ArticlePDF Available
The ‘Right’ Level for the Superannuation Guarantee:
A Straightforward Issue by No Means
11 January 2020
Gaurav Khemka, Yifu Tang and Geoff Warren*
College of Business and Economics
The Australian National University
We deploy a stochastic life-cycle model to examine how differing levels of the superannuation guarantee
(SG) impact on the welfare of individual Australians under existing superannuation, tax and pension
eligibility rules. Our main focus is the effect of various assumptions on the optimal SG, emphasising the
role of income and the retirement objectives of the individual. The analysis supports estimating the gains
and losses from changing the SG for various individuals, and associated impacts on net government
revenue. We find the optimal SG to vary substantially with income and objectives. While our baseline
analysis indicates a SG of below the current level of 9.5%, higher estimates emerge if access to the Age
Pension is excluded, and if the SG is used as a mechanism to self-insure against living to a very old age,
being forced into early retirement, or incurring lower investment returns. We conclude that the case for
raising the SG above 9.5% depends on the underlying assumptions, with the policy objectives that the
SG is intended to achieve being critical.
* Corresponding author: Email, Ph. +61 411 241 091
Acknowledgements: We thank the following for their valuable comments or assistance: Anthony Asher, Jenni
Bettman, Nathan Bonarius, Adam Butt, Bruce Chapman, Ross Clare, Brendan Coates, John Evans, David Haynes,
Tim Higgins, William Lim, Aaron Minney, Stuart Mules, Andrew Podger, John de Ravin, David Service, Bruce
Thomson, Guy Thorburn and Zili Zhu. We also thank participants in the ANU RSFAS Actuarial Brown Bag on
4 November 2019, the Australasian Actuarial Education and Research Symposium in Melbourne on 28-29
November 2019, and the ANU RSFAS Research Camp on 4-5 December 2019.
Page 1
Table of Contents
1. Introduction 2
What Our Analysis Reveals 2
Considerations for Whether to Raise the SG 4
Assumptions Needed for an SG of 9.5% or Above 4
Our Own View 4
2. Analysis Set-Up 5
The Model 5
Member Objectives and Utility Function 7
Income Levels 8
Scope of the Analysis 9
3. Baseline Results – Optimal Superannuation Guarantee 12
4. How Member Welfare Changes with the Superannuation Guarantee 14
5. Net Impact on the Government Budget 16
6. Sensitivity and Scenario Testing 17
Utility Parameters 20
Required Income Stream 21
Investment Assumptions 23
Policy Environment 24
Scenarios 26
7. Deciding on the SG Rate 27
8. Conclusions 29
References 30
Page 2
1. Introduction
Whether the superannuation guarantee (SG) should be increased from the compulsory rate of 9.5% is currently a
topic of much debate (see for example: Daley and Coates, 2018; Mercer, 2019; Rice and Bonarius, 2019). There
are two dimensions along which the discussion often falls short. First, is a tendency to focus on outcomes during
retirement, and the question of what needs to be put aside to ensure an adequate lifestyle once retired. This
overlooks the fact that saving through the SG can come at a cost to the individual, to the extent it reduces the
amount available for pre-retirement consumption (see Evans and Razeed, 2019) or saving via other mechanisms.
Second, analysis of how the SG impacts on the welfare of individuals (denoted fund ‘members’ here) could be
improved in a number of respects. It typically focuses on median outcomes, or projected outcomes for selected
fund members either across income levels or through cameos. Further, deterministic assumptions are often
employed regarding asset returns, investment strategies and drawdown strategies are often employed. The
landscape is more dynamic and complex than these simple formulations. Investment returns are uncertain,
members can react to changes along the path, and (most importantly) members may differ along a large number
of dimensions. Finally, existing analysis is quite hazy around how welfare should be defined and measured. The
objective function for individual members remains an open issue, as does the policy objective of the SG itself.
We tackle these issues head-on through evaluating a range of SG levels using expected utility as a metric in the
context of a stochastic life-cycle model that embeds the key rules under which the Australian superannuation
system operates. Basing the analysis around such a model offers two main advantages. First, it condenses the
temporal and investment risk trade-offs into a single measure. Saving via the SG entails shifting consumption
from pre-retirement to post-retirement, while exposing the member to uncertainty related to investment markets.
Utility functions allow a ‘score’ to be attached to all possible outcomes across each period and through time,
which can be added up to deliver overall expected utility. If a particular SG leads to higher expected utility, then
it implies that the member is better-off placing that money into superannuation after taking into consideration
both the time and risk dimensions. The helps to address what is a complex problem that is hard to reduce to simple
metrics such as shortfall measures (see Butt and Khemka, 2015). Second, we use utility functions to investigate
how the optimal SG may vary with member objectives. Specifically, we examine the relation between desired
post-retirement consumption and pre-retirement consumption, as well as a range of other assumptions. Our
analysis uses reference dependent utility functions inspired by prospect theory, under which members target either
a replacement rate or one of the ASFA retirement standards (ASFA, 2019) during the post-retirement phase.
Our model is designed to analyse the trade-off between pre-retirement and post-retirement consumption from the
member’s perspective. It assumes that the SG leads to lower take-home pay and hence reduces pre-retirement
consumption, but then adds to post-retirement consumption which is exposed to random investment returns and
uncertain time of death. The member maximises their expected utility of consumption over their lifetime, subject
to existing Australian rules governing tax, superannuation and the Age Pension and related supplements. We
initially conduct a baseline analysis that estimates the optimal SG across differing income levels and three
objective functions. We then investigate the impact of altering various assumptions, including: the availability of
the Age Pension; the assumption that the member saves to self-insure against the possibility of living to a very
old age or retiring early; the cost of the SG being partly borne by the employer rather than the member; various
investment and drawdown strategies; differing asset returns; and changes to risk tolerance. This allows us to
identify what matters for determining the appropriate SG, and how the optimal SG varies across differing types
of members depending on their income levels and objectives.
What Our Analysis Reveals
The main message is that there is no one-size-fits-all optimal SG, which in turn can vary substantially with
assumptions. Our analysis generates a wide range of optimal SGs. This highlights that a single SG rate is a blunt
instrument being applied against a background of significant member heterogeneity, along with a marked
sensitivity to assumptions. The most influential variables include: assumed member objectives; income level; the
availability of the Age Pension; whether the aim is for the member to self-insure against risks related to early
retirement, longevity and/or investment returns; and whether the cost of SG is borne by the employer or comes
out of lower take-home pay for the member. Other variables impacting on the optimal SG by up to a few
percentage points include time discounting, and how the post-retirement consumption target and drawdown
strategy are characterised. The most influential variables matter in the following way:
Page 3
Member objectives – We see the most relevant objectives as those where the member aims to save enough to
support a post-retirement consumption target that sits below pre-retirement levels, subject to providing for a
basic level of minimal income. We hence focus on ASFA modest at low incomes, and the replacement rate
and ASFA comfortable objectives at medium-to-high incomes. Our baseline analysis generates an optimal SG
that ranges from 3.5% to 9% under the replacement rate and ASFA comfortable, but only 2.5%-3% under
ASFA modest. These results emerge because an SG below 9.5% makes sufficient headway towards the post-
retirement targets once the impact of SG contributions on pre-retirement consumption and the presence of the
Age Pension and supplements are taken into account. Further, an SG of 9.5% or above is indicated if ancillary
objectives are included of becoming a self-funded retiree, or self-insuring against some combination of
investment, longevity and early-retirement risk.
Income – Income interacts with the Age Pension, which is more valuable for lower income members and thus
becomes less influential as income increases. Income also matters more when the objective is to attain a fixed
post-retirement consumption target as per the ASFA standards, under which it is optimal to just save enough
to reach the target after accounting for the Age Pension. The net consequence is that the optimal SG under the
ASFA standards is much higher for ASFA comfortable than ASFA modest and declines with income. For
instance, our baseline optimal SGs fall from 9.0% to 4.5% under ASFA comfortable in moving from an income
$60,000 up to $120,000. Under a replacement rate objective where the post-retirement consumption target
scales up and down with pre-retirement income and hence consumption, the optimal SG rises with income
largely due to a decreasing contribution from the Age Pension. In this case, our baseline optimal SGs rise from
3.5% to 7.5% as income moves from $60,000 to $120,000.
Assumed role of the Age Pension – The availability of the Age Pension and related supplements looms large
in our results given that it provides a substantial head start towards securing any target as well as providing a
hedge against investment losses. We find that estimated optimal SGs increase substantially if the Age Pension
is excluded, exceeding 12% in the majority of cases. It hence matters whether the Age Pension is viewed as a
perennial income source that is openly available to all, versus a safety net such that the implied aim of the SG
is to support members in becoming self-funded retirees and hence avoiding relying on the Age Pension.
Hedging against risks Three key risks faced by members are investment risk, longevity risk and early
retirement risk. To gauge the impact of self-insuring against these risks, we run the analysis under
combinations of lower investment returns and the member saving enough to support consumption in case they
happen to live to age 102 or retire early at age 62. The results indicate that an SG of 9.5%, if not 12%, might
be supported if the SG is used for risk hedging. The case where the member retires at age 62 also proxies for
career breaks, which have a similar effect in that the loss of income both reduces contributions and creates a
need to fund consumption out of savings.
Issue of ‘who pays’ – There is considerable debate over the extent to which the SG has been associated with
lower take-home pay for the member, and the evidence and opinions seem mixed (see Evans and Razeed,
2019; Stanford, 2019). We address this issue by assuming that the SG comes entirely out of lower take-home
pay for the member as a baseline, then conducting sensitivity analysis to gauge the impact where a portion of
the SG is borne by the employer. This raises the optimal SG from the member’s perspective, which is hardly
surprising as they receive higher savings at diminished personal cost. The optimal SG is boosted by around
2.5% if the employer bears 50% of the cost, although impact varies with income and member objectives.
We generate estimates of the welfare loss from imposing a sub-optimal SG on members. A typical lifetime
welfare (i.e. utility) loss from a sub-optimal SGs equals up to 2% of average income over the 5-years prior to
retirement under the estimates we consider most relevant. This implies that the downside from setting the SG at
a sub-optimal level is meaningful but not overly large. Under our baseline analysis, increasing the SG from 9.5%
to 12% results in utility losses that are equivalent to reducing income by up to 1% for the majority of members.
We also calculate the net impact per member on the government budget associated with various SG levels. This
analysis tallies the personal tax and Age Pension effects arising from our analysis over a member’s lifetime. The
modelling suggests that increasing the SG from 9.5% to 12% would result in a net increase in government revenue
taken from individual members. Nevertheless, the estimated revenue changes are relatively modest, suggesting
that the fiscal impacts might not be a major factor. However, we do not aggregate our estimates across members,
and hence do not comment on the overall budget implications.
Page 4
Considerations for Whether to Raise the SG
We put forward the following five considerations for policy makers to take into account in deciding whether the
SG should be increased to 12% as planned:
(i) Specify the policy objectives that the SG is trying to achieve – This is the critical consideration in our view.
Is the aim to facilitate maintenance of a standard of living post-retirement that is related to that experienced
pre-retirement (e.g. replacement rate); or to deliver at least some basic level of post-retirement income (e.g.
ASFA standards)? Should the superannuation system be directed at reducing reliance on the Age Pension?
Should superannuation be the vehicle through which members self-insure against risks such as living to a
very old age, involuntary early retirement or lower investment returns? And should the SG be set at a higher
rate to cover for the possibility that members might not make optimal choices, e.g. invest too conservatively?
(ii) Decide how to trade-off gains and losses between membersThe variable impact of any SG increase across
members should be addressed, noting that it could be relatively detrimental for low income earners.
(iii) Asymmetry between setting the SG higher versus lower – The asymmetry relates to the fact that a member
can do nothing if the SG is set too high, but can contribute more if it is set too low. Retaining a lower SG
for flexibility might be balanced against the reluctance to contribute beyond the mandated minimum.
(iv) Where the burden falls of a higher SGThis issue includes whether the member or the employer pays, and
the question of what broader economic consequences might arise if employers bear the cost.
(v) Impact on the government budget – Our estimates suggest the impacts are not large on a per-member basis;
although we do not address the demographics of the member base and hence the total budget impact.
Assumptions Needed to Justify an SG of 9.5% and Above
Our baseline analysis generates an optimal SG of below 9.5% for the member objectives that we consider most
relevant. The two main assumptions required to justify an SG of 9.5% and above include aiming to use the SG to
replace the Age Pension, and the stance that superannuation should facilitate self-insurance against investment,
longevity and early retirement risks. Two issues arise in taking the stance that a higher SG should be imposed as
a self-insurance mechanism. First is whether superannuation is the appropriate hedging vehicle, relative to seeking
other solutions based on social security or some form of member pooling. Second is that increasing the SG to
insure against these risks would lead to over-saving if the risks do not come to fruition. The potential
consequences are that pre-retirement consumption could be reduced unnecessarily, and an increased probability
of members dying with substantial unused balances and hence not receiving full benefit from the savings.
The impact of other assumptions is more moderate, but some combination might justify an SG of 9.5% and above.
Assumptions that increase our optimal SG estimates (in rough order of impact) include: assuming that the
employer bears a portion of the cost associated with the SG; increasing the replacement rate target; lowering
expected asset returns; and the member investing more conservatively, which also has the impact of lowering
returns. In the case of the employer bearing part of the cost, note that our analysis views the SG from the member’s
perspective. It hence does not account for any associated macroeconomic effects on profits, employment,
inflation, or overall consumption and savings. Assumptions that reduce our optimal SG estimates (in rough order)
include imposing a time preference parameter (discount rate); decreasing the replacement rate; and assuming
higher returns, perhaps through the member investing in an optimal asset mix entailing higher risky asset weights.
We find that imposing the minimum drawdown rules generates mixed effects, but mainly reduces the optimal SG
as constraining consumption increases the probability of that savings are not fully utilised before death. Changes
to loss aversion only has small impacts under the reference dependent utility function for our set-up.
