The Solvency II discount rate: Nothing is simple kpmg.co.uk/solvencyii
Introduction Nick Ford Senior Manager Actuarial (Life Insurance) Tel: 020 7311 5913 nicholas.ford@kpmg.co.uk Liquidity premium, matching premium, symmetric adjusters, extrapolation methods and a number of other technical concepts have turned the fairly simple concept of a discount rate into a minefield under Solvency II. The following simplifies these concepts into one, easily digestible paper to explain the issues, why they re important and their impacts. Solvency II has raised a lot of methodological issues around the concept of risk based capital and market consistent valuations. Some of these are understandable how one determines an event that would only occur once in 200 years is clearly open to a myriad of calibration issues. However, slightly more surprisingly, there has been a lot of disagreement as to what rate should be used to discount liability cash flows for the economic balance sheet. For those familiar with Market Consistent Embedded Value (MCEV), some debate around a liquidity premium may be expected, but there has also been significant debate as to what the basic risk free rate curve is what it is based on, what it should ultimately trend to and how it should reach that ultimate rate. The following attempts to explain the latest guidance and the issues arising. 1 The Solvency II discount rate: Nothing is simple / June 2012
What is risk free? A risk neutral valuation requires a risk free rate more precisely, a duration dependent risk free rate curve. So what financial instruments can we invest in with no risk? In the early stages of Solvency II, the yields on government bonds were assumed to give the best indication of risk free indeed the discount rate term structures used in QIS3 and QIS4 in 2007 and 2008 respectively, were based on government bonds. However, it was then determined that a better view of risk free would come from the rate that banks lend to each other at i.e. the inter-bank swap rate. So, to set guidelines for Solvency II, can we simply state then that the risk free rate should be the inter-bank swap rate for a given currency? Not quite. There are still questions: When looking at swap rates, are we looking at a floating leg that references overnight, 3 month, or 6 month interbank rates? If we are not looking at overnight rates, is there some credit risk involved? What is the longest duration that swap rates are reliable for and what do we do for durations beyond that? We will tackle each of these questions in turn. The floating rate deposit period Interest rate swaps are typically based on short term LIBOR rates (LIBOR is short for London Inter Bank Offer Rate and is the rate at which banks will lend to each other). What we mean by this is that the floating leg of the swap will be calculated based on the LIBOR rate. Depending on the frequency of the swap payments, this may be the overnight, 3 month, 6 month or annual LIBOR rate. In the swap market a 5 year rate vs 6 month LIBOR refers to the rate on a swap of 5 years with semi-annual payments, where the floating leg is based on LIBOR. The question then is, why does this matter? Well, shortening the term of lending should decrease credit risk; overnight lending will almost always be repaid. However, swaps referencing overnight LIBOR are not common and it could not be considered that there is an active, deep, liquid and transparent market for them. All of these are required characteristics for the market upon which the Solvency II risk free rate is based. The most commonplace arrangement in the market is a swap based on a 6 month deposit period and it is these rates that are proposed for use as the basis for the Solvency II risk free rate, as they will have the required characteristics. Credit risk deduction? As the deposit period is assumed to be 6 months, there is potential for credit risk i.e. there is a risk of not earning the reference floating rate given that there is risk associated with depositing the notional amount with an institution for the 6 month period. As noted earlier, using an overnight deposit period (e.g. an overnight index swap) would largely eliminate credit risk, but does not provide sufficiently liquid and reliable data. Currently, the Solvency II regulations specify a deduction of 10 basis points to allow for this credit risk at all durations. A joint CFO / CRO Forum paper has analysed the appropriate deduction by comparing unsecured inter-bank lending with secured repurchase agreement (repos) rates a method commonly used to measure the impact of credit risk in swap rates. This shows the 10 basis point adjustment to be on the conservative side and the size begs the question of whether it is material enough to warrant any adjustment. The analysis performed on the size of this deduction (both from the CFO / CRO Forum and EIOPA) has concentrated on the perceived current margin for credit risk. However, given the assumption that the deduction should be a parallel shift at all durations, there is a question of whether a long term margin should be considered instead. Of course, there is also the important question of the realism of a parallel shift adjustment. However, EIOPA have commented that this approximation is necessary due to a lack of appropriate term dependent data. So, it looks likely that a 10 basis point deduction will remain and companies will need to consider how this aligns with the rates assumed for current economic capital purposes. The Solvency II discount rate: Nothing is simple / June 2012 2
