Longevity risk: Is the LAGIC stress margin fit for purpose?

Longevity risk is often described simply as the risk that individuals live longer than expected. For insurers, reinsurers and retirement income providers, longevity risk is, however, much more complex. To define it very succinctly, it is the risk of misestimation of best estimate assumptions on a benefit contingent on survival.

In order to capture that risk, the Australian prudential framework LAGIC LPS 115 prescribes a "longevity stress", before adjustment for diversification, of a 20 per cent decrease in the best estimate mortality rate for each age from the reporting date for the remaining duration of the liabilities.

While such liabilities were inconsequential and longevity risk was mostly hedged by large exposure to mortality risk, it made sense to prescribe a one-dimensional stress to keep the risk charge calculation simple. However, as exposure to longevity risk has grown over the last few years and is expected to keep growing in the context of the Retirement Income Covenant, there is a question around what level of exposure should trigger a more mature risk calibration approach.

Longevity risk framework best practice explained

Under the Solvency II framework, even relatively simple internal models recognise that longevity risk is not a single uncertainty but a combination of several distinct risk drivers. At a minimum, firms usually distinguish between base mortality risk and mortality improvement (trend) risk.

Base mortality risk reflects the possibility that the current level of mortality assumed in the best-estimate liability calculation is incorrect. In practice, insurers may break base mortality risk down further, for example, into:

Data risk – the possibility that mortality assumptions are derived from incomplete, poor-quality, or non-representative datasets.
Volatility risk – the random deviation of actual deaths from expected levels due to sampling variability in finite portfolios.
Selection or take-up risk – particularly relevant where policyholders have options (for example, to take a lump sum or to purchase lifetime income). The realised pool of policyholders may differ from the population assumed at pricing, leading to adverse selection.

Under internal models, trend risk then reflects the uncertainty surrounding how mortality will improve in the future. History shows that improvements do not occur at a steady pace.

When trend risk is calibrated using general population data (as in most cases), models often allow for the possibility that mortality improvements in the insured population may diverge from those observed in the wider population. This adds a sub-risk to the trend risk component of longevity risk.

For retirement income providers, this uncertainty is particularly important because even modest errors in trend assumptions can have a large impact on the value of lifetime income liabilities. Both stochastic methods calibrated on historic data and scenario-based methods are typically used to calibrate trend risk.

In the defined benefit (DB) pension world, another demographic uncertainty comes into play: marital status risk. Are pensioners married or single? If they are married, how old is their spouse? These questions matter because survivor benefits can extend pension payments well beyond the life of the member.

In practice, the information is not always known with certainty. In some schemes, the existence or age of a partner may only become clear when the member dies and the surviving spouse calls to claim a reversionary pension. This creates a modelling challenge: the proportion of members with spouses at retirement is not the same as the proportion of reversionary pensions actually paid. After all, the spouse must still be alive at the time the member dies. This is where conditional probability theory becomes useful.

All these issues observed across the global longevity market offer a peek at what longevity risk really looks like on a portfolio of lifetime income streams. It is far more complex than simply applying a 20% stress to mortality rates, as is done in simplified regulatory approaches.

As the Australian Treasury’s Guidance on Best Practice Principles for Retirement Income encourages the development of lifetime income product propositions, it is important to draw on this body of thinking about how longevity risk behaves in practice and how it should be managed in different contexts.

Yet, as the market evolves, we are not convinced the industry is fully grappling with what longevity risk actually looks like in the emerging landscape. The next part of this article turns to how longevity risk manifests in a market increasingly shaped by superannuation fund–led lifetime income solutions.

What longevity risk looks like for a super fund-issued lifetime income stream

The newer retirement income products introduced into the Australian market are largely investment-linked lifetime income streams, albeit with some variations in design. The core idea behind these products is to allow members, working with their financial advisers, to select an investment strategy in retirement that aligns with their risk appetite while still providing protection against longevity risk.

In principle, this approach offers an opportunity to move away from the traditional practice of shifting large savings abruptly into a fully de-risked annuity. From an investment perspective, such a sudden and complete asset de-risking at an arbitrary point in retirement can be difficult to justify.

For these products, the accumulated underlying assets continue to be managed by the superannuation fund and the investment risk sits with the member, i.e., no investment guarantees are provided, so the member's annual income varies subject to investment performance.

Longevity protection is provided by a life insurer, who undertakes to top up the account balance used to generate income, ensuring payments continue until the member’s death. However, because the level of income fluctuates with investment performance, there is uncertainty around the eventual value of the benefit the life insurer will ultimately provide.

This contrasts with a traditional life annuity, where the level of income is typically known in advance. With investment-linked lifetime income streams, the insurer’s exposure depends on the path of investment returns and the resulting income levels over time.

Any allowance for this uncertainty does not need to be explicitly identified in the current APRA LAGIC longevity stress methodology. Indeed, in this case, the asset risk charge and the insurance risk charge should not diversify linearly when combined. Under Solvency II, this is a case of non-linearity, that is a failure of the standard variance-covariance diversification method to accurately capture complex risk dependencies.

Developing a capital basis for the interaction between investment and mortality risk for investment-linked retirement income products with longevity protection is complicated. The leveraging effect of investment returns being higher than assumed and mortality being lower than assumed is not reflected in just applying the LAGIC 20 per cent statutory mortality stress.

As previously stated, this simple capital stress already doesn’t allow for the uncertainty of future mortality improvements, let alone the added complication of uncertain benefit payments due to investment return volatility. Best practice methodologies for assessing these various risks include stochastic modelling of the risks and their interactions to develop an appropriate capital stress.

What this means for the industry

As the retirement and longevity markets in Australia continue to grow, companies should apply a more principles-based capital framework to ensure appropriate levels of capital are held for the risk. At present, the longevity capital frameworks being used for new retirement income products are not as robust as will be required when the market is more material.