Our Own View
We see the case for increasing the SG to 12% as tenuous unless the stance is adopted that a primary aim is to use
superannuation to replace the Age Pension where possible. We are wary over the use of the SG to facilitate self-
insurance against risks, noting that this could lead to over-saving with its own issues and costs. Further, our
analysis does not account for assets outside of superannuation that could significantly lower the required SG for
some members. Our preference would be to see policy directed at supporting pooling solutions (as it has done
recently with rule changes for annuities), while ensuring an appropriate social security safety net is in place.
Page 5
2. Analysis Set-Up
Our analysis is primarily designed to evaluate the impact of different SG levels on utility generated from lifetime
consumption at the individual member level. We vary the SG from 0% to 20% in increments of 0.5% under
differing assumptions. This allows identification of the ‘optimal’ SG to the nearest 0.5% for each set of
assumptions, and supports calculation of the change in member welfare from varying the SG from its current
level of 9.5%. The impact on the government budget is also estimated at an individual level.
The Model
We model an individual member that consumes all their disposable income pre-retirement, after accounting for
taxation and SG contributions. The member only saves via the SG, and does not directly optimise their pre-
retirement consumption and saving decisions. Under this set-up, the SG acts to reduce pre-retirement consumption
and hence utility prior to retirement, i.e. there is a pre-retirement cost borne by the member. The superannuation
balance that arises from the SG then generates a utility benefit in terms of post-retirement consumption. The
model effectively trades off the loss of utility from reduced consumption pre-retirement against the gain from
additional consumption post-retirement, above that which is generated by the Age Pension and related
supplements. Post-retirement consumption is uncertain for two reasons. First is exposure to investment risk over
the life-cycle, captured in the analysis through running 10,000 simulations of asset returns. Second, the member
does not know when they will die, and hence there is uncertainty over the extent to which they experience the full
benefit of their savings. The aim is to discover how differing SG levels impact on the utility of consumption over
the entire life-cycle, where the key trade-offs relate to consumption pre-retirement versus post-retirement, and
exposure to the risk and returns associated with investing via a superannuation fund as well as mortality.
We are particularly interested in how the optimal SG varies with member objectives and income. How we address
these two key variables is outlined in the next two sub-sections, followed by a discussion of the scope of the
analysis. After undertaking a baseline analysis, we then conduct sensitivity testing around other assumptions to
ascertain the extent to which they also matter. Specifically, we investigate how the results change with the utility
parameters, time preference, asset returns levels, investment strategy, required drawdowns, desire to hedge
against longevity risk, early retirement (also acts as a proxy for career breaks), the availability of the Age Pension,
and the assumption that the SG directly reduces the amount available for pre-retirement consumption (i.e. the
‘who pays’ issue). We also run some scenarios involving combinations of assumption changes.
The main elements of the model appear in Figure 1, with items selected for sensitivity testing listed at the right.
Most model elements are straightforward. One element that we do not directly address is home ownership, which
has potentially important implications for the SG1 and we plan to model in a follow-up study. Providing for the
cost of housing should have limited consequences under the replacement rate analysis, which might be interpreted
as modelling a situation where housing costs are a component of both pre-retirement and post-retirement
consumption. For instance, a member who rents would be incurring rental costs throughout. The main
complication occurs under the AFSA standards, which establish a fixed target level that is designed for a
homeowner, while our analysis does not address how the member comes to be in possession of a home. In
particular, ASFA modest sets a consumption target that may be far too low for non-homeowners who need to pay
rent or equivalent. The ASFA modest results should thus be interpreted as relevant only for lower income earners
who are fortunate enough not to pay rent. For members who need to pay rent, the ASFA comfortable results may
provide a better reference, noting that AFSA comfortable exceeds ASFA modest by $15,787, and this difference
appears to moderately exceed the additional costs associated with renting.2 Another interpretation is that the
ASFA standard analysis is relevant for a member who has access to a home or other lodgings at minimal cost,
perhaps because they have been bequeathed a house or due to their family situation.
1 A member who buys a home is effectively investing in an asset that generates a stream of income in terms of rental services.
As such, home purchase should be modelled as a savings decision that may occur in parallel with savings via superannuation.
2 According to the Australian Bureau of Statistics (ABS 4130.0, Housing Occupancy and Costs, 2017-18), average rent paid
to private landlords was $399 per week or $20,748 per annum, with lone person households paying $316 per week or $16,432
per annum. However, non-homeowners may qualify rental assistance of up to $3,588 annually, while ASFA modest includes
housing costs of $3,550 (excluding water charges) that are typically not incurred by renters. Thus, while AFSA modest of
$27,814 appears far too low as a target for non-homeowners who need to pay rent, an ASFA comfortable target should be
consistent with a living standard above that implied by ASFA modest after adjusting for rent and these offsetting items.
Page 6
Figure 1: Main Modelling Assumptions and Inputs
Element Baseline Sensitivity Testing
Dynamic programming, solved numerically
Pre-retirement phase from age 25-66, post-retirement phase
runs from age 67 until death
Income either consumed or invested in superannuation
Utility defined over consumption; no bequests
Modelling conducted in real terms; all real quantities
implicitly assumed to inflate at a common rate
Member Full-time worker, no career breaks
No assets outside the superannuation account
Mortality based 2015-17 Australian Life Tables for males,
applied from age 67 with mortality improvement
Aims to save enough for funds
to last until age 102
Retire at age 62; draw income
from the super fund; access to
Age Pension from age 67
Wage income Nine indicative gross income levels (pre-SG) ranging from
$30,000 to $150,000 in $15,000 increments
Income age profile hump-shaped, following Freestone (2018)
Income tax rates for 2019-20 including Medicare levy
Contributions SG range of 0% to 20%; 0.5% increments
Member contributes assumed SG; no additional contributions
Contribution taxes (allowing for caps) and LISTO applied
75%, 50% and 25% of SG
borne by the employer
Investment Constant asset mix of 70% risky asset, 30% in risk-free asset
Risky asset real expected return of 5% compound, standard
deviation of 17.4% (per global equities; Credit Suisse, 2019)
Risk-free real return of 1%; fixed
Fee of $70 plus 0.80% of balance (0.87% at $100,000)
Effective tax rates are imputed assuming mix of income,
capital gains and franking; zero tax rate in retirement phase
Risky asset return ±1.5%
Optimised asset mix, with no
short selling or borrowing
Life-cycle fund with 90/10
asset mix until 25 years prior
retirement, then trend to 40/60
mix at retirement and beyond
Social security Eligible for basic Age Pension, pension supplement and
energy supplement in June 2019 (maximum value $24,063)
Asset and income means-tests applied
Analysis is run excluding the
Age Pension and supplements
Drawdown in
Optimised, subject to minimum drawdown rules (i.e. member
may opt to draw down more if beneficial)
Superannuation accessed upon retirement at age 67
Super accessed at age 62
Minimum drawdown only
Draws consumption target
utility function
Functional form as per prospect theory value function
Reference level (i.e. target) pre-retirement consumption based
on after-tax income at zero SG, so that SG is treated as ‘loss’
Three reference levels for post-retirement consumption:
- Replacement rate of 70% of average after-tax income during
5-years prior to retirement, assuming an SG of 0%
- ASFA modest ($27,814)
- ASFA comfortable ($43,601)
Parameters of Blake et al. (2013), including loss aversion
weighting parameter of 4.5, curvature parameter on losses of
0.88, and curvature parameter on gains of 0.44
Discount rate of 0% (time preference = 1.00)
Higher loss aversion:
weighting parameter of 5.5,
curvature parameter on losses
of 0.88, gain 0.22
Lower loss aversion:
weighting parameter of 3.5,
curvature parameter on losses
of 0.88, gain 0.66
Discount rate of 2% (time
preference = 0.98)
Replacement rates of 60% and
Our expected asset returns are another element of the analysis that requires elaboration. Our baseline gross real
asset returns of 5.0% compound on the risky asset and 1.0% on the risk-free asset align with those observed
historically. While current returns on offer in the markets might be below historical levels, we are simulating over
an entire lifetime where contributions are progressively invested over a pre-retirement phase that lasts 42 years,
with the post-retirement phase extending up to another 43 years. Under these conditions, the historical compound
real return is arguably the best guide for the gross returns that might be expected over the long haul, noting that
Page 7
the returns currently on offer are less relevant for contributions made over future years. Adjusting for our taxation
and fee assumptions results in a net expected real compound returns of about 2.8% over the pre-retirement phase
under our baseline 70/30 asset mix at balances of $100,000 and above.3 This is more conservative than the median
real return target of 3.5% for MySuper balanced funds at June 2019.4 We conduct sensitivity testing on the return
assumptions in Section 6, and find they make a moderate but relevant difference to the optimal SG.
Member Objectives and Utility Function
To capture different objectives, the model evaluates consumption outcomes using a reference dependent utility
function that reflects the value function under prospect theory (Kahneman and Tversky, 1979; Tversky and
Kahneman, 1992). The reference dependent utility function accommodates the notion that there exists some target
level of consumption, and operates by applying a penalty to below-target consumption and a discount to above-
target consumption. This penalty is largely driven by loss aversion, i.e. the desire to avoid outcomes below the
target. See Warren (2019) for the functional form and detailed discussion of the properties of this utility function.5
The utility function is tailored to differing member objectives through applying three specifications for the post-
retirement consumption target: a replacement rate, and two fixed consumption targets based around ASFA modest
and ASFA comfortable respectively. These are coupled with a pre-retirement consumption target that equals post-
tax income assuming an SG of zero. This set-up treats any SG during the pre-retirement phase as a shortfall
relative to target, on the intuition that it reduces pre-retirement consumption relative to what would otherwise
have occurred. Meanwhile, the SG contributes toward attaining the post-retirement consumption target. The
reference dependent formulation hence involves a search for the SG level that balances the respective shortfalls
versus target in the pre-retirement and post-retirement phases, noting that consumption occurring in the pre-
retirement phase is deterministic while that in the post-retirement phase is uncertain due to random investment
returns and mortality. The modelling effectively analyses two broad categories of objectives with respect to target
consumption during the post-retirement phase, each with their own particular implications:
Replacement rate target – A replacement rate implies a desire to maintain a standard of living related to that
attained pre-retirement, albeit with relatively lower post-retirement consumption being acceptable. The
baseline post-retirement target is set at 70% of post-tax income during the 5-years prior to retirement (see
OECD, 2012), assuming an SG of zero. The set-up is equivalent to targeting the replacement of 70% of the
consumption that would have occurred in the absence of the SG. This formulation effectively links the pre-
retirement and post-retirement consumption targets, which scale up and down together.
Fixed consumption target – Assuming a fixed target might be interpreted in two ways. First, there is a given
level of post-retirement consumption that is deemed acceptable to the member. Second, the SG should be set
from a policy perspective with a view to ensuring that members save enough to achieve a minimal level of
post-retirement consumption. We examine both the ASFA modest and ASFA comfortable retirement standards
for a single household at June 2019, which equal $27,814 and $43,601 respectively. This formulation will tend
to indicate an optimal SG that suffices to achieve the post-retirement target, and no more. Further, because the
post-retirement target is decoupled from pre-retirement, it becomes likely that the indicated optimal SG will
decline with income as a lower savings rate becomes required to attain the target. This is exactly what we find.
3 Net returns at lower balances are reduced by the fixed $70 account fee component.
4 Based on data from the Australian Prudential Regulation Authority (APRA), see
5 We conducted analysis under power utility, which aligns with the objective of optimising lifetime utility of consumption.
We do not report the results to avoid complicating the message. Power utility suffers from various shortcomings in the
context of this study. It treats pre-retirement consumption and post-retirement consumption as equally valuable (at least
where a zero discount rate is applied). This seems counterfactual to the extent that required consumption declines post-
retirement, noting that pre-retirement consumption may be boosted by work-related expenses like travel, and the cost of
raising a family. Further, many members may reduce post-retirement consumption because they become less active
especially in their later years, with the availability of universal public healthcare helping to meet rising medical costs later
in life. In addition, the optimal SG estimates were boosted under power utility through additional savings aimed at limiting
the lower tail of outcomes during retirement. This resulted in optimal SGs that often sat at the maximum of 20%, and
generated higher expected post-retirement consumption than observed pre-retirement. Saving to avoid poor outcomes thus
dominated over consumption smoothing under the set-up, aided by various constraints not usually present in life-cycle
models. (The reference dependent utility function did not suffer from this issue due to a lower curvature.)
Page 8
Our baseline utility parameters are aligned with relatively high loss aversion, although sensitivity testing as
reported in Section 6 reveals that the results are not very sensitive to parameter changes. We adopt the parameters
of Blake et al. (2013), which include a curvature parameter on gains of 0.44, a curvature parameter on losses of
0.88, and a weighting (loss aversion) parameter on losses of 4.5. These place a heavy discount on above-target
consumption, and a large penalty on below-target consumption. This generates a strong preference for avoiding
consumption below target while giving limited credit for exceeding the target. The baseline analysis applies a
zero discount rate, i.e. time preference parameter of 1.00. This assumption implies no intertemporal preference
for earlier consumption over later consumption, i.e. both are considered equally valuable. Whether a discount rate
should be applied in the context is an open question. It is plausible that a younger member might reasonably place
lower value on post-retirement consumption. The issue also relates to the purpose of the SG. There is ample
evidence that people can be myopic and may heavily discount the future.6 This may prevent some from saving an
adequate amount for retirement, even though it may be in their own interests. We investigate the potential impact
of time discounting by running sensitivity testing including a discount rate of 2% pa, i.e. applying a time
preference parameter of 0.98. This implies that consumption occurring (say) 30 years in the future generates only
55% of the utility to an equivalent consumption outcome today.
Income Levels
We run the analysis over nine income levels, which we denote income level one (L1) through to income level
nine (L9). Our income range extends from $30,000 to $150,000 in $15,000 increments, which is taken to represent
average pre-retirement income from age 25 to age 66 prior to personal income tax and the SG. We settled on this
approach after finding none of the available income distribution statistics to be fully satisfactory. The main aim
is to establish how the optimal SG varies with income, meaning that the exact income levels being analysed are
not critical provided that they span a meaningful range. Our income range encompasses the effect of gender7 and
under-employment to the extent that women and part-time workers earn lower incomes, such that the results at
the lower income levels may be interpreted as more relevant for these two groups. We discuss how our income
range relates to available income distribution statistics at the end of this sub-section.
