06 Aug 2019 Defined Contribution Pension Retirees and Longevity Risk

It is hardly news to note that a growing proportion of retirees belong to Defined Contribution (rather than Defined Benefit) pension schemes, and that this proportion will continue to grow in the coming years. These retirees have considerable freedom and flexibility in the investment strategy and income drawdown decisions that they implement in retirement (and, in the UK, the abolition of compulsory annuitisation in 2015 has clearly furthered this flexibility). Today, very little of new retirees’ pension pots are invested in a form of annuity, or indeed any other asset that provides some form of longevity protection. If the retiree lives longer than he or she anticipates, the pension asset pot may therefore prove insufficient to sustain the ongoing living costs of the retiree. There is substantial uncertainty in the longevity of an individual retiree, and this risk may therefore be highly consequential.

Thinking about Longevity Uncertainty

When considering the uncertainty in the longevity outcome of the individual retiree, it can be useful to think it in about a couple of components:

• There is some significant uncertainty in how mortality rates of the population (or relatively homogenous sub-sets of the population) will change over time, especially over a period of decades. The mortality tables and improvement factors that actuaries use in their long-term projections can only ever be fallible estimates of the future. Society, and its technologies, economy and government policies are all constantly changing in unpredictable ways that have complex implications for the long-term changes in life expectancies of different groups. No one can therefore be particularly sure what the mortality rate of, say, females age 85 will be in 2040. This is, of course, why life assurers and annuity providers are required to hold regulatory solvency capital in respect of the mortality and longevity risks on their balance sheets.
• An individual faces a second, and much larger, form of longevity uncertainty – even if the above difficulties with long-term prediction are overcome such that the future collective mortality experience is known with certainty, the experience of the individual life is still subject to very significant statistical variation or ‘sampling error’. When a mortality rate of 0.01 has been experienced by a collective of lives in a given year, no one died 0.01 times. For every 99 people who lived to the end of the year, one unlucky person has died. As a result of this form of uncertainty, a male age 65 with a life expectancy of around 17 years may also have a 10% probability of actually living 10 years longer than that and making it to 92. And he may have something like a 1% chance of making it to 100 and beyond. This presents a very significant financial planning risk for individual DC retirees.

Transferring and Pooling Longevity Risk

The obvious and traditional solution to this longevity risk problem is to invest the retirement pot in an immediate annuity. This fully transfers both of the above types of longevity uncertainty from the individual to an annuity provider. But this has fallen out of favour with DC retirees for a variety of reasons, which include: the annuity is highly inflexible and largely irreversible; it reduces inheritance proceeds in the event of an unexpectedly early death; the current very low interest rate environment makes an annuity significantly more expensive than has been the case in the past (though that point also applies to virtually all financial assets today).

So, what is to be done? One potentially attractive compromise solution could be to invest, at retirement, a portion of the retirement asset pot in a deferred annuity with a start date that is, say, 20 years from the retirement date. The deferred annuity would act as a form of longevity tail risk hedge for the retiree. It would mitigate the financial consequences of living much longer than expected, without tying up all of the retirement assets today. The remaining retirement asset pot could be managed to produce income until the end of the deferral period.

This deferred annuity approach has not, however, been widely used by DC retirees to date. This is, at least to some degree, because deferred annuities, like immediate annuities, seem very expensive today. And this is partly an inevitable consequence of the impact of very low interest rates on the cost of very long duration assets. Life assurance companies have also pointed out that deferred annuities are very longevity-capital-intensive under Solvency II, and this puts further upward pressure on the pricing of these annuities.

I am not a longevity risk expert and I do not have a considered view on the degree of severity of the Solvency II longevity Solvency Capital Requirement (or the impact of the Risk Margin on annuity technical provisions). But as we noted above, the deferred annuity is a form of longevity risk tail hedge, and, as such, it is quite natural that it generates significant longevity risk for the insurance company and that this generates a correspondingly significant capital requirement.

