Towards Optimising the Decumulation of Retirement Savings

This article summarises findings from our recently published Paper that reviews risks faced in the decumulation phase, and considers the problem of finding a more optimal strategy for the decumulation of retirement savings.^[1]

We introduce an innovative decision framework that allows for a retiree’s differing appetites for liquidity and investment risks.

Exploring liquidity versus investment risks helps to map optimal decumulation strategies in two dimensions of risk, suggesting that there is no one-size-fits-all strategy for every retiree.

Almost all retirees face the problem of deciding how to best spread, or “decumulate”, their hard-earned savings for consumption over the remainder of their lives.

In general, the retiree seeks to convert their savings into a desired standard of living (SOL) over retirement by choosing a strategy under a complex web of preferences, risks and constraints. Nobel laureate William Sharpe famously coined decumulation to be the “nastiest, hardest problem in finance” ^[2], attributing the difficulty to the sheer dimensionality of the problem.

To help retirees navigate this complexity, the Australian Government implemented the Retirement Income Covenant (RIC) for superannuation trustees^[3]. The RIC emphasises the need to assist retirees in managing risk and liquidity needs in addition to sustaining a high SOL in retirement. We introduce a framework that is consistent in spirit with this aspect of the RIC.

In our research, we seek a strategy that best satisfies some specified retirement plan and is in agreement with the retiree’s (potentially differing) appetites for the risks to this retirement plan. Hence, we need (i) a retirement plan, (ii) a set of risks we want to consider, and (iii) a decision framework to best achieve (i) subject to (ii). This involves making a choice between predefined decumulation solutions.

We consider a retirement plan which aims to maintain a chosen SOL, while keeping flexibility to adjust spending as liquidity needs are incurred over time. The retirement plan has two features:

1. The retiree sets a “consumption target” in real terms to be sustained over time, at least until first needing liquidity (e.g., long-term care, deposit for children’s home, bequest).
2. The retiree sets aside a portion of retirement savings as a “liquidity reserve” intended to support future liquidity needs, such as bequest (upon death) or long-term care (before death). The liquidity reserve will deliver the lowest adequate bequest, provided the consumption target can be maintained until death.

In summary, the retiree plans to sustain their consumption target over time without having to draw upon their liquidity reserve, at least until first needing liquidity. We allow the retiree to have differing appetites for two very different risks associated with their retirement plan:

1. Investment risk: the retiree may be averse to the prospect of running out of savings due to an unfavourable sequence of investment returns. For modelling purposes, we consider the retiree to be a mean-variance investor in relation to bequest, based on a degree of investment risk “aversion”.
2. Liquidity risk: the retiree may be intolerant towards the prospect of having insufficient savings to fund liquidity needs before or upon death (e.g., long-term care or bequest, respectively). For modelling purposes, we consider the retiree to set a threshold limit on the probability of liquidity shortfall (i.e. not having enough liquidity), which we refer to as liquidity risk “tolerance”.

Now that we have set a retirement plan and discussed the associated risks, we can describe how the optimal decumulation strategy is determined. We allow the retiree to bundle a basic strategy of account drawdown with one choice of lifetime income stream from the “Annuity Family”, which we define as a suite of annuity-like strategies with some common features:

• Savings allocated to the Annuity Family are inaccessible after purchase.
• Payments from the Annuity Family are expected to be stable for life in real terms.

However, the Annuity Family is designed such that each member delivers a total return with significantly different risk exposures and protection guarantees, as described below and summarised in the table:

• An annuity provider grants a “longevity protection guarantee” when future mortality crediting rates are guaranteed at the time of purchasing the policy. These mortality crediting rates reflect a form of illiquidity premium in excess of the return on invested assets.
• An annuity provider grants an “investment protection guarantee” by investing assets in inflation-indexed bonds to earn a risk-free return in real terms.

The retiree will have to pay a risk-adjusted price for a longevity protection guarantee and/or forego the market risk premium for the investment protection guarantee.

In other words, the retiree must forego some expected returns in exchange for any protection guarantees purchased, which illustrates the trade-off between consumption and protection that is at the heart of the decumulation problem.

Finding optimal strategies in the Australian context

We implement our framework in the Australian context by calibrating our assumptions accordingly and producing the following results. 

We consider a 67-year-old homeowner with $595,000 of retirement savings, as recommended at June 2023 by the Association of Superannuation Funds of Australia (ASFA) for a comfortable standard of living^[4].

As a simplification, we did not consider the impact of the age pension in our own analysis. For this retiree, we considered modest and comfortable standards of living (SOL) defined by ASFA, as well as the famous 4% withdrawal rate popularised by William Bengen as historically “safe”^[5].

In the illustrations below, we only considered a liquidity reserve of 20% of retirement savings, which is permitted to grow with interest over time. This choice is arbitrary, and in the Paper, we also briefly considered 0%.

