[Skip to content]

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

Pensions: On the wrong track?

In recent years, longevity swap transactions have become reasonably common. Typically, a protection buyer pays a fixed series of premiums representing the expected liability cash flows plus a margin, for a pre-agreed set of annuitants and receives from a protection provider (usually, but not always, an insurer or reinsurer) a floating series of claims equal to the actual payments to those annuitants. These transactions are very long-dated — 60-year trades are typical.

Several commentators suggest that we may be at an inflection point in terms of demand for this product. At the same time, the credit crunch provides a useful backdrop to reflect on how these trades have been structured in the past and whether certain tweaks should be considered going forward. While these ultra-long swaps clearly provide a close match to the underlying liabilities being hedged, they lead to an unacceptable and poorly understood increase in counterparty credit risk.

Probability of default
A crucial first piece of analysis is to try to estimate the probability of default of protection providers. One can infer the annual probability of default from current credit default spreads — the cost of protecting against bond default. The probability of default is simply the credit default swap (CDS) spread divided by the expected loss in the event of default. Using loss given default of 50% of face amount as a crude estimate, one might conclude that a counterparty whose CDS trades at 2% per year has an annual probability of default of around 4%. Note that public CDS quotes are not available for the smaller, unrated or new players and therefore these firms are excluded from the analysis.

Now look at the typical protection providers active in the market — we do so on a no-names basis, but the analysis can be reproduced for a particular name by referring to publicly available information on Bloomberg or similar data providers. Three years ago, typical protection providers’ CDSs traded at 0.09% with modest differences from name to name. In other words, one would pay £0.90 to protect against a default costing up to face value of £1000. Thus the market ascribed an annual probability of default of about 0.18% and annual probability of survival of 99.82%. Using simple compounding, one might estimate a 98% chance of survival after 10 years and 90% after 60 years. So a ‘set and forget’ approach for long-term hedges was quite defensible — counterparties were not expected to go bust.

Post-credit crunch, the picture has changed substantially. Today’s CDS spreads on the main participants in the longevity market are between four and 28 times wider than they were in 2006. The result is the current market suggests that, for protection providers with strong credit, the probability of surviving for 10 years is 93% and for 60 years 64%. For weaker protection providers (but still household names), the probability of surviving for 10 years is 59% and for 60 years virtually nil.

Note that CDS probabilities contemplate formal, legal default. Doubtless a far more likely scenario is substantial credit downgrade, but not default, and the protection provider being put into run-off with potential deterioration in quality of service and concomitant deterioration in the health of the protection buyer’s risk manager who now needs to worry about the counterparty daily. That’s not quite as bad as an actual default, but it remains highly sub-optimal.

It is also worth noting that the default probabilities suggested by CDS spreads are, in general, at their lowest level in the past 12 months — a few short months ago, the probabilities of survival would have been lower. The market is therefore saying that most longevity swaps transacted to date are more likely to default than run to term. Longevity swaps can no longer be perceived as ‘set and forget’ hedges.

Does collateral solve the problem?
Some will argue that it does not really matter if a protection provider suffers a credit event if an appropriate collateral infrastructure is in place. In theory, that may be so. In reality, many protection buyers will be sorely disappointed. The Lehman and AIG sagas show how collateral mechanisms can fail even for liquid derivative instruments whose market price is unambiguous. Illiquid trades, such as those based on longevity, introduce significant incremental complexity.

For example, many of the longevity swaps to date have contemplated an annual marking to model. The marking typically requires some form of actuarial consensus of fair value to be reached. A variety of escalation processes are contemplated in order to ensure lack of bias in fair value. This is reminiscent of Nero fiddling while Rome burns. It can take years for the process to fully conclude, namely culminating in negotiated commutation and/or forced novation, especially if one counterparty is behaving in an awkward or unhelpful manner. The reality is that stressed credits can become defaulted credits well before the natural conclusion of such processes.

Worse is the fact that the contracts typically seek to reflect an amorphous actuarial fair value, rather than price. If there is substantial deviation between the two, the protection buyer will find itself paying up, potentially significantly, in order to replace the hedge. Pricing dislocations should be expected in the longevity market in years to come where there are few potential counterparties, many of whom have similar risk profiles. One example would be UK-regulated life assurers whose capital positions, and therefore pricing, for any annuitant risks can be expected to change substantially post- Solvency II. Another example is reinsurance counterparties whose return on capital expectations is a function of hard and soft market cycles. Traditional collateral mechanisms may break down quite easily, especially in times of stress.

What is the solution?
In time, a trading market in longevity may develop. Contract standardisation, transparency and liquidity will help mitigate the problems described above as it becomes easier for the market to value longevity risk in a consistent manner. But it may be quite some time before this happens. In the meantime, the market should consider shorter-dated transactions with a fixed contract term and a maturity value calculated using a commutation methodology whose inputs are deterministic and/or observable. This approach also allows collateral to be calculated more regularly during the transaction and enables all sides to easily quantify exactly how much money they are expecting to make or lose.

The most common objection here is that using a pre-defined formula erodes the efficiency of the hedge. However, a cleverly constructed commutation factor based on observable data and deterministic calculations has a surprisingly high degree of correlation — over 90%. The unspoken disadvantage of this approach, and one that I believe is a bigger obstacle than the maths described above, is that it catalyses uncomfortable questions about credit events and the use of forwardlooking extrapolation methodologies at contract inception. There is no question that it is much easier to avoid these issues when negotiating a contract with a view to ‘agreeing to agree’ a solution if and when a counterparty’s credit deteriorates. But this is simply poor risk management.

It is also worth noting that significant anecdotal evidence suggests that shortening transactions will substantially increase demand for longevity risk, particularly among capital market participants who are interested in taking the risk but for shorter maturities. This should eventually help to foster more competition with potentially more attractive pricing available.

The credit crunch has helped to put into sharp focus the difficulties of structuring long-term illiquid hedge contracts. The market is currently telling protection buyers to expect many hedge counterparties to default. Those protection buyers who have existing contracts should monitor them carefully and, if appropriate, renegotiate terms. Those who are considering hedging now should consider shorter-dated structures. In essence, it is better to be very confident of a hedge which is highly, albeit not 100%, effective than to have a ‘perfect’ hedge which fails to perform when really needed.


Eugene Dimitriou works in RBS’s global banking and markets division