nominal cashflows) are discounted at nominal rates of interest as usual. Alternatively, inflation-linked cashflows can be projected with inclusions for inflationary increase, and then discounted at nominal rates.
Negative Discount Rates
It follows that in the era of very low interest rates, there is a possibility of having to resort to negative discount rates when discounting real cashflows (a very low nominal interest rate can lead to a negative real interest rate after adjusting for inflation). IFRS 17 doesn’t specify that discount rates have a floor of zero and using negative discount rates is not non-compliant or out of the question! In practice, negative discount rates are still extremely exceptional. In our opinion, firms should not use them unless they believe it is absolutely necessary, and if doing so should provide strong justifications for doing so in the notes.
Discounting Over Very Long Terms
Active and observable markets for bonds usually do not contain products with terms greater than 30 years (for example, the 30-year UK Treasury bond is the longest-term gilt which is regularly traded). Many Life Insurance products have term lengths out-stripping this. IFRS 17’s only requirement is that “if data is available for an observable market, then this data must be used”. Any current methods of extrapolation which are already in use (eg constant spot rates, constant forward rates, flat curve after 30 years) are also permitted to be used under IFRS 17.
Time for the recap…
- Discounting methods have a direct major impact on valuations, and accounting standards have a motive to regulate them to ensure companies achieve prudence and faithful representation.
- IFRS 17 allows for either flat rates or curves to be used. In practice, a curve is almost always more appropriate.
- Changes in discounting assumptions over time generate income and expenses. Under IFRS 17, it is the entity’s choice whether this is displayed under Profit / Loss or Other Comprehensive Income.
- Under IFRS 17, discount rates should normally be the risk-free rate plus an illiquidity premium. These are generally more than the risk-free rate and less than the yields on backing-assets.
- Discount rates can be derived by starting at the risk-free rate and estimating the illiquidity premium (“bottom-up”) or starting at backing-assets’ yields and stripping the non-illiquidity premium (“top-down”). The results of the two approaches do not have to equate, but IFRS 17 requires disclosure of the method used.
- Firms must discount real cashflows at real interest rates and nominal cashflows at nominal interest rates.
- Negative discount rates are not ruled-out in IFRS 17 and there is mixed opinion within the industry as to whether or not they should be applied in certain, rare cases.
- Yield-curve extrapolation methods in use today for discounting cashflows over more than 30 years are still applicable under IFRS 17.
Stay tuned for APR’s next article in the IFRS 17 series. See you then!
Ajay Kotecha
February 2019
[1] Whilst not all government bonds are fully liquid and risk free, they are treated as such in the majority of cases and generally used as the benchmark for all other asset classes.
[2] See IFRS 17 paragraph B84.