Note that to map the optimal strategy in two dimensions, we made strong economic and mortality assumptions to be able to produce “smooth” results at acceptable run times. We assumed a constant rate of population mortality throughout retirement, as calibrated to the Australian Life Table 2019-21 published by the Australian Bureau of Statistics, paired with the longevity improvement factors published by the Australian Government Actuary. We randomised this constant rate of mortality to reflect known estimation uncertainty around these improvement factors. Finally, we assumed real investment returns follow Brownian motion, calibrated to historical ASX 200 returns and the inflation target set by the Reserve Bank of Australia.

The decision to purchase from the Annuity Family

For each of the above consumption targets, we map the optimal decumulation strategy in two dimensions of risk, as determined by our decision framework. This will either be a basic strategy of “pure” account drawdown (in purple), or some combination of account drawdown with one member of the Annuity Family defined above.

We make some key observations from the above plots:

• As the consumption target rises (moving from left to right), the retiree requires more liquid savings to fund consumption and has less leftover savings to purchase protection guarantees. Hence, pure drawdown becomes more optimal especially given greater appetite for risk (lower right of each plot), and there becomes no ability to satisfy lower levels of liquidity risk tolerance (the blank area in the right most plot).
• As the consumption target falls (moving from right to left), the retiree requires less liquid savings to fund consumption and has more leftover savings to purchase protection guarantees. Hence, strategies from the Annuity Family become more optimal especially given lower appetite for risk (upper left of each plot), most notably at higher levels of investment risk aversion (as explored further below).

The optimal allocation of retirement savings

We also tracked: (i) the overall proportion of accumulated savings allocated to the market, or to the risk-free asset (inflation-indexed bonds); and (ii) the proportion of savings allocated to a drawdown facility, or a member of the Annuity Family.

In both cases, those proportions add up to 1. Below, we take a liquidity risk tolerance of 0.2 and analyse the optimal allocation of retirement savings for varying levels of investment risk aversion. Note that these are essentially analysing the vertical “slice” at 0.20 in the above first two maps.

As investment risk aversion rises in the above plots:

• There is a gradual allocation of wealth away from the market to inflation-indexed bonds, reflecting an uptake of guaranteed investment protection. We note this uptake is faster under lower consumption targets where the retirement plan is less restrictive.
• There is a growing preference for annuities, with the retiree substituting mortality credits for the market risk premium to avoid jeopardising the retirement plan with a lower expected return on savings.
• For the Bengen 4% Rule, the retiree will eventually purchase guaranteed longevity protection (green IIA) to further decrease their variance of bequest. We might expect this “jump” (from yellow to green) to smoothen if we were to permit the simultaneous purchase of more than one type of annuity from the Annuity Family.

Concluding comments

We find that the retiree’s optimal choice of overall decumulation strategy will depend on their consumption target and appetites for different risks, indicating that there is no one-size-fits-all strategy for all Australian retirees. For instance, a retiree with $595,000 who targets the Modest SOL might be better off by partially annuitising via a GSA. However, the median superannuation balance is actually well below $595,000 at retirement, forcing many retirees to pursue higher consumption targets relative to their balance where pure drawdown becomes more optimal. The observed tendency towards pure drawdown in Australia might therefore not be so irrational after all.

We also illustrate that there is value in integrating annuities with market-linked strategies. The features of a retirement plan will limit the retiree’s ability to purchase protection against the risks of decumulation. Considering the illiquidity premium of annuities (via mortality credits) simultaneously with the market risk premium may allow the retiree to better adjust their allocation of savings to their appetite for the different risks of decumulation. We knew this intuitively already, but we offer a quantitative framework that describes this explicitly.

Of course, our results are limited by the assumptions required to produce inferable results, such as excluding the impact of the age pension. More generally, they exclude some of the key individual circumstances of a given retiree, and hence should not be construed as general investment advice. That being said, our results are broadly consistent with observed behaviours in the Australian market, which lends support to the validity of our framework. We also believe our framework presents a relatively effective way to map and communicate options to someone who is about to retire, as it showcases the trade-offs across different strategies in rather simple terms.

References

^[1] Our Paper can be downloaded for free from  https://link.springer.com/article/10.1007/s10203-024-00448-y  – Optimal strategies for the decumulation of retirement savings under differing appetites for liquidity and investment risks. Avanzi, B and De Felice, L. s.l. : Springer, Decisions in Economics and Finance, 2024.

^[2] Sharpe, W. Tackling the ‘Nastiest. Hardest Problem in Finance’. s.l. : Bloomberg, 2017.

^[3] APRA and ASIC. Implementation of the retirement income covenant: Findings from the APRA and ASIC thematic review. 2023. Report 766.

^[4] ASFA. The ASFA Retirement Standard. [Online] June 2023. https://www.superannuation.asn.au/resources/retirement-standard/.

^[5] Determining withdrawal rates using historical data. Bengen, W. 1994, Financial Planning Association Journal.