Warning: technical actuarial post ahead! This is a guest post from my colleague, Michael Dermody.
On Friday, APRA issued a letter to all CEOs and Appointed Actuaries of life insurers detailed its proposal for calculating the illiquidity premium in the discount rate for certain liabilities as part of the Life Insurance and General Insurance Capital REview (LAGIC).
Background
The illiquidity premium in the calculation of some life insurance liabilities is a technical aspect of the proposed new capital standards to recognise the illiquidity of those liabilities (notably long term annuities). Those liabilities are not liquid, either for the insurer or the policyholder. Instead, they are long term liabilities. The liquid commonwealth government bond market rate includes, in its discount rate, an effective market rate for the liquidity of those securities. So a true risk free rate would be expected to be slightly higher than the commonwealth government bond rate.
APRA has responded to industry feedback that this is an important aspect of the valuation of those liabilities, and hence should be allowed for in valuing those liabilities.
Overview
Key aspects of the proposal are that:
- The illiquidity premium is added to the risk-free discount rat efor the purpose of calculating the liabilities for immediate life annuities, term certain annuities, fixed term/rate products and funeral bonds. For all other life insurance products and for general insurance no illiquidity premium is permitted.
- For the first 10 years, the illiquidity premium added to the government bond rate is 15% x A spread (to govt bond)+ 15% x AA spread (to government bond) with a maximum of 150 basis points (and a minimum of zero). The A, AA spreads to government bond must come from RBA Table F3. After 10 years, the illiquidity premium is fixed at 20 basis points.
- This produces an illiquidity premium 78 basis points at 31/12/2011 (for first 10 years, and 20 basis points thereafter). This calculation was based on spreads of AA spread of 203bps/A of 320bps.
- The formula applies to overseas as well as Australian business (mostly for simplicity as APRA’s view was that the volume of overseas business was too small to need a separate calculation
Public Policy implications
In our submissions to APRA we noted our view that adopting a risk-free discount rate as the “base rate” is flawed and inconsisent with other international capital rameworks and current thinking.
In particular, it suffers from disadvantages including:
- inconsistency with banking regulation – ADIs do not value their liabilities at “risk free”. There is no requiremenet to value term deposits, swap or other derivative book exposures, or loan liabilities, at “risk free” so as to generate large capital charges on these liabilities.
- A 99.5% framework inherently recognises that ultimately there is no guarantee on insurers, and failure (albeit at a low probability) is a risk. The frisk free disocunt rate implies a 100% POS (0% failure risk) which is inconsistent with this objective.
This is unecessary and will place a burden on the life insurance industry in its attempt to meet consumer and public policy demands to provide competitive income stream and longevity products.
The addition of the illiquidity premium goes some way to addressing this for certain products, however, we believe the overall resulting discount rate remains overly conservative.
Industry Capital Levels
The current discount rate for regulatory prudential capital requirements for life insurers is capped at the maximum of swap rates. The spread between BBSW and commonwealth government bond yields varies over time and for different durations.
Depending on the period selected, applying APRA’s method to historic data can produce discount rates higher or lower than swap. However, at 31 December 2011 the swap spread over the bond rate was approximately the same as the APRA illiquidity premium of 78 basis points. However, because of increased credit spread related changes (and where there are significant cashflows beyond 10 years) overall capital levels for annuities are likely to increase.
For QIS5 for Solvency II a formula of 50% of (Spread -0.4%) was adopted (with spread based on the spread of a corporate bond index to swap. Using this formula to estimate discount rates for Australia at 31 December 2011 gives significantly higher discount rates than those derived using APRA’s proposed formula.
Other Aspects
The illiquidity premium is added to forward rates (rather than, for example, zero coupon spot rates). This means that a cash flow at, say, year 12 is discounted for 10 years using a margin of 78bps and two years of 20bps.
Stress scenarios
The credit stresses in the draft LPSs were calibrated on market data – and therefore include the movement in the illiquidity premium. Therefore the stresses to discount rates are now different for liquid/illiquid liabilities.
Pre allowance for diversification, the illiquidity premium in teh post credit stress environment increases by 30 basis points.
APRA has included a further requirement that the overall credit spread risk charge cannot be less than that calculated using default factors alone. This minimum should rarely apply.
Conservative
APRA recognises that they have adopted a conservative approach. They also considered two other approaches:
- Negative basis method (subtracting from corporate bond yields CDS premia); and
- In deciding on the degree of conservatism to apply, APRA also considered the spreads on semi-government bonds. These bonds have minimal credit risk and higher yields than Commonwealth government securities but are more liquid than corporate bonds.