Finding a Longevity Risk Transfer Solution that fits the Customer Need

Let’s take a step back here and consider what is fundamentally driving the longevity risk capital requirement of the deferred annuity before we (rightly or wrongly) shoot the Solvency II messenger. The insurance company is holding longevity risk capital to support the risk that the collective longevity experience of a large pool of annuitants is longer than is assumed in the pricing of the product. If an insurer writes deferred annuities on 10,000 different males age 65, there is a risk that the actual average longevity outcome is, say, 3 years longer than the actuarial best estimate basis at the time the product was written (this is the approximate scale of the stress assumption required in the Solvency II Standard Formula longevity risk capital calculation).

This risk arises because of the inherent fallibility in the forecasting of long-term changes in population life expectancy. But that is not the risk that the DC retiree cares about! The newly-retired DC member is not particularly worried about everyone living, on average, 3 years longer than currently expected. That is not the main source of his or her financial risk. He or she is worried (or should be worried) about the financial consequences of him or her as an individual living to 92 or 102. This is of much greater consequence than the actuarial forecast of 82 turning out to be 2 or 3 years shorter than the actual collective experience. It is the second category of longevity risk discussed above, rather than the first, that is the individual’s primary problem. But that is a risk that the insurance company can pool and (largely) diversify away…it therefore does not need to hold a material amount of longevity risk capital to support that risk!

This suggests that the problem with the deferred annuity is perhaps not (only) that it is too capital-intensive, but that the capital requirement is being created by transferring risk that the potential annuitant retiree isn’t really concerned about transferring. A more capital-efficient, and therefore presumably more cost-effective, longevity protection product would be one where the annuitant bears the risk of changes to collective mortality experience (which may vary by, say, 3 years and is highly capital-intensive for the insurer), whilst transferring the idiosyncratic risk of their individual experience to the insurance company (which may vary by 10 or 20 years, and is highly capital-light for the insurer).

With such a product, when the overall actual mortality experience of the group is heavier (lighter) than the original pricing basis, the level of annuity payments will be adjusted upwards (downwards) over time, but each annuitant will receive their annuity payment until their ultimate death. The limiting case of this idea is a tontine-type approach. In a tontine, the starting annuity income is based on the assumption that everyone lives forever, and survivors’ annuity incomes then rise with every death that occurs. But there is no reason why the starting annuity income couldn’t be based on a more sensible mortality basis that is still more prudent than the actuarial best estimate. For example, if we used the mortality basis of the SII longevity stress to determine the starting annuity income level, the product would not generate any initial longevity stress risk capital for the life assurer. Annuity income would then be increased as and when the emerging actual annuity mortality experience proved heavier than the prudent assumption of the longevity stress basis. This is similar in concept to a with-profit annuity. The Collective Defined Contribution (CDC) initiative provides another mechanism that could allow a retiree group’s mortality experience to be pooled and shared whilst providing individual longevity protection.

There are many product and solution variations along this theme that could be explored – for example, is the starting annuity income a guaranteed minimum (which therefore could consume longevity capital in the future) or can the annuity income be reduced if actual mortality experience proves lighter than the starting pricing basis; is the longevity risk pooling across a specific closed pool of annuitants, or by reference to a wider population or longevity index; is the product offered in both immediate and deferred formats; etc.

The specific mechanisms to implement this form of collective longevity risk-sharing in a practical and appealing way would be a matter for insurance companies’ product development and marketing teams. It would inevitably result in a more complicated product than a vanilla immediate or deferred annuity. But it would be an innovation that could better meet customer needs and provide a more cost-effective and much-needed individual longevity risk solution.

Acknowledgments and Disclaimer

I would like to express my thanks to John Taylor, with whom I have recently had interesting discussions relating to this topic. All views and any errors are my own. This article is written in a personal professional capacity as a Fellow of the Institute and Faculty of Actuaries, and the views expressed herein are not intended to represent those of my employer.