To generate an income profile over the pre-retirement phase, we take the nine average pre-retirement incomes as
a reference point. The model of Freestone (2018) is then used to generate hump-shaped income profiles to match
each average level. Under this model, income at age 25 is about 78% of the average. Income grows to peak at age
47 at 111% of the average, implying 1.6% real growth over this period. Income then tapers off to reach an average
over the 5-years prior to retirement that is 10% below the average, and 19% below the peak. The real growth rate
from age 25 to age 62-66 is 0.4%. Details of the income distribution appear in Figure 2.
Figure 2: Wage Income Distribution, Most Relevant Objectives and Related Income Targets
Income Level L1 L2 L3 L4 L5 L6 L7 L8 L9
Base Income, Age 25 $23,278
$34,917 $46,557
$81,474 $93,113 $104,752
Maximum, Age 47 $33,303
Average, Age 25–66 $30,000
Average 5-years Prior
Retirement, Age 62–66 $27,028
$94,596 $108,110
Most relevant
Related income targets $27,814 $27,814 $43,601
* AM is ASFA modest, AC is ASFA comfortable, RR is replacement rate
6 Notable behavioural effects that give rise to a bias toward the present include myopic loss aversion (see Benartzi and Thaler,
1995) and hyperbolic discounting (see Laibson, 1997).
7 Females differ to males in a propensity to earn lower incomes, a higher likelihood of career breaks and longer life
expectancy. The main implication of the first two differences is lower lifetime income, and is thus covered by our income
distribution. Our sensitivity tests involving early retirement also cover for career breaks to some extent.
Page 9
Figure 2 also indicates the member objectives that we consider ‘most relevant’, and the related income targets.
For income L1 and L2, we consider ASFA modest as most relevant. At these lower incomes, a replacement rate
objective implies a living standard that is possibly below the poverty line, while ASFA comfortable implies a
living standard above that enjoyed pre-retirement. For income L3 to L9, we view both ASFA comfortable and
the replacement rate as competing candidates for plausible objectives, depending on whether the aim is to achieve
a set standard of living, or to achieve income and consumption reflecting that enjoyed pre-retirement.
It is worth expanding on how our baseline 70% replacement rate relates to the pre-retirement income distribution.
A 70% target based on average income over the 5-years prior to retirement equates to a replacement rate of 63%
based on average pre-retirement income over the working phase from age 25 to age 66. However, we base our
replacement rate target on after-tax income at a zero SG, which supports isolating the impact of the SG itself.
This means that our quoted replacement rate is a higher percentage of disposable income in the presence of any
SG. For instance, under an SG of 9.5%, our replacement rate target for income L5 equates to 76% of disposable
income in the 5-years prior to retirement, and 69% of average pre-retirement income over age 55 to age 66.
To place our income distribution in context, we refer to Australian Bureau of Statistics (ABS) data on employee
earnings8 and household income9 from 2017-2018. The lowest income L1 of $30,000 sits at around the 20th
percentile for total earnings per employee of $29,120, but is well below the 10th percentile for earnings per full-
time employee of $46,608. It is positioned in-between the 10th and 20th percentile for household income of
$24,544 and $38,896 respectively. Our highest income L9 of $150,000 is above the 90th percentile for both total
and full-time earnings per employee of $123,084 and $142,428 respectively, but sits about half-way between the
70th and 80th percentile for household income. The central income L5 of $90,000 is close to median household
income, although it is nearer to the 60th percentile for full-time earnings and 75% percentile for total earnings per
employee. It also happens to be close to ordinary-time adult average weekly earnings at May 2019 of $89,840.10
We believe that the range we span is reasonable and meaningful, noting that those on incomes below $30,000
may be receiving support from the government or elsewhere that is not explicitly taken into account in our
analysis. In addition, incomes above $150,000 comprise a minority of very wealthy people who should be able
to cater for themselves, and might not be considered the prime concern for determining SG policy.
Scope of the Analysis
Implementation of the type of model applied in this study requires making judgements around which elements to
include or leave out, and the degree of complexity to admit. Our aim is to incorporate all elements that might be
crucial for setting the SG. Figure 3 summarises the key elements that are included in the model on the left, and
notes some of the more notable elements that are excluded on the right. We consider the key elements captured
by our modelling to be as follows:
We analyse a single wage-earner that faces the trade-off between consuming their available income, or saving
for retirement via their superannuation fund;
We assume the member contributes to their superannuation fund only to the extent that the SG dictates;
The superannuation (and retirement) fund invests in either a risky asset (equities) or a risk-free asset (cash);
The member has access to the Age Pension plus related supplements, and is thus relying on both the Age
Pension and drawdowns from their retirement fund to support consumption in retirement;
We allow for the tax and other rules surrounding both personal income and superannuation, as well as the
pension eligibility rules;
We assume that the member holds a 70/30 balanced fund without attempting to optimise their asset mix;
We assume that the member optimises their drawdown decisions;
The member faces uncertainty over the post-retirement consumption that their savings will generate due to
randomness in both investment returns and mortality.
8 Based on ABS 6306.0- Employee Earnings and Hours, Australia.
9 Based on ABS 6523.0 - Household Income and Wealth, Australia.
10 Or $1,727.70 per week, per ABS 6302.0 - Average Weekly Earnings, Australia.
Page 10
Figure 3: What We Include and Excluded from the Analysis
Included in the Model Notable Exclusions From the Analysis
Member either consumes their income, or saves a portion of
it by investing via their superannuation fund
Wage income follows a pre-determined hump-shaped path
during the pre-retirement phase
Taxation and other rules related to personal income and
Impact of uncertain mortality
Availability of the Age Pension plus supplements
Member contributes the SG only to their superannuation fund
Two assets: risky with stochastic real returns, risk-free with
fixed real returns
Member invests in a 70/30 balanced fund (which our analysis
reveals as sub-optimal)
Member optimises drawdowns, subject to meeting the
minimum drawdown rules
No assets outside of superannuation, including
notably the potential to invest in a home
Income uncertainty (e.g. early retirement and
career breaks) is not addressed under the
baseline analysis, but is investigated under the
sensitivity tests
Modelling performed in real terms without
allowing for the possibility that average
incomes, the Age Pension, consumption targets
and asset returns may inflate at differing rates
Household effects are not considered, with
modelling based around a single individual
without any bequest motive
No other assets are available for investment,
including annuities
The assumption that the member holds a 70/30 fund while optimising their drawdowns requires comment. It is
an open issue whether the optimal SG should be estimated by assuming that the member behaves optimally, or in
line with some commonly observed practice. Assuming that the member behaves optimally avoids confounding
the SG findings with potentially sub-optimal choices, perhaps as a consequence of poor financial literacy,
behavioural biases or shortcomings in the policy environment itself. On the other hand, members do not
necessarily behave optimally with regard to either investment or drawdown decisions. Applying a 70/30 asset
mix recognises that the optimal investment strategy is unlikely to be followed as it entails much higher risky asset
weights than most members might be willing to accept. Our model generates risky asset weights that average
between 75% and 85% across the life-cycle, depending on the member objective. By contrast, the optimal
drawdown strategy that emerges is far more plausible, and entails drawing down the target in most circumstances
(subject to the minimum drawdown rules). Against this background, our sensitivity testing gauges the impact of:
optimised asset weights; a life-cycle investment strategy under which the risky asset weight transitions from 90%
to 40% at retirement; and the member strictly following the minimum drawdown rules.
Five notable exclusions from the analysis are listed on the right of Figure 3. These are discussed below:
(a) Absence of other assets outside of superannuation – This is potentially an important omission: effectively
we are modelling only pillar one and pillar two of the retirement incomes system. Including other assets in our
model should lower the estimated optimal SG, potentially significantly. Other forms of savings can act as a direct
substitute for saving via superannuation, and hence may impact on the SG required to support a desired level of
post-retirement consumption. Housing is the asset that matters most in this context. Individuals who buy a family
home enter into a form of semi-committed savings, and effectively buy an income stream in terms of the rental
services that may extend through into retirement. Further, housing is advantaged to the extent that it is exempt
from income and capital gain tax, and does not qualify for the assets test that determines Age Pension eligibility.
For these reasons, housing might be viewed as a genuine alternative to superannuation; and the optimal SG for
those with a family home might be expected to be considerably lower than for those without. An analysis of
housing and its policy implications is a topic deserving of its own analysis, but is treated as beyond scope here.
(b) Uncertainty over pre-retirement incomeOur baseline model assumes that pre-retirement income is known,
although this issue is partly addressed during sensitivity testing. The most important source of uncertainty relates
to the length of time spent in employment. Career breaks, periods of part-time work and early retirement all have
the dual effect of lowering contributions into superannuation and creating the need to fund consumption when
income dries up. Our sensitivity test is framed around early retirement, and indicates that a meaningfully higher
SG is required when the length of time in employment is reduced by 5 years. Nevertheless, this issue is more
complex than captured by the analysis in this study. In particular, it is not clear that a higher SG is the best way
Page 11
to deal with income uncertainty. First, if income is uncertain, or there is a possibility of career breaks, then it
makes sense that some level of precautionary balances should be built to be readily accessible in times of need.
These precautionary balances might be better held outside of superannuation,11 thus providing one reason why
the SG might be set at a lower level to provide more room for building precautionary savings elsewhere. The
availability of social security (e.g. unemployment benefits) and capacity to borrow also needs to be taken into
account. Further, the effect of career breaks on savings via superannuation might be mitigated by members
contributing more when they can.12 The other consideration is that neither career breaks nor early retirement are
certain. It is debatable whether it is appropriate to set a higher SG for all members ‘just in case’ some may suffer
a career break or retire early. The impact of income uncertainty on the optimal SG is a complicated issue that
requires a more complex model to be addressed properly.
(c) Analysis in real terms, assuming that all amounts inflate at a common rate The possibility that some
quantities may inflate at different rates has been referred to as relating to the ‘deflator’ used, although we think
this is better seen as a matter of whether to allow for differential real growth rates. One key issue is the relation
between economy-wide real income growth, the post-retirement consumption target and the Age Pension. Also
of relevance is the basis of the assumed asset returns, which we set with reference to historical CPI-adjusted
returns. Our analysis might be viewed as consistent with a world where economy-wide incomes and hence the
Age Pension do not grow in real terms, so that all variables inflate with prices (e.g. CPI). Note that we do allow
for real income growth at an individual level under the model of Freestone (2018), equating to 1.6% between age
25 and age 47 and 0.4% between age 25 and the 5-years prior to retirement.
The effect of economy-wide real income growth on the optimal SG estimates is unclear, and will depend on
assumptions and various interactions. First, economy-wide real income growth would raise the dollar value of
contributions above what we have assumed, thus increasing the balance at retirement available to support post-
retirement consumption. Second, the impact on the post-retirement consumption target depends on how the target
is formulated. Under the replacement rate objective, the target will rise in lock-step with any additional real
growth in income. Calculations indicate that the percentage increase in a replacement rate target would be
approximately twice that of the balance at retirement. However, the extent to which fixed consumption targets
such as the ASFA standards increase with economy-wide real income growth is unclear: these targets might rise
at a rate somewhere between wage and price inflation. Third, economy-wide real income growth would raise the
level of Age Pension, given that it is linked to average weekly earnings. This opens up the possibility that the Age
Pension might grow at a higher rate than the consumption required to retain a given standard of living during the
post-retirement phase relative to what is implicit under our analysis, to the extent that living standards are linked
to prices rather than wages. The net effect from these various influences is not straightforward.
(d) Household effectsWe model a single-income household with male mortality with mortality improvements
(i.e. allowing for increasing life expectancy). Modelling a household rather than an individual would add much
complexity. It would require making allowance for the possibility of dual income streams and dual superannuation
funds, applying Age Pension payments for couples, addressing dual longevity risk (the funds need to last to the
second survivor), and making adjustments for shared living costs. These various elements are at least partly
offsetting. We suspect modelling a household would make a marginal rather than substantial difference to the
results. The fact that we do not allow for bequests is worth noting, as it implies zero value is attached to any
residual left in the retirement account at death. The idea that the residual value is worthless will be counterintuitive
for many members. Including a bequest motive in the analysis should tilt the results towards a higher optimal
SG. On the other hand, it is debatable whether public policy should be set to accommodate bequests.
(e) Limited asset universeWe assume that the available assets comprise only a risky asset (‘equities’) and a
risk-free asset (‘cash’). Adding additional assets would allow more effective portfolios to be built, which might
lower the SG required. This would be particularly the case where other assets might make a post-retirement
consumption target easier to attain at lower risk. In particular, the availability of annuities might make a
difference, to the extent that they deliver certainty of income and help to hedge longevity risk. Nevertheless, we
expect that adding additional asset classes to the analysis would make only a modest difference.
11 The hardship provisions, under which individuals can access their superannuation under extreme circumstances, go some
way to mitigating this issue. However, they are a very poor substitute for having readily accessible savings to draw on.
12 This option of contributing more at a later date is generally not available under early retirement.
Page 12
3. Baseline Results – Optimal Superannuation Guarantee
We start by presenting our baseline estimates for the ‘optimal’ SG. The baseline sets the foundation for further
analysis aimed at gauging how the SG varies with assumptions, which is the primary focus of this study. Figure
4 reports the matrix of optimal SG estimates by income and member objectives, based on running the model for
SG levels from 0% to 20% in increments of 0.5%. The grey shading highlights results under the objectives we
consider most relevant, which include ASFA modest at income L1 and L2, either the replacement rate or ASFA
comfortable at L3 through to L9.
Figure 4: Optimal Superannuation Guarantee under the Baseline Analysis
Income Level
Member Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
70% Replacement Rate 0.0% 0.5% 3.5% 5.0% 6.0% 6.5% 7.5% 8.0% 8.5%
ASFA Comfortable 15.0% 12.0% 9.0% 7.0% 6.0% 5.0% 4.5% 4.0% 3.5%
ASFA Modest 3.0% 2.5% 2.0% 1.5% 1.0% 1.0% 1.0% 1.0% 1.0%
Two stark findings emerge. The first is the large variation in the optimal SG. There is no single SG that suits all.