At what duration does the data become unreliable? Interbank swaps are typically short duration instruments. However, insurance liabilities are typically of a significantly longer duration. Therefore, it is necessary to determine the longest duration on the swap curve for which we have reliable data and extrapolate until we have a curve that is of long enough duration to be usable for discounting insurance liability cash flows (if we imagine we sell a Whole Life product to a woman aged 25, the cash flows could easily extend for 70 years). For each currency, the last liquid point needs to be determined i.e. the point at which we ignore market data and start to extrapolate. Currently, the last liquid point for the major currencies has been assumed as follows: Currency EUR GBP USD Last liquid point 20 years 50 years 30 years Source: QIS5 Technical Specification Risk-free interest rates & Omnibus 2 The last liquid point for the Euro swap curve has come under much scrutiny and has been reduced compared to, say, QIS5. However, there has been little information or scrutiny of the last liquid point for GBP or USD curves. One may expect this to occur at some point which might result in a reduction in the last liquid point for these currencies as well. A reduction in this point will make the extrapolation technique much more important for GBP denominated liabilities. How do we extrapolate? There are many different extrapolation methods in existence and EIOPA have considered the following: The Smith-Wilson method The Nelson-Siegel method The Svensson method (in a macroeconomic formulation) The method used and developed by Barrie & Hibbert A linear approach (as adopted for QIS5) The Smith-Wilson method is currently the one being proposed for use. As with any extrapolation method, certain assumptions are required the key assumptions being: Entry point of extrapolation (i.e. last liquid point as discussed above) The ultimate forward rate (UFR) (i.e. the rate that the curve should converge to) The maximum period of convergence (i.e. at what point the curve will be considered close enough to the UFR) The speed of convergence (i.e. how quickly the curve will converge) Each of these assumptions is open to judgement for example, we have already discussed how the Euro entry point of extrapolation has shifted over the last year or so. Ultimate Forward Rate Similarly, there is debate regarding the ultimate forward rate (UFR). The UFR is driven by the long term expected real interest rates and long term expected inflation. Theoretically, there are other drivers of the long term forward rate, but these are ignored and are not further discussed in this paper or considered by the Solvency II rules. The current Solvency II regulations set a UFR of 4.2% for every economy. This is split as 2.2% long term growth and 2% inflation. The 2% long term inflation assumption is consistent with the inflation target of most central banks, although there are various economists who argue that this inflation target (particularly for the European Central Bank) is unrealistically low and should be revised. The current Solvency II regulations assume that the UFR is the same for all economies. Again, this assumption is open to debate. Work performed for QIS5 demonstrated that there should be at least 3 different buckets of economies to reflect different long term inflation environments for example, Japan should arguably have a lower long term inflation assumption given the long period of deflation seen there, whereas Turkey has had consistently high inflation and has had a target of between 5% and 7.5% over the last 4 years. The long term real growth rate assumption is also seen to be quite different across economies over the long term. Even when considering only the second half of the 20th century (the first half being heavily impacted by high inflation due to the wars), real returns ranged from 1% in the Netherlands to 4.7% in France. However, the average return over all economies over the second half of the century is 2.3%, which gives some justification to the 2.2% assumption. So, although there is some backing to the assumptions made on the UFR, there are still arguments around the 3 The Solvency II discount rate: Nothing is simple / June 2012
appropriateness of a one rate for all approach. Some commentators (e.g. the Groupe Consultatif) have suggested that the local supervisors have the best knowledge of their respective economies and should be able to use that knowledge to derive their own calibration. But then to allow local regulators to set one assumption could open a whole can of worms, with industry arguing that all assumptions should be set by local regulators! Convergence Once we know where we start and where we should converge to, we then need to determine how that should happen and how quickly. The Economic Monetary Affairs Committee (ECON) has suggested that the risk free curve should converge to the UFR (or, at least, not be materially different) 10 years after the last liquid point. However, EIOPA have suggested that the period of time should be 40 years. Clearly these are significant differences in approach that could have material impacts on the technical provisions. For the Euro, this would mean that we would assume that after 30 years we reach the ultimate forward rate (i.e. the last liquid point is 20 years and the period to convergence is 10 years). Some observers have noted that this is too short a period of time and it is worth noting that the shorter the period of time until the UFR is reached, the more sensitive the extrapolated portion of the yield curve is to the choice of UFR. Note that the Wilson-Smith model does not specify a point when the UFR is reached instead it incorporates a speed of convergence ( ). The Solvency II default value for is 0.1. Where this does not lead to convergence (within 3 basis points) by the assumed