Executable Strategy / Hedgable
Other than the probability of adequacy of 99.5%, APRA has generally provided limited details of the underlying scenarios or principles underlying the new standards (i.e. are we trying to fund run-off, transfer third party, executable strategy, ongoing manage business).
Interestingly in the letter APRA makes some reference to availability of securities in making its decision.
The proposed cut off point of 10 years between the formula-driven illiquidity premium and the proposed long-term rate of 20 basis points was chosen to recognise that there are few corporate bonds with maturity beyond 5 years, whilst semi-government bonds are available in reasonable quantities for maturities up to 12 years.
The illiquidity premium of 20 basis points applying to forward rates beyond 10 years is a conservative estimate of the long-term average illiquidity premium that might be accessible to investors in the future.
There are few corporate bonds on issue with maturity greater than 5 years. It would be more difficult for an insurer to access the illiquidity premium implicit in corporate bond yields for maturities greater than 5 years.
Anti-Cyclical
A positive feature of the illiquidity premium is that it is market sensitive and may potentially reduce the volatility of the capital position of a life company.
Next Steps and Ongoing Issues
APRA is seeking feedback and expects to finalise the proposals relating to illiquidity premium by September 2012.
The Institute of Actuaries has been involved in providing analysis with regards to the illiquidity premium to APRA but did not provide any recommendation in relation to methodology of determining illiquidity premium or appropriate illiquidity premium to adopt.
However, given the extensive feedback that APRA has already received on the subject of the “base” discount rate as well the illiquidity premium, we would be surprised if any changes, other than at the margins, are introduced.
Nonetheless, it is worth recapping on some of the key issues that have emerged during the debate (beyond those already discussed above
- Broaden application to other products – illiquidity relates to the impact on value because of difficulties or costs of conversion to cash. It is arguable that an illiquidity premium is applicable to liabilities other than those envisaged by APRA. For QIS5 products were divided into several “buckets” of various illiquidity (for example incorporating conventional business) and compared with the single bucket for APRA. If the current low discount rates persist, the importance of the illiquidity premium and industry demands for broadening the application may increase.
- Materiality – the liquidity premium is of most interest (and of public policy concern) to long term liabilities with low levels of insurance type risks. We are sympathetic to the view that there is limited value in complicating the discount rate derivation where there are highly uncertain insurance risks of substantially greater importance
- Accounting standards – both the life (AASB 1038) and general insurance (AAS 1023) accounting standards require risk-free discount rates which relate to the nature of the liabilities. It is possible that a similar such definition may be adopted in an international insurance accounting standard. Debate is ongoing whether this definition should be interpreted as requiring liquidity premium based on the liquidity characteristics of the liability.
Overall outcome
This issue of discount rates for long term illiquid liabilities is one that has been difficult around the world. It is even harder in Australia where there are very few debt securities of any kind with terms matching the liabilities. APRA, while acknowledging it has been conservative, seems to have reached a workable outcome that responds to most of the issues raised by the industry.
I can’t help feeling the industry has got the wrong end of the stick on risk-adjusted valuation. In the 1990s we fielded arguments from the general insurance industry (or at least the riskier participants in it) that the presence of risky assets backing the technical provision should result in a lower valuation for the technical liabilities, since the expected return was higher. The Institute of Actuaries made it clear that it supported the paradigm of risk free + risk margin – ie. a value always higher than the risk free rate, to reflect the liability risk.
It seems to me that this is making a similar argument – that the presence of liquidity risk makes the liabilities less valuable. The implication seems to be that a liability of 100 should be valued at, say 95. I would argue that ‘less valuable’ in this sense makes the liability 105, not 95.
Surely, to take this liability off your hands you would have to pay me *more* (not less) than the equivalent risk-free portfolio?
I think applying a stochastic discount factor approach (a la Cochrane) will back me up on this.
Hmm. now that I investigate further, the argument is not the backing of the liabilities with illiquid assets (such as corporate bonds) although there are references to this in the literature which simply serve to muddy the waters. In that case I think my argument is correct – you can’t reduce liabilities using the argument of a reduced value of the asset backing.
In this case the annuitant is illiquid, so the value of that asset (and hence the life company liability) is reduced.
The issue is not about what assets are backing the liabilities, but rather whether the liabilities are liquid or illiquid. If the liabilities are illiquid (which is generally true of the liability of a lifetime annuitant) then the appropriate discount rate is the risk free discount rate for an illiquid asset.
An illiquid asset, by definition, is not tradeable, and therefore its discount rate is not observable. So the approach APRA is proposing is attempting to define that an illiquid risk free discount rate.
In this definition, the assets backing the liabilities are not relevant – they could be highly geared equities, and the liability valuation would be exactly the same.
I agree with your view that a liability valuation should not depend on the asests backing it (unless, of course, in the case of a unit linked liability, there is a contractual link).