The second is that the optimal SG sits below the current level of 9.5% under the most relevant objectives as
highlighted with grey shading, where the range is 2.5% to 9.0%. Indeed, an SG of 9.5% or above is indicated
only under the ASFA comfortable target for L1 and L2, where the optimal SG is 15% and 12% respectively.
Recall that ASFA comfortable sits above the level of pre-retirement consumption at incomes L1 and L2.
Estimation of optimal SGs of below 9.5% under the most relevant objectives will come as a surprise to many
readers. Hence we now provide a detailed explanation of why our model generates these outputs, placing
emphasis on the estimates for the most relevant objectives. Under the reference dependent utility function, the
model directly addresses the question of how much a member needs to save in a bid to reach a post-retirement
consumption target. Noting that sacrificing consumption pre-retirement is treated as a loss under our set-up, the
model is in effect trying to locate the SG that balances the utility loss incurred pre-retirement, against the utility
loss if the SG remains lower than required to achieve the post-retirement consumption target. Further, in most
(but not all13) cases, the post-retirement consumption target is lower than the pre-retirement consumption target.
Lower saving is also required where the Age Pension is doing much of the heavy lifting in terms of attaining the
consumption target, as is the case for low income earners in particular.
Life expectancy is another factor at play. Uncertain mortality has the effect under our model of placing a lower
weight on consumption at older ages. For example, the probability of surviving to age 90 is 47%, meaning that
consumption outcomes at age 90 might be considered as having a 47% weighting. The model is thus trading off
certain reductions in pre-retirement consumption against increases in post-retirement consumption that are
uncertain due to both fluctuations in asset returns and the chance that the consumption generated may not be
experienced if the member does not survive. The model consequently references life expectancy (of age 89) in
identifying the SG that optimises lifetime utility, and implicitly attaches a low weight on the possibility that the
member might survive to a very old age.
Figure 5 provides intuition by plotting the consumption profile that emerges under a replacement rate target for
income L5, applying the ‘optimal’ SG of 6.0%. Note that an SG of 6.0% is also optimal under the ASFA
comfortable target at L5, hence this chart covers both objectives at this income level. The double grey line plots
‘available’ pre-retirement income and the post-retirement target, with the latter downwardly-dislocated. The
model indicates that in most cases it is optimal to drawdown the target until the retirement account balance is
exhausted, resulting in the member then moving onto the Age Pension. Drawdowns are subject to the minimum
drawdown rules, which may require the member to withdraw more than the target where higher investment returns
occur. (This explains the bumps in the 75th percentile line.) The utility-based model thus trades off the reduction
in pre-retirement consumption against the risk of ending up on the pension later in life. The consumption profile
appearing in Figure 5 indicates that exhaustion of the retirement account occurs at a median age of 91, with the
13 ASFA modest exceeds pre-retirement post-tax income at L1, as does ASFA comfortable at L1 and L2.
Page 13
25th percentile at age 87 and the 75th percentile at age 102. In essence, the model suggests that the utility loss from
giving up 6% of pre-retirement consumption is equal to the utility benefit of consuming the post-retirement target
well into the 80’s with a 75% likelihood, with more than a 25% chance of sustaining the consumption target
beyond age 100. The model implies it is worth taking these odds, with the downside related to ending up on the
Age Pension later in life.
Figure 5: Consumption for Income Level 5 under Replacement Rate Target at the Optimal SG
The manner in which the optimal SG estimates vary with income also require explanation. Three factors are at
Age Pension – As the Age Pension provides a capped level of income support during retirement, it is of greater
value to lower income earners relative to higher income earners. It is hence more influential in limiting the
need for a higher SG for those on lower incomes.
Tax effects – The differential between the tax rate applied to personal income and superannuation vary with
income in manner that makes contributing to superannuation less beneficial for those on lower relative to
higher incomes. The Australian marginal personal income tax rates in 2019-20 are 19% for income between
$18,201 and $37,000, then 32.5% for additional income up to $90,000, 37% for additional income up to
$180,000, and 45% on any income above $180,000. Meanwhile, superannuation is taxed at 15% on
contributions, 15% on income, and 10% on capital gains. Members on incomes below $37,000 – encapsulating
L1 and L2 in part – notionally incur higher marginal tax rates on superannuation than regular income, although
this is offset by the low income superannuation tax offset (LISTO) of up to $500 (which we incorporate in the
modelling). Meanwhile, superannuation can act to lower effective tax rates for those in the higher tax brackets.
Nature of the post-retirement consumption target – Where the post-retirement consumption target is fixed in
level terms (i.e. the ASFA standards), a declining saving rate is required to attain that target as income
increases. On the other hand, under a replacement rate the post-retirement consumption target is linked to pre-
retirement income and hence consumption scales up with income. In this case, there is limited interaction
between income and the optimal SG arising from the target itself.
The combination of the above effects leads to decline in the optimal SG with income under both ASFA
comfortable and ASFA modest, where the dominant effect is that a lower portion of income needs to be saved to
attain the fixed target as income increases. Under the replacement rate objective, the optimal SGs rise with income
25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105
Real Consumption
75th Percentile
25th Percentile
Estimated for member with income
L5 at their 'optimal' SG of 6.0%
Page 14
levels. Here the dominant influence is the tailing off in the value of the Age Pension relative to income, with tax
effects playing a supporting role.
Putting it all together, under the ASFA comfortable target of $43,601, the optimal SG starts at 15.0% at L1 then
progressively declines. Noting that AFSA comfortable sits above average pre-retirement income for income L1
and L2, the range of the relevant results under ASFA comfortable extend from 9.0% at L3 down to 3.5% at L9.
Under ASFA modest, the maximum Age Pension plus supplements of $24,063 is just $3,751 short of the
consumption target of $27,814. Hence a very low SG is required to secure the target. The optimal SG under ASFA
modest is only 3.0% for L1 and 2.5% for L2 (the level we consider most relevant); then declines to only 1.0% at
L5 and above. Under the replacement rate objective, the optimal SG increases over the most relevant range from
3.5% at L3 up to 8.5% at L9. It stands at 0% for income L1 and 0.5% for L2, as the Age Pension is above or near
the target. These estimates highlight how purported member objective, income level and the existence of the Age
Pension are key determinants of the optimal SG for any particular member.
In summary, the relatively low SGs indicated by our model arise from a combination of requiring enough savings
to achieve a post-retirement consumption target that is less than pre-retirement consumption, accounting for
access to the Age Pension, and balancing the utility cost of a certain reduction in utility pre-retirement against
an uncertain gain in utility post-retirement. We suspect that other commentators generate higher SGs in part
because they examine the post-retirement phase in isolation, and do not characterise the SG in terms of a trade-
off against pre-retirement living standards that involves uncertainty over the post-retirement pay-off due to
unknown investment returns and mortality. Further, some commentators may be explicitly or implicitly giving
insufficient credit to the Age Pension, which is a key factor in generating our results particularly for low income
earners. We examine the impact of excluding the Age Pension, as well as altering other assumptions, under the
sensitivity testing reported in Section 6.
4. How Member Welfare Changes with the Superannuation Guarantee
We now use our baseline estimates to gauge the individual welfare effects associated with differing SG rates. To
make the analysis more tangible, we convert expected utility into certainty equivalents expressed in terms of
lifetime consumption. This provides a utility-based metric where the units directly relate to yearly consumption
levels, and hence can be interpreted in economic terms. Given the use of reference dependent utility functions,
we calculate the constant deviation of consumption from the target across all ages that generates the same utility
as estimated under the simulations. We denote this ‘certainty equivalent distance from target’. Figure 6 plots the
respective certainty equivalents under the utility functions for income L3, L6 and L9, except for ASFA modest
where we plot them for L1, L2 and L3. The vertical double-line representing the current SG of 9.5%. Figure 6
provides a visualisation of how expected utility varies with the SG across a range of incomes and member
objectives. The utility functions all reveal a fairly distinct peak, sometimes with relatively steep slopes around
the optimal level.
Page 15
Figure 6: Certainty Equivalents Metrics at Three Income Levels
Notes on Certainty Equivalents Metric:
The series plotted represent levels of lifetime real
consumption, expressed as a difference from target
The highest points on each curve represent the
optimal SGs as reported in Figure 4.
Focus should be placed on how the certainty
equivalents vary with the SG: the overall level of
each series is of limited consequence.
The slope of the curves around the optimal SG
provide an indication of degree of the potential
welfare losses from imposing a sub-optimal SG on
a particular member.
Figure 7 presents the difference in certainty equivalent estimates for an SG of 9.5% relative to the estimated
optimal SG, and for an SG of 12% versus and an SG of 9.5%. Panel A reports (real) dollar differences; while
Panel B expresses these differences as a percentage of income in the 5-years prior to retirement. Across the
member objectives that we consider most relevant, the 9.5% SG is sub-optimal an imposes a cost up to the
equivalent of about 2% of income. The losses sit above 1% of pre-retirement income for income L3 and L4 under
the replacement rate, L8 and L9 under ASFA comfortable, and level L1 and L2 under ASFA modest. In all cases,
the loss occurs because a 9.5% SG is too high. The estimates also indicate that increasing the SG from 9.5% to
12% would impose a loss in utility equivalent to a reduction of 0.2% to 1% in income under the most relevant
objectives. Our welfare estimates depend on our model, and are valid only for the baseline assumptions. They
flag the possibility that increasing the SG to 12% might harm more members than it helps. Nevertheless, the
reduction in welfare is estimated to be relatively modest.
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%
Distance from Target
Superannuation Guarantee
70% Replacement Rate Target
Income Level 3
Income Level 6
Income Level 9
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%
Distance from Target
Superannuation Guarantee
ASFA Comfortable Target
Income Level 3
Income Level 6
Income Level 9
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%
Distance from Target
Superannuation Guarantee
ASFA Modest Target
Income Level 1
Income Level 2
Income Level 3
Page 16
Figure 7: Certainty Equivalent Gains (Losses) at 9.5% and 12% Superannuation Guarantees
Difference in Certainty Equivalent
Consumption or Distance from Target
Income Level
L1 L2 L3 L4 L5 L6 L7 L8 L9
PANEL A: Dollar Difference (Real)
9.5% SG less Optimal SG
Replacement Rate -1,784 -1,401 -946 -645 -414 -239 -133 -62 -20
ASFA Comfortable -258 -72 -3 -131 -375 -664 -1,018 -1,418 -1,844
ASFA Modest -475 -779 -1,234 -1,741 -2,166 -2,580 -3,025 -3,492 -3,958
12% SG less 9.5% SG
Replacement Rate -296 -378 -475 -529 -526 -458 -400 -343 -290
ASFA Comfortable 200 72 -117 -322 -509 -643 -782 -922 -1,058
ASFA Modest -259 -356 -473 -599 -719 -804 -906 -1,017 -1,132
PANEL B: % Income 5-Years Prior Retirement
Average Income, 5-Years Prior Retirement 27,028 40,541 54,055 67,569 81,083 94,596 108,110 121,624 135,138
9.5% SG less Optimal SG
Replacement Rate -6.6% -3.5% -1.8% -1.0% -0.5% -0.3% -0.1% -0.1% 0.0%
ASFA Comfortable -1.0% -0.2% 0.0% -0.2% -0.5% -0.7% -0.9% -1.2% -1.4%
ASFA Modest -1.8% -1.9% -2.3% -2.6% -2.7% -2.7% -2.8% -2.9% -2.9%
12% SG less 9.5% SG
Replacement Rate -1.1% -0.9% -0.9% -0.8% -0.6% -0.5% -0.4% -0.3% -0.2%
ASFA Comfortable 0.7% 0.2% -0.2% -0.5% -0.6% -0.7% -0.7% -0.8% -0.8%
ASFA Modest -1.0% -0.9% -0.9% -0.9% -0.9% -0.8% -0.8% -0.8% -0.8%
5. Net Impact on the Government Budget
We calculate the total impact of different SG levels on the government budget at the individual member level by
summing the effect on taxation and the Age Pension and supplements over the entire lifetime. The post-retirement
inputs are weighted by probability of survival, thus generating an expected total impact allowing for uncertain
life expectancy. Our calculations include the following inputs:
Inflows to the government:
Income tax pre-retirement, including the Medicare levy
Superannuation contributions tax
Taxes on superannuation earnings pre-retirement
Outflows from the government:
Low-income superannuation tax offset, up to $500 pa for individuals with income of $37,000 or less
Age Pension and supplements
Franking credits claimed within the retirement account
This is of course a partial analysis of the interactions between an individual and the government, ignoring other
effects such as the impact on company tax and the GST. Hence the absolute numbers are less meaningful than
how they vary with the SG. We also take no account of time discounting, and do not consider the mix of member
cohorts in the economy that could lead to aggregate impacts that differ significantly from our indicative estimates.
In particular, tax revenue is accrued pre-retirement while the impact on outlays via the Age Pension and other
items is largely incurred post-retirement, meaning the net present value will exceed our estimates based on a
simple sum over time from the government’s perspective. These timing differences will also interact with
population demographics in determining the aggregate budget impact.