time, will be recalibrated. Summary of basic risk free rate Much of the debate in the UK has concentrated on the adjustment to the basic risk free rate to allow for a proportion of the spread, however, there is also much debate, subjectivity and judgement in the derivation of the basic risk free rate itself. Although it is not such an issue for the UK, as the GBP swap curve is considered reliable for longer (i.e. currently to 50 years), there is still potential that this assumption could be revisited. Any companies with significant Euro denominated exposures should be keeping a close eye on these judgements and assess the impact that each one of them causes. They should also be considering if their own internal view is the same as that which will ultimately be arrived at by the European Parliament and if it is not, how they wish to manage their business. It is also worth noting that, currently, the proposal is that EIOPA will publish basic risk free curves monthly. The industry is pushing for weekly publications and, for the major currencies, daily. The reason being that companies are required to ensure solvency on a continuous basis therefore, in order to do this, they need to know the regulator provided curve on a more frequent basis than monthly. As with all other areas of uncertainty, it will be interesting to see where this debate lands but it may require companies to produce their own tools for deriving the appropriate curves. The Solvency II discount rate: Nothing is simple / June 2012 4
What should I add to risk free? As noted above, in the UK, the main issue on discount rates has related to whether there should be an allowance for a premium above risk free for certain insurance liabilities namely, those related to annuity contracts. So, why should these contracts be treated any differently to any others? The cash flows associated with annuity contracts are more predictable than those associated with other types of insurance contracts and they are easier to match with actively traded assets. Typically, gilts and corporate bonds are used to match annuity liability cash flows and these bonds will be held until maturity. The yield on a corporate bond is higher than that on a gilt (or swap) and the addition is known as the risk premium. The risk premium allows for certain risky characteristics of the bond e.g. that the issuer of the bond may default, that the issuer of the bond may get downgraded and that the bond may be difficult to turn into cash quickly i.e. it is illiquid. The holder of the bond can mitigate against the risk attached to the last portion of the premium in a simple way by holding the bond until maturity. Given this, the argument goes that the holder (i.e. the insurance company) should capture the value of that liquidity risk premium in their valuation. However, there are also other allowances within that spread which don t relate to default, downgrade or liquidity. To avoid debate about whether the premium actually relates to liquidity or other elements as well, the term liquidity premium was replaced by matching premium. More recently, it has been viewed that the term premium is potentially confusing and therefore, the latest Solvency II text refers to a matching adjustment. The matching adjustment But whatever it may be called, it looks likely that an adjustment above the EIOPA derived basic risk free rate will be allowed for certain insurance obligations (typically annuities). The battle across Europe has been to set rules to be able to quantify this adjustment fairly, openly and consistently (although it is fair to say that there are some countries/parties across Europe who are still not convinced that such an adjustment should be allowed for at all). There have been Level 2 guidelines set out in a document released only to certain stakeholders (hence not public) in October 2011 which discuss the details of when the adjustment should apply and how it should be calculated. More recently (March 2012), the proposed amendments to the Level 1 text (known as Omnibus 2), were released publicly. These largely echo the Level 2 guidelines, although there are some interesting differences. So, from what we currently know, a company can apply a matching adjustment to a defined portfolio of insurance contracts and corresponding assets where the following conditions are met: The products to which the adjustment is applied have no future premiums, no possibility of lapse (or if so, there is a surrender value such that no loss is made) and the only underwriting risks that exist are longevity and expense. Therefore, we are basically looking at immediate annuity contracts. The assets should have cash flows that are fixed and bond-like and that match the liability cash flow profile in the same currency, or such that any mismatch is not material. The cash flows of the assets cannot be altered by the issuer or other party. The insurance contracts and assets must be managed and operated separately from the other activities of the company. The assets in the assigned portfolio must have a credit quality that is above the minimum level considered to be investment grade. This final bullet is ambiguous and is a current area of uncertainty. The investment grade wording is a new addition in this text and adds uncertainty as the minimum credit rating considered to be investment grade is typically BBB-. Therefore, it may be that BBB and above is permitted, with no limits on the amount of BBB. However, if investment grade, as written, means BBB, then you would only be able to include assets of rating A or above in your assigned portfolio - or maybe BBB+! Clearly this requires clarification. In addition, there are the following requirements: The impact of applying the adjustment must be publicly disclosed The contracts to which the matching adjustment are applied must be written in the country where the company has been authorised The supervisor must approve the application of the matching adjustment These final three bullets are also new additions in the matching adjustment text. Each of them is being argued against by the industry. Once a company has determined the appropriate portfolio of assets and liabilities which meet the above criteria, they must calculate the matching adjustment itself. This is not particularly straightforward. 5 The Solvency II discount rate: Nothing is simple / June 2012
Figure 1 Gross redemption yield of assets Matching adjustment Fundemental spread Single risk free rate weighted by liability cash flows Allowance for downgrade Allowance for default OR (IF HIGHER) 75% long term average spread The basic idea is that the company calculates: The gross redemption yield of the assets matching the liabilities A single rate equal to the basic risk free curve weighted by the liability cash flows The difference between these two is the total spread above risk free on the assets. Then we have to deduct the amount of that total spread that is related to default and downgrade, and then the remaining amount is the matching adjustment. The amount related to default and downgrade is referred to as the fundamental spread. This is show pictorially in Figure 1. However, this fundamental spread is then subject to a limit it must be at least 75% of the long term average spread over the basis risk free for assets of the same class, duration and credit quality. The fundamental spread will be calculated based on additional specifications given in the Level 2 Delegated Acts. However, additional specifications are likely to include a minimum loss given default and the number of years of data required to calculate the long term average spread. Publicly available reports suggest that these are likely to be 30% and 30 years, respectively. Therefore, in order to assess the matching adjustment, a company will need to show that the cash flows match and that they meet all the other criteria listed above, and then they must split the assigned assets into those of the same duration, type and rating. Then, for each bucket of assets, assumptions for probability of default and loss given default and the loss given downgrade must be derived likely based on rating agency data. Separately, for each bucket, the long term spread must be derived. Using this, the matching adjustment can then be calculated. One of the key issues companies face is in deriving an appropriate long term spread. Typically, data on the types of bonds that insurance companies hold does not go back 30 years. Therefore, if that requirement is maintained in the Level 2 Delegated Acts (when released), approximations will need to be made to derive the long term spread. The Solvency II discount rate: Nothing is simple / June 2012 6
What all of this means is that, in times where the current spread is very close to the long term spread, the matching adjustment will be 25% of the total spread. This is significantly lower than that currently allowed for in Solvency I calculations. However, where the current spread is far above the long term spread, a greater proportion of the spread can be taken as the matching adjustment. Intuitively, this seems sensible i.e. when spreads widen, a greater proportion is due to liquidity (or, to be pedantic, non default and downgrade!) however, even on this issue there is some debate. Figure 2 demonstrates the way the matching adjustment reacts to market conditions What about when we are in times of financial stress? The Level 1 text allows for particular circumstances where EIOPA observes a stressed situation of financial markets for a given currency and that this situation, in EIOPA s view, will be temporary and exceptional and will lead to companies selling Figure 2 Q4 2007 Q1 2009 Q4 2011 Approximate matching adjustment as proportion of the total spread 20% 75% 60% Source: KPMG analysis (based on a typical asset portfolio used to back UK annuity liabilities this is based on the matching adjustment that would apply to the liabilities backed by assets that meet the requirements a large and substantial part of their fixed income securities portfolio. In these cases, EIOPA will determine and publish an adapted basic risk free rate curve. This adjustment to the basic risk free rate curve has previously been known as the counter-cyclical premium, although this terminology appears to have been dropped. The important consideration for this adapted curve is that it can only be used for certain illiquid liabilities (which is a new requirement, introduced in the most recent Level 1 text) and EIOPA will determine the adjustment. However, EIOPA will need to publish how they will calculate the adjustment and how they will determine when a stressed situation is observed. This is required so that companies can manage their solvency positions appropriately. 7 The Solvency II discount rate: Nothing is simple / June 2012