Page 17
Figure 8: Net Revenue to the Government – Tax Collected less Pension and Related Supplements
$'000 per Individual over their
Lifetime (Constant 2019 Dollars)
Income Level
L1 L2 L3 L4 L5 L6 L7 L8 L9
PANEL A: Net Revenue Estimates
Superannuation Guarantee of 0% -416 -238 -21 196 420 660 905 1,150 1,396
Superannuation Guarantee of 9.5%
Reference Dependent, Replacement Rate -412 -220 31 337 580 825 1,074 1,324 1,571
Reference Dependent, ASFA Comfortable
-401 -199 58 320 579 845 1,113 1,378 1,638
Reference Dependent, ASFA Modest -410 -219 22 267 511 768 1,029 1,290 1,548
Superannuation Guarantee of 12%
Reference Dependent, Replacement Rate -401 -202 60 381 627 866 1,111 1,355 1,598
Reference Dependent, ASFA Comfortable
-381 -168 100 367 625 885 1,145 1,402 1,654
Reference Dependent, ASFA Modest -399 -200 48 298 545 798 1,056 1,314 1,568
PANEL B: Changes in Net Revenue
Superannuation Guarantee 9.5% vs. 0%
Reference Dependent, Replacement Rate 4 18 52 141 160 165 169 173 175
Reference Dependent, ASFA Comfortable
15 39 80 124 158 185 208 227 241
Reference Dependent, ASFA Modest 6 20 43 71 91 108 124 140 152
Superannuation Guarantee 12% vs. 9.5%
Reference Dependent, Replacement Rate 11 18 29 44 46 42 37 32 27
Reference Dependent, ASFA Comfortable
20 31 42 47 46 40 33 25 17
Reference Dependent, ASFA Modest 12 19 26 31 33 30 27 24 19
By capturing the main revenue and outlay items that are affected by the SG we provide an indication of how
altering the SG may impact on the flows between the government and individuals. Figure 8 reports the net revenue
associated with an SG of 0%, 9.5% and 12% in Panel A, then presents the change in moving from an SG of 0%
to 9.5% and from 9.5% to 12% in Panel B. Focusing on the change estimates appearing in Panel B, we find that
the SG increases net government revenue across an individual’s lifetime at all income levels, albeit by reducing
what are net outlays at L1 and L2. Our results indicate that increasing the SG from 9.5% to 12% would boost net
revenue in the range of $12,000 to $47,000 under the most relevant objectives. The numbers do not seem large
bearing in mind that they are lifetime totals. The impact is less for those on high incomes. For example, the
increase in net revenue taken by the government under the most relevant objectives stands between 12% and 34%
of one year’s income at L7 to L9, and between 42% and 70% for L1 through to L6.
6. Sensitivity and Scenario Testing
We test the sensitivity of the optimal SG estimates to a wide range of assumptions. This reveals which other
assumptions are important, in addition to income level and member objectives as examined under the baseline
analysis. The sensitivity analysis reveals that changing the following assumptions produces a meaningful change
in the estimated optimal SG:
Higher SG indicated (in rough order of significance)
Ignoring access to the Age Pension, implying a desire (or policy objective) for self-funding
Early retirement at age 62, thus self-funding consumption until accessing the Age Pension at age 67
Member saves so that the funds might last until age 102, i.e. self-insuring against longevity risk
Employer bears a substantial portion of the cost of the SG, rather than the member
Lower asset returns, via decreasing expected returns on the risky asset by -1.5%
Lower exposure to risky assets, through a life-cycle strategy reaching 40% in risky assets at retirement
Greater required drawdowns, e.g. higher replacement rate
Page 18
Lower SG indicated (in rough order of significance)
Adding a discount factor, consistent with a preference for current over future consumption
Optimised asset mix, reflecting higher average risky asset weights
Higher asset returns, via increasing expected returns on the risky asset by +1.5%
Imposing the minimum drawdown rules (decreases the optimal SG in most cases, but not all)
We also run two groups of scenarios to evaluate combinations of changes, focusing on sets of assumptions that
might provide justification for a higher SG. The first group captures the effect of using the SG as a hedging
mechanism against investment risk in conjunction with either longevity or early retirement risk. The second group
includes ‘kitchen sink’ scenarios that invoke combinations of: excluding the Age Pension; the employer bearing
50% of the cost of the SG; saving to maintain consumption at target through to age 102; retiring early at age 62;
lower investment returns; and imposing the minimum drawdown rules.
Figure 9 presents an overview of the sensitivity and scenario analysis. It sets out changes made to the assumptions,
and provides an indication of the directional impact on the optimal SG in the final column. The sensitivity and
scenario analysis is arranged under five categories, which provides the structure below under which the results
are reported and discussed.
Page 19
Figure 9: Overview of the Sensitivity and Scenario Analysis
Sensitivity Test /
Input Assumptions
Baseline Sensitivity Testing Indicative Impact on
Optimal SG
Utility Parameters
Higher loss aversion Slope 4.5, Curv(-) 0.88,
Curv(+) 0.44
Slope 5.5, Curv(-) 0.88,
Curv(+) 0.33
Very small, if any
Lower loss aversion Slope 4.5, Curv(-) 0.88,
Curv(+) 0.44
Slope 3.5, Curv(-) 0.88,
Curv(+) 0.66
Very small, if any
Higher discount factor 0% p.a. 2% p.a. Lower
Required Income Stream
Higher replacement rate 70% of income age 62-66 80% of income age 62-66 Higher
Lower replacement rate 70% of income age 62-66 60% of income age 62-66 Lower
Minimum drawdown Optimised, subject to
minimum drawdown
Draw no more than the
minimum drawdown
Mixed, but mainly lower
Draw the target Optimised, subject to
minimum drawdown
Draw the target (or min.
drawdown) until retirement
account exhausted
Modestly higher
Longevity risk self-insured Expected mortality Assume death at age 102 Much higher
Retire early Retire at age 67 Retire age 62, draw income
until access pension age 67
Very much higher
Investment Assumptions
Higher risky asset return 5.0% compound 6.5% compound Lower
Lower risky asset return 5.0% compound 3.5% compound Higher, except for ASFA modest
Optimised asset mix 70% risky / 30% risk-free Dynamically optimised Lower
Life-cycle strategy 70% risky / 30% risk-free 90% risky to age 42, trends
to 40% risky at retirement
Policy Environment
Age Pension Age Pension available Fully self-funded Very much higher
Who bears the cost Member Employer bears 25/50/75% Higher
Self-insuring against
combinations of two risks
See above 3.5% risky asset return
plus death at age 102
3.5% risky asset return
plus retire at age 62
Meaningfully higher: SG of
9.5% or 12% is justified across a
much broader range of income
levels and member objectives
Combinations of changes:
Fully self-funded
Employer pays 50%
3.5% risky asset return
Death at age 102
Minimum drawdown
See above Four combinations Very much higher optimal SG,
often exceeding 12%
Page 20
Utility Parameters
Figure 10 reports the optimal SG estimates at differing utility parameters. The matrix of revised optimal SG
estimates across income and member objectives appears at the top, with changes from the baseline estimates
appearing below with ‘..’ indicating no change. Grey shading highlights the objectives that we consider to be
most relevant at each income level. Changing loss aversion has only small effects. This stems from our set-up.
The reference dependent analysis primarily balances utility losses pre-retirement versus post-retirement, while
the modelling produces a tendency for the member to draw the target until the funds run out (see Figure 5).
Outcomes that indicate above-target consumption only occur in a modest proportion of the simulations (if
required by the minimum drawdown rules). When this occurs, the gain is heavily discounted by the curvature
parameter applied. This effectively means that the optimal SG remains dominated by outcomes in the realm of
losses. Hence only minor changes occur in response to the varying the reference dependent utility parameters,
with at most a ±0.5% change in the optimal SG occurring in some instances.14
Figure 10: Sensitivity to Utility Parameters
14 We also tested the parameters of Tversky and Kahneman (1992), which include a weighting parameter on losses of 2.25
and a curvature parameter of 0.88 on both gains and losses. These parameters often deliver an optimal SG of 20% and above.
Investigation reveals that this stems from the combined effect of only a small discount being applied to above-target
consumption, and the fact that compounding returns over long horizons can generate considerable wealth accumulation. The
consequence is that utility then becomes maximised through investing to access post-retirement consumption that is well
above-target, i.e. the member becomes willing to take on risk in pursuit of higher post-retirement consumption in order to
boost lifetime utility of consumption. These results seem at odds with both observed behaviour and the purported purpose of
the SG, and hence we do not report them. They also suggest that the Tversky and Kahneman (1992) parameters may be
unsuitable for addressing multi-period investment problems where returns compound over time.
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Higher loss aversi on
70% Replacement Rate 0.0% 0.5% 3.5% 4.5% 6.0% 6.5% 7.5% 8.0% 8.5%
ASFA Comfortable 15.0% 12.0% 9.0% 7 .0% 6.0% 5.0% 4.5% 4.0% 3.5%
ASFA Modest 3.0% 2.5% 2.0% 1.5% 1.0% 1.0% 1.0% 1.0% 1.0%
Lower loss aversi on
70% Replacement Rate 0.0% 0.5% 3.0% 5.0% 6.0% 7.0% 7.5% 8.5% 9.0%
ASFA Comfortable 15.0% 12.0% 9.5% 7 .5% 6.0% 5.5% 4.5% 4.0% 3.5%
ASFA Modest 3.0% 2.5% 2.0% 1.5% 1.5% 1.0% 1.0% 1.0% 1.0%
Higher discount rate
70% Replacement Rate 0.0% 0.0% 2.0% 3.0% 3.5% 3.5% 4.0% 4.0% 4.0%
ASFA Comfortable 9.0% 6.5% 5.0% 4.0% 3.5% 3.0% 2.5% 2.5% 2.0%
ASFA Modest 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
Higher ri sk or l oss aversion
70% Replacement Rate .. .. .. -0.5% .. .. .. .. ..
ASFA Comfortable .. .. .. .. .. .. .. .. ..
ASFA Modest .. .. .. .. .. .. .. .. ..
Lower risk or loss aversion
70% Replacement Rate .. .. -0.5% .. .. 0.5% .. 0.5% 0.5%
ASFA Comfortable .. .. 0.5% 0.5% .. 0.5% .. .. ..
ASFA Modest .. .. .. .. 0.5% .. .. .. ..
Higher discount rate
70% Replacement Rate .. -0.5% -1.5% -2.0% -2.5% -3.0% -3.5% -4.0% -4.5%
ASFA Comfortable -6.0% -5.5% -4.0% -3.0% -2.5% -2.0% -2.0% -1.5% -1.5%
ASFA Modest -3.0% -2 .5% -2.0% -1.5% -1.0% -1.0% -1.0% -1.0% -1.0%
Income Level
Page 21
Imposing a discount rate of 2% (time preference parameter of 0.98) makes more of a difference, reducing the
optimal SG by between -1.5% and -4.5% across the most relevant objectives as shaded in grey. At income L5 for
example, a 2% discount rate reduces the optimal SG by -2.5% under the replacement rate and ASFA comfortable
targets. This provides a guide to the effect of our baseline assumption that the member has no preference in favour
of current consumption over future consumption. Further, this sensitivity test applies a consistent discount rate as
typically specified under rational models. It is possible that behavioural effects such as hyperbolic discounting
(see Laibson, 1997) might heighten the preference for the present over the future even further.
Required Income Stream
We undertake five sensitivity tests designed to evaluate changes to the income stream that the member requires.
Results are reported in Figure 11.
Differing replacement rates – We test both 80% and 60% replacement rates, versus the baseline assumption
of 70%. This changes the optimal SG by a magnitude of between 1.5% and 3.0% across the relevant income
range spanning income L3 and above. Increasing the replacement rate to 80% generates an optimal SG above
9.5% only at income L7 and above, but still falls short of indicating an SG of 12%. The higher 80%
replacement rate also indicates what might occur if the 70% replacement rate was based on average pre-
retirement income from age 25 to 66, which would be equivalent to applying a replacement rate of about 77%
relative to average income in the 5-years prior to retirement.
Applying the minimum drawdown rulesAssuming that the member draws income from their retirement
account in line with minimum drawdown rules generates changes in the optimal SG that are mixed in sign,
but with a decrease occurring in the majority of instances. The mixed results reflect interactions between
constraining post-retirement consumption in cases where the minimum drawdown rate is not binding, and the
manner in which pre-retirement and post-retirement consumption are traded off under the utility function. The
main reason for the decreases in optimal SG is that the minimum drawdown rules constrain consumption to
below the target, making it less attractive to sacrifice (certain) pre-retirement consumption in order to gain an
increase post-retirement consumption that occurs mainly later in life, and hence are down-weighted by
mortality. In essence, our model is indicating that there is little point in forcing members to save more if they
are unlikely to make use of those savings (given the probability of dying before they drawdown fully on their
retirement account). Also contributing to the results is a relatively flat utility curve flat under this sensitivity
test, meaning that small changes can produce sizeable shifts in the optimal SG.
Drawing down the targetThis sensitivity test assumes in all instances that the member draws the maximum
of any consumption target or the minimum drawdown rate in. Some small increases in the optimal SG result,
typically of 0.5%. This reflects the fact that the optimal drawdown strategy under the reference dependent
objectives is to draw the target in most instances anyway (see Figure 5), except under limited circumstances
such as where high investment returns make it optimal to further increase drawdowns.
Self-insuring to age 102 The baseline analysis assumes that the member takes into consideration their
mortality rate in accordance with the 2015-17 Australian Life Tables. This implies that the member optimises
with an eye on their life expectancy, with outcomes later in retirement having lesser influence because the
probability of survival decreases with age. This set-up might be interpreted as underweighting any desire to
hedge against longevity risk, i.e. the need to save enough to ensure that the funds last ‘just in case’ of survival
to a very old age. To test the impact of a desire to self-insure against longevity risk, we run the analysis
assuming that the member is certain to live to age 102. According to the 2015-17 Australian Life Tables, the
probability of survival to age 102 is 3.6%. Thus, we consider this a conservative test to the extent that few
members might want to self-insure against such a low probability event, while many could be satisfied with
the Age Pension coupled if they happen to reach a very old age. Modelling as if the member expects to live to
age 102 generates mixed but generally meaningful increases in the optimal SG, ranging from +1.0% to +6.0%
across the most relevant objectives. Impacts of +3% or greater are observed under the replacement rate
objective at income L6 and above, and under the ASFA comfortable objective at income L4 and below. This
takes the optimal SG to 9.5% or greater for these estimates. The average increase in the optimal SG versus the
baseline is +2.9% across the most relevant objectives as shaded in grey.
Page 22
Figure 11: Sensitivity to Assumptions Regarding the Required Income Stream
Early retirement at age 62 This sensitivity test investigates the circumstance where the member spends
five years less than the full term earning income and hence contributing, and supports consumption over that
period entirely by drawing down on their superannuation balance. Assuming retirement 5-years earlier at age
62 has a significant impact on the optimal SG estimates, increasing them by between 3.0% and 8.0% across
the most relevant objective range as shaded in grey. Under the replacement rate objective, all increases sit in
the range of +4% to +5.5%. Increases of +7.5% and +5.5% are observed for ASFA modest at L1 and L2,
noting that low income earners are now required to save in order to support consumption over 5-years with no
access to the Age Pension. The increase under ASFA comfortable is +8.0% at L3, declining to +3.0% at L9.
The average increase in the optimal SG versus the baseline is +5.2% across the most relevant objective range;
and the majority of optimal SG estimates now sit above 9.5%. An SG of to 12% is indicated at income L6 and
above under the replacement rate objective, and at L4 and below under AFSA comfortable.
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Differing replacement rates
80% Replacement Rate 0.0% 2.5% 5.0% 6.5% 8.0% 9 .0% 10.0% 10.5% 11.5%
60% Replacement Rate 0.0% 0.0% 1.0% 3.0% 4.0% 5 .0% 5.5% 6.0% 6.5%
Mini mum drawdown rules
70% Replacement Rate 0.0% 0.5% 3.5% 4.5% 4.5% 4 .0% 4.0% 10.0% 11.5%
ASFA Comfortable 11.0% 8.5% 6.5% 5.0% 4.5% 3.5% 3.5% 3.0% 2.5%
ASFA Modest 3.5% 2.5% 2.0% 1.5% 1.5% 1.0% 1.0% 1.0% 1.0%
Drawing down the target
70% Replacement Rate 0.0% 0.5% 3.0% 5.0% 6.0% 7 .0% 7.5% 8.5% 9.0%
ASFA Comfortable 15.5% 12.5% 9.5% 7.5% 6.0% 5.5% 4.5% 4.0% 3.5%
ASFA Modest 3.0% 2.5% 2.0% 1.5% 1.5% 1.0% 1.0% 1.0% 1.0%
Self-i nsure to age 102
70% Replacement Rate 0.0% 0.5% 4.5% 6.5% 8.5% 10.5% 12.0% 13.5% 14.5%
ASFA Comfortable 20.0% 18 .0% 13.5% 10.5% 8.5% 7.5% 6.5% 6.0% 5.0%
ASFA Modest 5.0% 3.5% 3.0% 2.0% 2.0% 1.5% 1.5% 1.5% 1.0%
Retire earl y at age 62
70% Replacement Rate 4.0% 4.5% 8.5% 9.5% 10.5% 12.0% 13.0% 13.5% 14.0%
ASFA Comfortable 20.0% 20 .0% 17.0% 13.5% 11 .0% 9.5% 8.0% 7.0% 6.5%
ASFA Modest 10.5% 8.0% 6.0% 5.0% 4.0% 3.5% 3.0% 2.5% 2.5%
Differing replacement rates
80% replacement rate .. 2.0% 1.5% 1.5% 2.0% 2.5% 2.5% 2.5% 3.0%
60% replacement rate .. -0.5% -2.5% -2.0% -2.0% -1.5% -2.0% -2.0% -2.0%
Mini mum drawdown rules
70% Replacement Rate .. .. .. -0.5% -1.5% -2.5% -3.5% 2.0% 3.0%
ASFA Comfortable -4.0% -3.5% -2.5% -2.0% -1.5% -1.5% -1.0% -1.0% -1.0%
ASFA Modest 0.5% .. .. .. 0.5% .. .. .. ..
Drawing down the target
70% Replacement Rate .. .. -0.5% .. .. 0.5% .. 0.5% 0.5%
ASFA Comfortable 0.5% 0.5% 0.5% 0.5% .. 0.5% .. .. ..
ASFA Modest .. .. .. .. 0.5% .. .. .. ..
Self-i nsure to age 102
70% Replacement Rate .. .. 1.0% 1.5% 2.5% 4.0% 4.5% 5.5% 6.0%
ASFA Comfortable 5.0% 6.0% 4.5% 3.5% 2.5% 2.5% 2.0% 2.0% 1.5%
ASFA Modest 2.0% 1.0% 1.0% 0.5% 1.0% 0.5% 0.5% 0.5% ..
Retire earl y at age 62
70% Replacement Rate 4.0% 4.0% 5.0% 4.5% 4.5% 5 .5% 5.5% 5.5% 5.5%
ASFA Comfortable 5.0% 8.0% 8.0% 6.5% 5.0% 4.5% 3.5% 3.0% 3.0%
ASFA Modest 7.5% 5.5% 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.5%
Income Level
Page 23
The early retirement sensitivity test is important as it covers for a range of situations where income is impaired,
thus leading to lower contributions to the superannuation fund coupled with a likely need to fund consumption
out of savings (or borrowings). In addition to early retirement, career breaks and periods of part-time employment
can also leave a member in a comparable position. We contend that basing our sensitivity test around retiring 5-
years early while fully-funding consumption from the superannuation account amounts to a very conservative
evaluation of these situations. First, the assumption that income falls to zero and the member then draws down
on their savings to maintain a targeted level of consumption is an extreme assumption. It is possible that many
members who incur a loss of wage income may receive support from other sources, including social security such
as unemployment benefits or carer’s payments, paid maternity leave, redundancy payments, and so on. This might
particularly be the case for lower income earners, where we do not model all the support payments that they may
have available. Many members could also have latitude to reduce consumption below the assumed target during
the time they are not earning income. Second, our modelling assumes that superannuation is the only source of
savings. In practice, many people will have access to other forms of savings on which they might draw. Third,
those suffering career breaks may have the ability to top-up superannuation contributions at a later date. We thus
believe that our early retirement analysis might be seen as establishing a reasonable upper limit on the potential
impact on the optimal SG of the risk of failing to remain in full-time employment over the full working term.
A related point is we understand that lower income earners are more likely to retire early, but also have lower life
expectancy. Thus, while lower income earners may be more exposed to early retirement risk, they may also be
less exposed to longevity risk. Under these circumstances, the risk of under-saving in case of early retirement
may be partly offset by a lesser need for saving to allow for the possibility of living to an old age.
Investment Assumptions
The sensitivity analysis around the investment assumptions takes two directions, with the results appearing in
Figure 12. The first is varying the expected risky asset expected returns by ±1.5% relative to the baseline of 5.0%
compound. This translates to a reduction in total portfolio returns of a bit over ±1% under the constant mix of
70% in the risky asset and 30% in the risk-free asset assumed in the baseline analysis. Increasing the risky asset
returns to 6.5% decreases the optimal SG by up to -1.5% across the range of most relevant objectives. Conversely,
lowering the real risky asset return 3.5% increases the optimal SG by a comparable amount.
The second direction applies alternative investment strategies. Imposing an optimised asset mix has the effect of
increasing risky asset exposure and hence portfolio returns, which reduces the optimal SG. The optimisation
generates an optimal glide path that is u-shaped and centred around the point of retirement, where the median
risky asset weight at L5 is 54% under the replacement rate, 59% under ASFA comfortable and 88% under ASFA
modest. The average risky asset weights over the life-cycle are relatively high, averaging between 75% and 85%
across the three objectives. The optimal SG decreases by up to -2.0% across the most relevant objective range
(average decrease of -1.3% versus the baseline). We also apply a life-cycle strategy entailing a 90% weight in the
risky asset until 25-years prior to retirement (i.e. age 42), then linearly decreasing to reach a risky asset weight of
40% at retirement (i.e. age 67) and beyond. This has the effect of reducing risky asset exposure to an average of
58% over the life-cycle (lower still on an asset-weight basis), which reduces portfolio returns. This produces an
increase in the optimal SG of up to 1.0% across the most relevant objective range (average increase of +0.5%).
The investment sensitivity results reflect the need for higher savings to attain a given post-retirement consumption
target when investment returns are lower. They also highlight the potential impact of two aspects. First is the
increase in SG that might be needed to cater for the risk that investment returns are lower going forward than
experienced historically. As a rough rule of thumb, the SG might be set about 1% higher if superannuation fund
returns are expected to be 1% lower. Second is the effect of catering for the possibility that members might choose
sub-optimal investment strategies, perhaps due to behavioural biases. The difference in the optimal SG between
the life-cycle and optimal investment strategy is up to +3% across the most relevant range (averaging about +2%).
The life-cycle strategy might be taken as a guide to the conservative investment strategies followed by some
members, to the extent that it de-risks as retirement approaches and maintains a relatively conservative asset mix
throughout retirement. Our model suggests that it is optimal for members to maintain an asset mix that can
generate higher returns over the long run, and limit the degree of de-risking going into retirement. The 70/30 asset
mix used in the baseline analysis sits is the middle ground. The issue that arises for setting the SG is whether
policy should accommodate sub-optimal investment behaviour by members, versus being based on a more
optimal asset mix, perhaps coupled with other measures to encourage better investment decisions.
Page 24
Figure 12: Sensitivity to the Investment-Related Assumptions
Policy Environment
The sensitivity testing presented so far assumes access to the Age Pension, and that the member bears the full
cost of the SG which directly lowers their take-home pay below what it otherwise would have been. Testing the
sensitivity to these assumptions reveals both to be important, with the Age Pension being critical. We use the
heading ‘policy environment’ for lack of a better term, noting that these two aspects relate to whether the policy
objective might be to ensure more people become self-funded retirees, as well as how any increase in the SG is
structured. It is also plausible that some members might aspire to becoming self-funding retirees, and hence prefer
not to take the Age Pension into account in deciding how much to place in their superannuation fund. Another
interpretation of the analysis excluding the Age Pension is that it reveals the total amount of savings required to
maximise utility under each member objective.
Figure 13 reports estimates of the optimal SG where the value of the Age Pension and related supplements are
set to zero. The impact is particularly large, with the optimal SG increasing between 5.0% and 17.0%, with an
average of +8.7% across the most relevant objectives. The optimal SG sits at 9.5% across all estimates with the
exception of ASFA comfortable at income L9; and above 12% except for ASFA comfortable at income L7 to L9.
This confirms the importance of the Age Pension in the generation of the baseline estimates, which indicate
optimal SGs in the 2.5% to 9% range. It highlights the significant value of the Age Pension to Australian retirees.
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Higher ri sky asset return
70% Replacement Rate 0.0% 0.5% 2.5% 4.0% 5.0% 5.5% 6.0% 7.0% 7.5%
ASFA Comfortable 12.5% 10.0% 7.5% 6.0% 5.0% 4.5% 3.5% 3.5% 3.0%
ASFA Modest 3.0% 2.0% 1.5% 1.5% 1.0% 1.0% 1.0% 1.0% 0.5%
Lower risky asset return
70% Replacement Rate 0.0% 0.0% 3.5% 5.5% 6.5% 7.5% 8.5% 9.0% 9.5%
ASFA Comfortable 17.0% 14.0% 10.5% 8.0% 7.0% 6.0% 5.0% 4.5% 4.0%
ASFA Modest 2.0% 1.5% 1.5% 1.0% 1.0% 1.0% 0.5% 0.5% 0.5%
Optimi sed asset mix
70% Replacement Rate 0.0% 0.5% 3.5% 3.5% 4.5% 5.0% 5.5% 6.0% 6.5%
ASFA Comfortable 11.5% 9.0% 7.0% 5.5% 4.5% 4.0% 3.5% 3.0% 3.0%
ASFA Modest 2.5% 2.0% 1.5% 1.0% 1.0% 1.0% 1.0% 0.5% 0.5%
Life-cycle: 90/10 to 40/60
70% Replacement Rate 0.0% 0.0% 3.5% 5.5% 6.5% 7.5% 8.0% 9.0% 9.5%
ASFA Comfortable 17.0% 13.5% 10.0% 8.0% 6.5% 5.5% 5.0% 4.5% 4.0%
ASFA Modest 2.5% 2.0% 1.5% 1.0% 1.0% 1.0% 1.0% 1.0% 0.5%
Higher ri sky asset return
70% Replacement Rate .. .. -1.0% -1.0% -1.0% -1.0% -1.5% -1.0% -1.0%
ASFA Comfortable -2.5% -2.0% -1.5% -1.0% -1.0% -0.5% -1.0% -0.5% -0.5%
ASFA Modest .. -0.5% -0.5% .. .. .. .. .. -0 .5%
Lower risky asset return
70% Replacement Rate .. -0.5% .. 0.5% 0.5% 1.0% 1.0% 1.0% 1.0%
ASFA Comfortable 2.0% 2.0% 1.5% 1.0% 1.0% 1.0% 0.5% 0.5% 0.5%
ASFA Modest -1.0% -1.0% -0.5% -0.5% .. .. -0.5% -0.5% -0.5%
Optimi sed asset mix
70% Replacement Rate .. .. .. -1.5% -1.5% -1.5% -2.0% -2.0% -2.0%
ASFA Comfortable -3.5% -3.0% -2.0% -1.5% -1.5% -1.0% -1.0% -1.0% -0.5%
ASFA Modest -0.5% -0.5% -0.5% -0.5% .. .. .. -0.5% -0.5%
Life-cycle: 90/10 to 40/60
70% Replacement Rate .. -0.5% .. 0.5% 0.5% 1.0% 0.5% 1.0% 1.0%
ASFA Comfortable 2.0% 1.5% 1.0% 1.0% 0.5% 0.5% 0.5% 0.5% 0.5%
ASFA Modest -0.5% -0.5% -0.5% -0.5% .. .. .. .. -0.5%
Income Level
Page 25
Figure 13: Sensitivity to Removing the Age Pension
Figure 14 reports the optimal SG when it is assumed that a portion of the cost is borne by employers (implying
increased labour costs to business), with the remainder paid for by the member by reducing the income available
for consumption. We compile estimates under the assumption that the employer bears 25%, 50% and 75% of the
cost, noting that 100% would indicate an SG as high as possible from the member’s perspective. Unsurprisingly,
the optimal SG increases with the proportion of cost borne by the employer. The magnitude of the change also
varies with both income and member objective. Under the 50% assumption for instance, the increase in optimal
SG sits in the +1% to +4.5% range under the most relevant objectives as shaded in grey, averaging +2.5% versus
the baseline. This is sufficient to lift the optimal SG above the current level of 9.5% at L6 and above under the
replacement rate objective, and at L4 and below under ASFA comfortable. These results are hardly surprising, as
the member receives higher savings at diminished personal cost. Our model does not account for any associated
macroeconomic effects on profits, employment, inflation, or overall consumption and savings that might impose
a burden through other channels. This complicates interpreting the estimates, to the extent it is possible that some
of the cost initially borne employers might be passed back to members via higher prices or reduced employment.