Does any of this impact the capital requirement? In a word...yes... Typically, in economic capital modelling, market practice is to assume a widening of spreads in a stress as part of the allowance for credit risk (often named spread risk). It will also be assumed that a greater proportion of that widened spread can be allocated to liquidity (or non default and downgrade). Looking at a combination of the draft Level 1 and Level 2 texts, there are three areas to consider in the standard formula: (1) the adjustment to fundamental spread in the spread risk sub module, (2) the spread adjustment mechanism and (3) the counter-cyclical premium risk sub-module. Fundamental spread in spread risk In the Solvency II spread risk calculation, for any assets where a matching adjustment is applied, the fundamental spread should not be assumed to widen as much as the total spread. This effectively means that it is assumed that a greater proportion of the total spread is allocated to liquidity. Spread adjustment mechanism In addition, the impact of spread widening will be assumed to change depending on where in the financial cycle we are much in the same way as the equity symmetric adjuster. This will mean that, when spreads are high, the spread risk capital will be lower and when spreads are low, the spread risk capital will be higher. However, there is a clear interaction between this and the adapted risk free curve and the application of the matching premium. The below table identifies when the spread adjustment mechanism applies (with references to the relevant articles of Level 1): Balance sheet treatment Capital requirement Basic risk free Use EIOPA derived basic risk free curve Can apply spread adjustment Adjusted basic risk free / CCP Matching adjustment Use EIOPA derived adapted risk free curve [Article 77a(2)] Use EIOPA derived basic risk free curve + company derived matching adjustment [Article 308c] Can not apply spread adjustment if gives a positive impact [Article 106a (4)] Can not apply spread adjustment at all [Article 308c (5)] (interpreted as meaning for those contracts/ assets where a matching adjustment is applied) This effectively says that, when a company is already taking credit in the balance sheet for a premium above basic risk free, they cannot also benefit in the capital requirement if they are using the standard formula. This is likely to make capital management slightly more difficult as there are effective offsets between balance sheet and capital requirements. Counter-cyclical premium sub-module Publicly available literature has suggested that, where a countercyclical premium applied (remember, this is now called the adapted risk free curve), it should be assumed, in stress, that this is removed. With the new terminology it is not yet clear whether a similar sub module will apply, although we expect it will. The Solvency II discount rate: Nothing is simple / June 2012 8
Conclusion There are many areas of uncertainty in Solvency II and a number of them are understandably very complex. However, the above outlines that it is not only quite complex and debatable areas that are uncertain, but also areas where, at a high level, it would seem fairly straightforward to get an agreement. Companies should be ensuring that they are up to speed with exactly what is occurring no matter how big or small they are, as these are issues that will impact everyone. There is still a long way to go until Solvency II is finalised and how much will come down to industry lobbying or political intervention is unclear but there are a number of areas that could impact the fundamental balance sheet and companies must have clarity as to what these are and the potential impacts of them. 9 The Solvency II discount rate: Nothing is simple / June 2012
Contacts If you would like more information on this report please contact the author using the details on page 1 or your usual KPMG contact. If you have any other Solvency II queries, we have a number of specialists who will be happy to assist. See below for their details. Phil Smart UK Head of Solvency II Tel: 020 7311 5134 phil.smart@kpmg.co.uk Solvency II Proposition Leads Michael Rallings Pillar 1 (Life Insurance) Tel: 020 7311 6061 michael.rallings@kpmg.co.uk Roger Jackson Pillar 1 (Non Life Insurance) Tel: 020 7694 5484 roger.jackson@kpmg.co.uk Pheng Lim Asset Allocation Tel: 020 7694 5394 pheng.lim@kpmg.co.uk Michael Crawford Pillar 2 (ORSA) Tel: 020 7311 1446 michael.crawford@kpmg.co.uk Danny Clark Pillar 3 Tel: 020 7311 5684 danny.clark@kpmg.co.uk Jon Dowie Data and Systems Tel: 020 7311 5295 jon.dowie@kpmg.co.uk Steve Liddell Programme Assurance Tel: 020 7694 5409 steve.liddell@kpmg.co.uk Brid Meaney Programme Delivery Tel: 020 7311 5470 brid.meaney@kpmg.co.uk Jane Portas Structuring and Groups Tel: 020 7311 5437 jane.portas@kpmg.co.uk Paul Merrey Strategy Tel: 020 7694 5276 paul.merrey@kpmg.co.uk Stuart Secker Tax Tel: 020 7311 5366 stuart.secker@kpmg.co.uk For Actuarial Life Insurance specific queries, please contact any of the below: Tim Roff Tel: 020 7311 5001 tim.roff@kpmg.co.uk Ferdia Byrne Tel: 020 7694 2984 ferdia.byrne@kpmg.co.uk Nick Dexter Tel: 020 7311 5443 nick.dexter@kpmg.co.uk Richard Care Tel: 020 7694 2890 richard.care@kpmg.co.uk John Jenkins Tel: 020 7311 6199 john.jenkins@kpmg.co.uk Trevor Jones Tel: 020 7311 5874 trevor.jones@kpmg.co.uk Gavin Palmer Tel: 020 7694 5885 gavin.palmer@kpmg.co.uk Simon Perry Tel: 0117 905 4080 simon.perry@kpmg.co.uk www.kpmg.co.uk/solvencyii The information contained herein is of a general nature and is not intended to address the circumstances of any particular individual or entity. Although we endeavour to provide accurate and timely information, there can be no guarantee that such information is accurate as of the date it is received or that it will continue to be accurate in the future. No one should act on such information without appropriate professional advice after a thorough examination of the particular situation. 2012 KPMG LLP, a UK limited liability partnership, is a subsidiary of KPMG Europe LLP and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative, a Swiss entity. All rights reserved. The KPMG name, logo and cutting through complexity are registered trademarks or trademarks of KPMG International. RR Donnelley RRD-270487 June 2012 Printed on recycled material.