Figure 14: Sensitivity to Portion of Cost Borne by Employer
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
70% Replacement Rate 14.0% 16.0% 15.0% 14.5% 14.0% 14.0% 14.0% 13.5% 13.5%
ASFA Comfortable 20.0% 20.0% 20.0% 17.5% 14.0% 12.5% 10.5% 9.5% 8.5%
ASFA Modest 20.0% 18.0% 13.5% 10.5% 9.0% 7.5% 6.5% 6.0% 5.0%
70% Replacement Rate 14.0% 15.5% 11.5% 9.5% 8.0% 7.5% 6.5% 5.5% 5.0%
ASFA Comfortable 5.0% 8.0% 11.0% 10.5% 8.0% 7.5% 6.0% 5.5% 5.0%
ASFA Modest 17.0% 15.5% 11.5% 9.0% 8.0% 6.5% 5.5% 5.0% 4.0%
Income Level
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Employer co ntri butes 25%
70% Replacement Rate 0.0% 0.5% 3.5% 5.5% 7.0% 8.0% 9.0% 1 0.0% 10.5%
ASFA Comfortable 18.0% 14.5% 10.5% 8.5% 7.0% 6 .0% 5.5% 4.5% 4.0%
ASFA Modest 4.0% 3.0% 2.0% 2.0% 1.5% 1.5% 1.0% 1.0% 1.0%
Employer co ntri butes 50%
70% Replacement Rate 0.0% 0.5% 4.5% 6.5% 8.0% 10.0% 11.0% 12.0% 13.0%
ASFA Comfortable 20.0% 17.5% 13.0% 10.0% 8.5% 7.0% 6.5% 5.5% 5.0%
ASFA Modest 5.0% 4.0% 3.0% 2.5% 2.0% 1.5% 1.5% 1.5% 1.0%
Employer co ntri butes 75%
70% Replacement Rate 0.0% 0.5% 5.5% 8.5% 11 .0% 13.5% 15.0% 16 .5% 17.5%
ASFA Comfortable 20.0% 20.0% 17.5% 13.5% 11.0% 9.5% 8.5% 7.5% 6.5%
ASFA Modest 7.5% 5.5% 4.0% 3.5% 3.0% 2.5% 2.0% 2.0% 1.5%
Employer co ntri butes 25%
70% Replacement Rate .. .. .. 0.5% 1.0% 1.5% 1.5% 2.0% 2.0%
ASFA Comfortable 3.0% 2.5% 1.5% 1.5% 1.0% 1.0% 1.0% 0.5% 0.5%
ASFA Modest 1.0% 0.5% .. 0.5% 0.5% 0.5% .. .. ..
Employer co ntri butes 50%
70% Replacement Rate .. .. 1.0% 1.5% 2.0% 3.5% 3.5% 4.0% 4.5%
ASFA Comfortable 5.0% 5.5% 4.0% 3.0% 2.5% 2.0% 2.0% 1.5% 1.5%
ASFA Modest 2.0% 1.5% 1.0% 1.0% 1.0% 0.5% 0.5% 0.5% ..
Employer co ntri butes 75%
70% Replacement Rate .. .. 2.0% 3.5% 5.0% 7.0% 7.5% 8.5% 9.0%
ASFA Comfortable 5.0% 8.0% 8.5% 6.5% 5.0% 4.5% 4.0% 3.5% 3.0%
ASFA Modest 4.5% 3.0% 2.0% 2.0% 2.0% 1.5% 1.0% 1.0% 0.5%
Income Level
Page 26
The sensitivity testing investigates the impact of changing one assumption at a time. We now adjust combinations
of assumptions. The first set of scenarios assumes that the SG is used to hedge against a combination of risky
asset returns being -1.5% lower coupled with either longevity risk, as represented by assuming death at age 102,
or early retirement (and other income-related risk) by assuming retirement at age 62. We do not include hedging
against longevity and early retirement at the same time as this would amount to an element of ‘doubling-up’,
noting that lower income earners are more likely to both retire and die earlier. Figure 15 reveals that using the SG
to hedge against investment risk plus one other risk boosts the majority of estimates to above 9.5% across the
most relevant range, and might provide support for increasing the SG to 12%. This indicates that an SG level in
the 9.5%-12% range could be justified if members were expected to self-insure against at least two of these three
key risks.
Figure 15: Hedging Scenarios
The second set of ‘kitchen sink’ scenarios alters combinations of the following assumptions, with the results
reported in Figure 16:
Excluding the Age Pension and related supplements (denoted ‘Pension’ in Figure 16)
Employers bear 50% of the cost of the SG (denoted ‘cost’)
Self-insuring against living longevity risk to age 102 (denoted ‘age 102’)
Self-insuring against early retirement at age 62 (denoted ‘age 62’)
Lowering real returns on the risky asset by 1.5% to 3.5% compound (denoted ‘return’)
Drawdowns at the minimum drawdown rate (denoted ‘min’)
Unsurprisingly, the optimal SG estimates turn out to be substantially higher, sometimes hitting the imposed
maximum of 20%. We consider these to be an aggressive set of assumptions that deliver optimal SG estimates
which seem questionably high. The point of this exercise is to demonstrate that an increase in the SG might be
justified if it were designed to cover for a range of additional considerations that were not incorporated in our
baseline case. This provides a segue into the next section, which discusses the issue faced by policy makers of
‘what is in, and what is out’ in deciding whether to go through with plans to increase the SG to 12%.
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Lower returns + death age 102
70% Replacement Rate 0.0% 0.5% 5.5% 8.0% 10.0% 12.5% 14.5% 16.0% 17.5%
ASFA Comfortable 20.0% 20.0% 16.5% 12.5% 10.5% 9.0% 8.0% 7.0% 6.5%
ASFA Modest 5.5% 4.5% 3.5% 2.5% 2.0% 2.0% 1.5% 1.5% 1.5%
Lower returns + retire age 62
70% Replacement Rate 4.0% 5.0% 10.0% 11.0% 12.5% 14.0% 15.0% 15.5% 16.5%
ASFA Comfortable 20.0% 20.0% 20.0% 15.5% 12.5% 11.0% 9.5% 8.5% 7.5%
ASFA Modest 11.0% 8.5% 6.5% 5.0% 4.5% 3.5% 3.5% 3.0% 2.5%
Lower returns + death age 102
70% Replacement Rate .. .. 2.0% 3.0% 4.0% 6.0% 7.0% 8.0% 9.0%
ASFA Comfortable 5.0% 8.0% 7.5% 5.5% 4.5% 4.0% 3.5% 3.0% 3.0%
ASFA Modest 2.5% 2.0% 1.5% 1.0% 1.0% 1.0% 0.5% 0.5% 0.5%
Lower returns + retire age 62
70% Replacement Rate 4.0% 4.5% 6.5% 6.0% 6.5% 7.5% 7.5% 7.5% 8.0%
ASFA Comfortable 5.0% 8.0% 11.0% 8.5% 6.5% 6.0% 5.0% 4.5% 4.0%
ASFA Modest 8.0% 6.0% 4.5% 3.5% 3.5% 2.5% 2.5% 2.0% 1.5%
Income Level
Page 27
Figure 16: Kitchen Sink Scenarios
7. Deciding on the SG Rate
Whether the SG should be increased to 12% as planned is currently a matter of considerable debate. Our analysis
indicates that the issue arises against a background where the impact of the SG on members may vary substantially
depending on member objectives and income levels, and where the appropriate level of the SG also depends on
the underlying assumptions. Our primary recommendation is that the policy objective of the SG needs to be
more clearly specified. How the interests of various members are to be traded-off is also pivotal, given that the
impact of the SG is not evenly felt. Other considerations such as the broader effects on employers, the economy
and the government budget might be also addressed. But arguably these are not as important as establishing clear
policy objectives and deciding how to deal with disparate welfare effects. We suggest the following five
considerations be addressed in making any decision on whether to increase the SG:
(i) Define the policy objective that the SG is trying to achieve – This is a fundamental issue with various
What is the superannuation system trying to facilitate? Two key choices are either (a) catering for post-
retirement living standards at a level related to that enjoyed pre-retirement, implying a replacement rate
Objective L1 L2 L3 L4 L5 L6 L7 L8 L9
Pension/age 102/return/min
70% Replacement Rate 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0%
ASFA Comfortable 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 19.5% 17.0% 15.0%
ASFA Modest 20.0% 20.0% 20.0% 19.5% 15.5% 13.5% 11.5% 10.5% 9.0%
Pension/age 62/return/min
70% Replacement Rate 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0%
ASFA Comfortable 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 17.5% 15.5%
ASFA Modest 20.0% 20.0% 20.0% 19.5% 16.0% 14.0% 12.0% 10.5% 9.5%
Cost/age 102/return/min
70% Replacement Rate 0.0% 1.0% 7.5% 10.5% 17.5% 20.0% 20.0% 20.0% 20.0%
ASFA Comfortable 20.0% 20.0% 20.0% 20.0% 18.5% 16.0% 14.0% 12.5% 11.0%
ASFA Modest 7.5% 5.5% 4.0% 3.5% 3.0% 2.5% 2.0% 2.0% 1.5%
Cost/age 62/return/min
70% Replacement Rate 0.1% 1.0% 11 .0% 15.5% 20.0% 20.0% 20.0% 20.0% 20.0%
ASFA Comfortable 20.0% 20.0% 20.0% 20.0% 20.0% 20.0% 18.5% 16.0% 14.5%
ASFA Modest 11.0% 8.5% 6.5% 5.0% 4.5% 4.0% 3.5% 3.0% 2.5%
Pension/age 102/return/min
70% Replacement Rate 20.0% 19.5% 16.5% 15.0% 14.0% 13.5% 12.5% 12.0% 11.5%
ASFA Comfortable 5.0% 8.0% 11.0% 13.0% 14.0% 15.0% 15.0% 13.0% 11.5%
ASFA Modest 17.0% 17.5% 18.0% 18.0% 14.5% 12.5% 10.5% 9.5% 8.0%
Pension/age 62/return/min
70% Replacement Rate 20.0% 19.5% 16.5% 15.0% 14.0% 13.5% 12.5% 12.0% 11.5%
ASFA Comfortable 5.0% 8.0% 11.0% 13.0% 14.0% 15.0% 15.5% 13.5% 12.0%
ASFA Modest 17.0% 17.5% 18.0% 18.0% 15.0% 13.0% 11.0% 9.5% 8.5%
Cost/age 102/return/min
70% Replacement Rate .. 0.5% 4.0% 5.5% 11.5% 13.5% 12.5% 12.0% 11.5%
ASFA Comfortable 5.0% 8.0% 11.0% 13.0% 12.5% 11.0% 9.5% 8.5% 7.5%
ASFA Modest 4.5% 3.0% 2.0% 2.0% 2.0% 1.5% 1.0% 1.0% 0.5%
Cost/age 62/return/min
70% Replacement Rate 0.1% 0.5% 7.5% 10.5% 14.0% 13.5% 12.5% 12.0% 11.5%
ASFA Comfortable 5.0% 8.0% 11.0% 13.0% 14.0% 15.0% 14.0% 12.0% 11.0%
ASFA Modest 8.0% 6.0% 4.5% 3.5% 3.5% 3.0% 2.5% 2.0% 1.5%
Income Level
Page 28
target, or (b) ensuring a basic living standard during retirement, implying a fixed target such as the ASFA
standards. There is also a case for considering some combination of the two, emphasising a basic living
standard for lower income earners and some degree of maintenance of living standards for those on higher
incomes. Under this approach, our baseline estimates point toward an SG in the 6%-9% range by applying
an ASFA comfortable objective between L3 and L5 and a replacement rate objective at L5 and above;
but an SG below 3% under ASFA modest at L1 and L2. While insufficient to justify the current SG of
9.5% let alone an increase to 12%, our sensitivity analysis indicates that an increase in the SG to 12%
does receive support if the policy objectives are broadened beyond that implicit in our baseline analysis.
Is the policy objective to reduce reliance on the Age Pension? The case for a higher SG would be
considerably strengthened if the aim was to suppor as many members as possible to become self-funded
retirees, with the Age Pension viewed as a safety net that is used only if really needed. In this case, our
estimates that ignore the availability of the Age Pension come into play from a policy perspective. Under
these estimates, an SG above 12% receives strong support.
Is facilitating self-insurance against risks one of the aims? Another issue is whether the SG should be set
to ensure that individuals save enough to generate adequate retirement income ‘just in case’ they happen
to: (a) suffer poor investment returns, (b) live to a very old age, and/or (c) are forced into involuntary
early retirement. The alternative is that other solutions are sought to deal with these three risks, such as
through social security or various pooling solutions. Our analysis suggests that desire to self-insure
against longevity risk or early retirement risk are the key elements, with each potentially increasing the
optimal SG in the order of 3% to 5% under the member objectives that we consider to be most relevant.
Hedging against the risk of lower investment returns than experienced historically might boost the
required SG by a further 1% or so. However, the benefit of hedging needs to be weighed against the
possibility that members may end up over-saving if the risks of concern do not come to fruition. Over-
saving incurs costs related to unnecessarily decreasing pre-retirement consumption and the possibility of
dying with substantial un-used balances.
Should SG policy accommodate sub-optimal choices by members? Many members do not make optimal
decisions with regard to either investments or drawdowns, at least by comparison with the optimal
strategies that emerge under our model. They appear to invest too conservatively given the long-term
purpose of superannuation, and often fail to drawdown enough. Our analysis suggests that setting policy
on the assumption that members will invest too conservatively indicates that that the SG might be set
about 2% higher than if they were assumed to invest more optimally. On the other hand, accommodating
the possibility that many members may fail to draw down enough due to following the minimum
drawdown rules hints at shaving the SG at the margin, lest it result in some members saving more than
they will use and hence dying with substantial unused balances.
(ii) Decide how to trade off welfare gains and losses across various member types – Given that a ubiquitous
optimal SG level does not exist, the issue of how to trade-off the interests of differing members should be
addressed. Possible approaches include maximising aggregate welfare, minimising losses, or focusing on
the equity considerations between low and high income earners. There is an argument that any compulsory
system should try to benefit the largest number of Australians while prioritising low income earners over
the rich, to the extent that the rich are more capable of looking after themselves. For instance, high income
earners may be more inclined to make additional contributions when required. If emphasis is placed on the
lowest income earners, the case for a higher SG becomes tenuous at least in the presence of the Age Pension,
to the extent that our analysis reveals that the SG may already be too high for members at income L1 and
L2 under an ASFA modest target.15
(iii) Asymmetry between setting the SG higher versus lower If the SG is set too high, there is nothing a
member can do to lower their contributions. If the SG is set too low, a member at least has the option to
contribute more. This would suggest erring on the side of a lower SG; or perhaps building in more flexibility
into the system that allows the SG to be tailored to member circumstances. On the other hand, the reluctance
of members to contribute beyond the mandated minimum due to behavioural influences is also relevant,
15 The situation may change for low income earners who need to pay rent, in which case ASFA modest is likely to be too
low. See discussion in Section 2, in particular footnote 2.
Page 29
raising the question as to what extent it is appropriate for the government to adopt a paternalistic stance
when setting policy.
(iv) Where the burden falls, and how this impacts on the broader economyThe impact of a higher SG will
depend in part on how it is implemented, in particular who pays for any increase. If the member pays, then
our analysis as reported in this paper is relevant in its own right. However, there may also be broader
macroeconomic effects to take into account. If an increase in the SG effectively comes out of profits with
no offsetting reduction in wages, then this might impact on economic activity and employment.
Alternatively, the cost of the SG may be passed through into prices, leading to higher inflation that unwinds
the effective benefit to members by reducing their spending power. Another issue is the broader impact of a
higher SG stemming from lowering consumption and raising savings, including whether members might
respond by substituting superannuation for other forms of saving.
(v) Impact on the government budget Our indicative estimates of the net revenue effects per individual
member suggest that a higher SG might be associated with a net increase in government revenue from a
majority of members over their lifetimes. Also of interest is that this revenue increase would impact more
on lower income earners. However, the numbers are modest, suggesting that the effects may not be overly
significant. How the revenue and outlay impacts associated with a higher SG interact with the various
member cohorts and the demographics of an aging population might also be considered, as well as other
impact such as on corporate taxation and GST receipts.
The above considerations and the evidence from our analysis suggests that justifying an increase in the SG to
12% requires adopting a particular stance on the policy objectives that the SG is aiming to achieve. The two key
objectives in this regard would be treating the policy aim as replacing the Age Pension (or supporting members
to become self-funded retirees), and/or using the SG as a mechanism for members to self-insure against some
combination of investment, longevity and early-retirement risk. Changes in other assumptions will also make a
difference at the margin, but seem unlikely to tip the balance towards increasing the SG in isolation.
8. Conclusion
This study applies a stochastic life-cycle model to estimate the optimal SG for members across nine different
income levels and three member objectives. We gauge the sensitivity of our results to changes in a wide range of
assumptions, and estimate the impact of various SG levels on member welfare and the government budget at an
individual-level. Our main contribution is to identify the factors that matter for determining the appropriate level
of the SG. In doing so, we highlight that the appropriate level can differ substantially across members; and with
the stance taken on what the SG is trying to achieve, and the particular assumptions underpinning the modelling.
We trust that our analysis will inform what has already been an intense policy debate, which is set to continue
under the current Retirement Income Review. To this effect, we suggest five considerations to take into account
in deciding whether the SG should be increased from 9.5% to 12% as proposed. Our results suggest that an
increase in the SG might be justified if the policy objective involves either replacing the Age Pension, or requiring
members to self-insure against various risks. We caution that setting the SG so that members self-insure gives
rise might result in over-saving if the risks do not come to fruition, with its own potential issues and costs. We
raise the point that there may be other solutions for insuring against risks related to social security and pooling,
rather than relying on the SG.
We understand that some of our findings are at odds with a strong belief in some quarters that increasing the SG
is essential to ensure adequacy during retirement. We believe the difference stems from the fact that our analysis:
(a) takes into account the impact of the SG on pre-retirement consumption; (b) treats the Age Pension as an
income stream that will continue to be available to all, rather than as a safety net that should only be accessed
when needed; (c) does not assume an aim of saving enough to self-insure against the risk of either low investment
returns, living to a very old age or early retirement; and, (d) is based on the assumption that the SG will be paid
for by the member out of their take-home pay. We strongly suspect that many of those arguing for a higher SG
are implicitly taking a different position on some, if not all, of these key assumptions.
Page 30
AFSA, 2019. ASFA Retirement Standard: Detailed Budget Breakdowns. Association of Superannuation Funds
of Australia, June 2019. Available at:
Benartzi, S. and Thaler, R.H., 1995. Risk Aversion or Myopia? Choices in Repeated Gambles and Retirement
Investments. Management Science, 45(3), pp.364-381.
Blake, D., Wright, D. and Zhang, Y., 2013. Target-Driven Investing: Optimal Investment Strategies in Defined
Contribution Pension Plans under Loss Aversion. Journal of Economic Dynamics and Control, 37(1), pp.195-
Butt, A. and Khemka, G., 2015. The Effect of Objective Formulation on Retirement Decision Making. Insurance:
Mathematics and Economics, 64, pp.385-395.
Credit Suisse, 2019. Summary Edition Credit Suisse Global Investment Returns Yearbook 2019. Credit Suisse
Research Institute, February 2019. Available at:
Daley, J. and Coates, B. 2018. Money in Retirement: More than Enough. Grattan Institute, November.
Evans, J. and Razeed, A. 2019. A Consumption v. Savings Analysis of Increasing the Superannuation Guarantee
Levy. Economic Papers (in press), see
Kahneman, D. and Tversky A., 1979. Prospect Theory: An Analysis of Decision Under Risk. Econometrica,
47(2), pp. 263-292.
Laibson, D., 1997. Golden Eggs and Hyperbolic Discounting, Quarterly Journal of Economics, 112(2), pp.443-
Mercer, 2019. A Review of Grattan’s Work on Super. Mercer, July.
Rice, A. and Bonarius, N. 2019. What is the Right Level of Superannuation Guarantee? Rice Warner, June.
Tversky, A. and Kahneman, D., 1992. Advances in Prospect Theory: Cumulative Representation of Uncertainty.
Journal of Risk and uncertainty, 5(4), pp.297-323.
Freestone, O., 2018. The Drivers of Life‐Cycle Wage Inequality in Australia. Economic Record, 94(307), pp.424-
OECD, 2012. OECD Pensions Outlook. OECD Publishing, Paris.
Stanford, J. 2019. The Relationship Between Superannuation Contributions and Wages in Australia. Centre for
Future Work, Australia Institute, November.
Warren, G.J., 2019. Choosing and Using Utility Functions in Forming Portfolios. Financial Analysts Journal,
75(3), pp.39-69.
... Reference dependent utility is useful for addressing retirement problems where there exists a specific outcome target, and can be based on the value function component of prospect theory (see Kahneman and Tversky, 1979;Tversky and Kahneman, 1992). 2 This functional form has been used to analyse retirement incomes by Blake et al. (2013) in a UK setting and Butt et al. (2019) and Khemka et al. (2020) in an Australian setting, and appears as equation (1): ...
Full-text available
An approach for designing a menu of comprehensive income products for retirement (CIPRs) is proposed and demonstrated. The approach entails for steps: defining and characterising member types based on selected attributes; specifying a utility function to capture the objectives and preferences of each member type; conducting analysis to identify and evaluate candidate investment and drawdown strategies, and hence select a product design; and communication to members. The latter uses attributes to describe the type of investor for which a product is designed, as well as setting out the key product features and the outcomes it may deliver. JEL classifications: D14, D15, E21, G11, G23
Full-text available
For a retiree who must maintain both investment and longevity risks, we consider the impact on decision making of focusing on an objective relating to the terminal wealth at retirement, instead of a more correct objective relating to a retirement income. Both a shortfall and a utility objective are considered; we argue that shortfall objectives may be inappropriate due to distortion in results with non-monotonically correlated economic factors. The modelling undertaken uses a dynamic programming approach in conjunction with Monte-Carlo simulations of future experience of an individual to make optimal choices. We find that the type of objective targetted can have a significant impact on the optimal choices made, with optimal equity allocations being up to 30% higher and contribution amounts also being significantly higher under a retirement income objective as compared to a terminal wealth objective. The result of these differences can have a significant impact on retirement outcomes.
Full-text available
Assuming loss aversion, stochastic investment and labor income processes, and a path-dependent target fund, we show that the optimal investment strategy for defined contribution pension plan members is a target-driven 'threshold' strategy. With this strategy, the equity allocation is increased if the accumulating fund is below target and decreased if it is above. However, if the fund is sufficiently above target, the optimal investment strategy switches discretely to 'portfolio insurance'. We show that under loss aversion, the risk of failing to attain the target replacement ratio is significantly reduced compared with target-driven strategies derived from maximizing expected utility.
Full-text available
We study how decision makers choose when faced with multiple plays of a gamble or investment. When evaluating multiple plays of a simple mixed gamble, a chance to win x or lose y, subjects show a sensitivity to the amount to lose on a single trial, holding the distribution of returns for the portfolio constant; that is, they display "myopic loss aversion." Many subjects who decline multiple plays of such a gamble will accept it when shown the resulting distribution. This analysis is applied to the problem of retirement investing. We show that workers invest more of their retirement savings in stocks if they are shown long-term (rather than one-year) rates of return.
Full-text available
Hyperbolic discount functions induce dynamically inconsistent preferences, implying a motive for consumers to constrain their own future choices. This paper analyzes the decisions of a hyperbolic consumer who has access to an imperfect commitment technology: an illiquid asset whose sale must be initiated one period before the sale proceeds are received. The model predicts that consumption tracks income, and the model explains why consumers have asset-specific marginal propensities to consume. The model suggests that financial innovation may have caused the ongoing decline in U. S. savings rates, since financial innovation increases liquidity, eliminating commitment opportunities. Finally, the model implies that financial market innovation may reduce welfare by providing “too much” liquidity.
The Superannuation Guarantee Levy (SGL) is scheduled to be increased from 2025, and there is evidence that an increase could be offset against wages. This paper uses a dynamic model to estimate the distribution of the impact of the SGL increase on both pre‐ and post‐retirement standards of living. The paper shows the increase in the SGL rate has the potential to reduce current consumption for the mean household below the “first level of financial stress” (derived from ABS (6530) Table 11.4) whilst only marginally increasing post‐retirement consumption. The SGL increase may not be an acceptable trade‐off between current consumption and retirement savings.
Utility functions offer a means to encode objectives and preferences in investor portfolios. The functions allow one to place a score on outcomes and then identify optimal portfolios by maximizing utility. The central theme of this article is that utility functions should be tailored to the investor. I discuss how an appropriate function might be chosen and demonstrate concepts for power utility and reference-dependent utility. A modeling approach is presented that may be applied without resorting to dynamic optimization. The selection of utility functions is illustrated for four investor types.
Why does wage inequality rise with age? Using panel data on male wages, we explore the relative importance of unobserved worker heterogeneity versus random wage shocks in explaining this life‐cycle trend in Australia. While we find significant heterogeneity in wage levels (via differences in starting wages), we find no evidence of systematic heterogeneity in wage growth. Instead, highly persistent wage shocks are found to account entirely for the rise in wage inequality with age. We also find evidence that the upward trend in wage inequality since the early 2000s reflects an increase in permanent rather than transitory wage inequality.
We develop a new version of prospect theory that employs cumulative rather than separable decision weights and extends the theory in several respects. This version, called cumulative prospect theory, applies to uncertain as well as to risky prospects with any number of outcomes, and it allows different weighting functions for gains and for losses. Two principles, diminishing sensitivity and loss aversion, are invoked to explain the characteristic curvature of the value function and the weighting functions. A review of the experimental evidence and the results of a new experiment confirm a distinctive fourfold pattern of risk: risk aversion for gains and risk seeking for losses of high probability; risk seeking for gains and risk aversion for losses of low probability. Copyright 1992 by Kluwer Academic Publishers
Analysis of decision making under risk has been dominated by expected utility theory, which generally accounts for people's actions. Presents a critique of expected utility theory as a descriptive model of decision making under risk, and argues that common forms of utility theory are not adequate, and proposes an alternative theory of choice under risk called prospect theory. In expected utility theory, utilities of outcomes are weighted by their probabilities. Considers results of responses to various hypothetical decision situations under risk and shows results that violate the tenets of expected utility theory. People overweight outcomes considered certain, relative to outcomes that are merely probable, a situation called the "certainty effect." This effect contributes to risk aversion in choices involving sure gains, and to risk seeking in choices involving sure losses. In choices where gains are replaced by losses, the pattern is called the "reflection effect." People discard components shared by all prospects under consideration, a tendency called the "isolation effect." Also shows that in choice situations, preferences may be altered by different representations of probabilities. Develops an alternative theory of individual decision making under risk, called prospect theory, developed for simple prospects with monetary outcomes and stated probabilities, in which value is given to gains and losses (i.e., changes in wealth or welfare) rather than to final assets, and probabilities are replaced by decision weights. The theory has two phases. The editing phase organizes and reformulates the options to simplify later evaluation and choice. The edited prospects are evaluated and the highest value prospect chosen. Discusses and models this theory, and offers directions for extending prospect theory are offered. (TNM)
ASFA Retirement Standard: Detailed Budget Breakdowns. Association of Superannuation Funds of Australia
AFSA, 2019. ASFA Retirement Standard: Detailed Budget Breakdowns. Association of Superannuation Funds of Australia, June 2019. Available at: