A system dynamics model of the individual decisions of a
salaried employee in saving for retirement
Johann de Villiers"', Josephine K Musango”, Alan C Brent?
* Correspondence Author,
' Department of Business Management, Stellenbosch Universi South Africa; Email: judv@ sun.ac.za
? School of Public Leadership; and the Centre for i Energy Studies, University,
South Africa; Email: oudntind manne: om sun.ac.Za
5 Department of Industrial Engineering; and the Centre for R ble Energy Studies,
University; South Africa; Email: Hoe sun.ac.Za
Abstract
This paper analyses the dynamic behaviour of an individual in salaried
employment saving for retirement. The employee decides what proportion of
their salary they will allocate to saving for retirement, and tries to ensure that
their realised retirement income meets their retirement income expectations.
The dynamic behaviour stems from changes that the employee makes in their
allocation in response to changing income levels and investment returns. The
preliminary model results illustrate that an early start to the process of saving
for retirement is important. Employees should evaluate the projected retirement
income periodically and should, when required, make strong and swift
adjustments in order to ensure that they will obtain a reasonable retirement
income. If not, they may be faced with having to take an unpleasant drop in
living standards during retirement.
Keywords: Salaried employees; savings; retirement; system dynamics
1 Introduction
One of the challenges facing everyone in the salaried working population is to
accumulate enough funds during their working life to keep them in their retirement.
Few people seem to meet this challenge adequately. Concern has been expressed in
the popular press about the pension saving behaviour of employees in South Africa.
Pickworth (2013) states that only one in nine employees in South Africa is on track to
accumulate enough pension savings to be able to retire with an adequate income of
about 75% of their current income at the age of 65 years.
There are various reasons why pension provision turns out to be inadequate. One of
these is employees changing jobs frequently and then spending their accumulated
retirement funds instead of preserving this for retirement. According to Pickworth
(2013) less than 10% retirement benefits are preserved when people change jobs in
South Africa. The authorities are currently considering changes in the pension fund
regulations which would encourage or compel employees to preserve their retirement
funds when changing jobs. Measures are also considered which would encourage
employees to save more for their retirement (Leach 2013).
Leach (2013) and Cameron (2013) discuss the various ways in which an individual
can reduce the shortfall in their pension provision. The most obvious remedy would
be to work longer and retire later. This will enable the employee to accumulate more
funds for retirement. It will also reduce the number of years the employee will spend
in retirement. The employee will need a smaller retirement fund to finance their
living costs over the shorter retirement period. Working longer should therefore be
encouraged whenever this is possible. Unfortunately many salaried employees work
for employers with a prescribed retirement age, and this decision is therefore often
not available to the employee.
A second option to manage the shortfall in retirement provision is for retirees to
reduce their standard of living in retirement. The retirees could purchase a life
annuity that pays them a guaranteed pension for the rest of their lives, or specify an
appropriate drawdown rate when managing the retirement funds themselves. They
would then be forced to drop their standard of living to the level they can afford on
this income. The level of life annuity payments varies with provider and with interest
rates. Illustrative figures are provided by 10X Investments (2013). According to these
figures, a 65 year old male purchasing an inflation adjusted annuity to be paid for the
remainder of his life will receive an income of 7.2% of the principal amount per year.
If the annuitant is married and needs an annuity that will also provide for his wife
possibly surviving him, the payment drops to 5.6% of the principal per year. If the
annuitant retires earlier, the pay-out will be smaller still.
Other methods of managing the accumulated savings during retirement are also
possible. De Villiers-Strydom and Krige (2014) study the post-retirement decision in
2
some detail. They undertake a very comprehensive study of the returns that could
historically be obtained in South Africa from different plans to convert the retirement
savings into a retirement income. They consider a living annuity strategy to be the
most efficient. They base their study on drawdown rates of 5.5% per year for males
aged 55 years at retirement, 6.2% for males aged 60 years, and 7.3% for males aged
65 years.
These drawdown rates would have been considered conservative some time ago, but
with current low interest rates even this could be considered too high. Finke, Pfau
and Blanchett (2013) are of the opinion that even a drawdown rate of 4% per year
may be too high to ensure that retirees do not deplete their funds during retirement.
A low drawdown rate means that an even greater capital amount needs to be
accumulated in order to finance a desired lifestyle during retirement. This shifts the
emphasis to a key variable that is within control of the employee: the amount that
they save for retirement while they are working. The individual decision by a salaried
employee about the amounts they save for retirement is therefore the focus of this
paper. Given the complexity of decision making process by the salaried employee,
this paper utilised system dynamics.
System dynamics is a methodology that was developed by Forrester (1961) in order to
support strategic decision making for complex problems (Sterman, 2000; Maani and
Cavana, 2007). The method has been used to study various aspects of the retirement
funding challenge. One of the earliest studies is the research of Shimada, Kameyama,
Uchino, Machida and Watanabe (1990). They use a system dynamics model to
simulate the population of Japan for the years 1963 to 2025 in order to study the
future claims against the national welfare system and the and premiums that will be
received to finance this. Using this model, they study the financial health of the
system, given the challenges of a projected aging population in Japan.
Viehweger and Jagalski (2003) undertake a similar analysis of the German public
pension system after it was reformed in 2002. They model the system (which was
changed from a pure public pay-as-you-go scheme to a reduced public system
augmented with private savings) for the period 2002 to 2050. They find that this
reformed system will be able to pay the pensions as required, at least up to 2030.
Chaim (2006 and 2007) uses system dynamics to focus on_asset-liability
management in pension funds. He applies this specifically to the challenge of
investing in appropriate assets to manage the risk of meeting the obligations of a
pension fund in Brazil.
Sapiri, Kamil, Tahar and Tumin (2010 and 2011) use system dynamics to study the
challenge for an unfunded defined benefit pension fund in Malaysia. The obligations
of this fund depend on the salary increases awarded to fund participants during their
working lives and longevity risk once they have retired. The benefits are paid to
retired government employees from the current budget, and the system dynamics
model is used to determine the expenditures required from this fund in different
scenarios. Sapiri, Kamil and Tahar (2013) expand this analysis by using the system
dynamics model to undertake sensitivity analyses. This enables the plan sponsor to
understand the exposure to the different risk factors.
The present paper focuses on a different challenge. Where the studies cited above
focused on the pension fund that has to meet its obligations, this paper looks at the
challenge of individuals that are saving for their own retirement. This saving can take
place within a defined contribution pension fund, or the saving can be undertaken to
supplement the income that will be received from a defined benefit or defined
contribution fund. The important focus is that we are modelling an individual
decision, and that the individual has to forfeit consumption during their working
lives to finance additional consumption during retirement.
2 The salaried employee saving for retirement
The challenge experienced by a salaried employee to save for retirement is modelled
as a system in which the employee decides what proportion of their income to
allocate to their retirement contributions. The causal loop diagram is shown in
Figure 1, and the sub-sections that follow present the dynamic hypothesis describing
this problem.
Projected conversion
Projected investment of capital to income
retums leading up to
as Belg during retirement
- Projected salary
Tetums on
/ +
+ +
fh “ie ge Retirement
income
+ (2p expectations
}
Actual salary
increases
Figure 1. Causal loop diagram of the system that describes the dynamic behaviour of
a salaried employee that is saving for retirement
2.1 Additional contributions feedback loop (B1)
The causal loop diagram shows three feedback loops that drive the behaviour in this
problem. The most central feedback loop is the additional contributions feedback
loop, through which the employee tries to adjust their total retirement savings in
order to meet their retirement objectives. The model assumes that the employee
makes normal contributions to an employer contribution plan, and can make
additional contributions to supplement this. The employee therefore considers their
accumulated retirement savings, estimates what retirement income this balance and
their current contribution policy will bring them, compares this to their retirement
income expectations to determine a retirement provision gap, and then adjusts the
additional contributions they make in order to try to close this gap. This is a
balancing loop, where, higher accumulated savings result in a higher projected
retirement income, which in turn results in a lower retirement provision gap, which
further results in a lower additional contributions, which lead to lower accumulated
savings.
This additional contributions feedback loop is the primary mechanism that governs
the planning process modelled here. If this loop is not functioning, employees are not
considering whether their current savings behaviour will lead to a retirement income
that will meet their expectations. They are therefore not making any adjustments to
their retirement contributions, and could end up disappointed it their income does
not meet their expectations in retirement.
2.2 Investment income feedback loop (R)
The second important loop in the retirement saving model is the investment income
loop. This is a reinforcing loop. The accumulated retirement savings are invested,
and receive an investment income. More savings means more returns, which means
more savings. This exacerbates the challenge of saving for retirement. If an employee
is “ahead” then this works in their favour, but is they fall behind or start late then this
loop reinforces their savings deficit. The investment income depends on returns in
the market, which is a stochastic variable. The present analysis does not model the
returns as a stochastic variable but allows for the variability of the returns to be
included by specifying the different returns for different years. (Including the
stochastic behaviour of the returns could be considered as a possible extension of the
model in the future.)
2.3 Standard of living feedback loop (B2)
The third feedback loop in the model is perhaps more subtle. It consists of an
employee having to adjust their current standard of living to finance additional
contributions to their retirement savings. Lowering their current standard of living
will also lower their retirement income expectations (because they would have
become accustomed to a lower standard of living, and the effects of this would
continue into retirement).
The standard of living feedback loop is thus a balancing loop. Higher additional
contributions to retirement savings would result in a lower standard of living, a lower
retirement income expectation, a lower retirement provision gap, and less additional
contributions to retirement savings required to close this gap.
2.4 Additional contributions to retirement savings
The most important variable in this model is the additional contributions that the
employees make to their retirement savings (over and above the “normal”
contributions). The form this takes is usually determined by tax or cost
considerations. If the contributions are tax-deductible for income tax purposes then
it would be advantageous to add the additional contributions to the “normal”
retirement fund. If they are not, any other savings vehicle could be considered to
keep these savings.
In the simplified model considered here, we do not distinguish between savings
added to the retirement fund and those saved separately. The additional savings
would therefore simply add to the accumulated retirement savings.
The additional amount that an employee will contribute to retirement funding is a
behavioural issue, and will differ from individual to individual. It depends critically
on how the employee views the future, and the extent to which they discount the
future. In some instances employees will not care about the future (in a “tomorrow
never comes”, or “we shall cross that bridge when we get to it” approach) and will
simply save the minimum prescribed by their employment contract and nothing
more. If that is the response, there is very little dynamic behaviour to model prior to
retirement. The employee will make no additional contributions, will have to make
do with whatever funds they have accumulated when they retire, will very likely end
up with insufficient funds to maintain their standard of living, and will be forced
either to work longer (if that is possible) or suffer a drop in standard of living in
retirement.
An alternative approach as represented here would be followed by pro-active
employees who are concerned about their standard of living in the future and who
are trying to position themselves for this future. They would calculate (either
explicitly or by undertaking a more intuitive assessment) the income they are likely
to receive from their current and future planned retirement savings, the income they
would need in retirement, and the gap between these two. They would then try to
close the gap by making additional contributions to their retirement savings. The
elements of the model used to describe this behaviour are discussed below.
2.5 Projected retirement income
To determine the projected retirement income the employee has to look into the
future and try to predict (either through explicit calculations or intuitively) the
retirement funds that they will be able to accumulate to retirement, and the rate at
which they will be able to convert this into a retirement income when they retire. For
this, the employee has to estimate the additional funds they are still going to
contribute to their retirement funds in the period to retirement, estimate the
investment returns that they will receive on their current retirement fund balance as
well as on the funds they are to contribute in the future (taking into account possible
salary increases in the future), and estimate the interest rates at retirement (because
this will influence life annuity rates as well as appropriate drawdown rates for other
investments). From this they can then calculate the retirement income that they can
be expected to receive from their present retirement plan.
2.6 Retirement income expectation
The employee must also predict (either through explicit calculations or intuitively)
the retirement income that they will expect to receive when they retire. This is
usually defined as a proportion of their expected pre-retirement income. This can
either be a proportion of the gross income they receive before retirement, or a
proportion of the expected income they can use for living expenses pre-retirement.
Once they have decided on a proportion of final income that they want to maintain
during retirement, and their final pre-retirement income, they can specify the
income that they expect to need in retirement.
2.7 Retirement income gap
The employee can now calculate the shortfall in retirement income provision, or the
gap between the retirement income expectation and the projected retirement
income. An employee taking pro-active steps to secure a sufficient income in
retirement would try to close a gap when it exists. To close this gap they will be
making additional contributions to retirement savings when this is required.
Closing the gap is a behavioural response, and difficult to predict or model. There are
infinite ways in which an individual could respond to a perceived gap. A very
common way of responding already referred to above is denial. The explicit or
intuitive assessment of the gap depends on so many assumptions about the future,
that is, it would be very easy for individuals in denial to convince themselves that
these calculations are so speculative that it is not necessary or appropriate to act on
them now. Many individuals also find it difficult to manage their personal
expenditures over a week or a month, and it would be even more difficult for them to
manage their expenditures and savings over a lifetime. Sadly, many individuals
would therefore simply do nothing to close this gap. This is likely to be the principal
reason for the shortfall in retirement provision for eight out of nine employees in
South Africa, as discussed by Pickworth (2013).
There are also many ways in which pro-active employees who want to close the
retirement provision gap can respond. They can close the gap quickly or try to do this
over a longer period. An increase in retirement contributions will require a drop in
living standards. This will limit the strength with which employees can respond. In
addition, there is the question on how to respond once the gap has been closed. If
they plan on reducing the additional retirement contributions they can do so quickly
or make the adjustment over a longer period.
In one way or another, pro-active employees will increase their additional retirement
contributions in response to a perceived retirement income gap. This will result in a
positive relationship between the gap and the additional contributions.
3. Preliminary model
3.1 Stock flow diagram
Figure 2 shows the stock flow diagram of the preliminary model used for the
preliminary analysis of the decisions of a salaried employee saving for their
retirement.
ception pst — Euctl ysaer lneeasas ao per
year
— Riihe \ ran ie
Projctad
= tom copa! —_
Amal addon
th roorenent
combutons
a '
* he am, Nel x
oman os ___—~ rae peryear SB
= ToMamal ee ;
Ber wet
investment ie zd ———~ , Policy
ri we ne ore
} v 3. aigesPan cioess "equa a
__ ims Amal ramet Fal oon F i aren ces ones 4 \
aos oad anlar alent ~ [os
f ey rotiement Pouicy ‘Timo aR Ot
testa ot }atiasay contbons POLICY VARIABLE Log omton
LOOKUP TABLE woot VARIABLE 2 stonhafmspor a
SO egttan | / Futurama | Se LOOKUP TA nN
pecenage fi { conbutons 7 \
| adjusment rate Futter adiiomta ~~ condartfor
Anmal salary \ cor 0 g— ee :
LOOKUP \ = etry EA Snes chy
TABLE ‘ ioc
Adin consbtns
Na. sspoprtonoting g—— Li,
Figure 2. Stock and flow diagram of the system that describes the dynamic behaviour
of a salaried employee that is saving for retirement
The decisions are modelled over the working life of the employee. The primary focus
of the analysis is on the stock variable describing the accumulated retirement
savings. This consists of the accumulation of three flow variables; the flow of
“normal” retirement contributions, the flow of additional retirement contributions,
and the flow of investment income.
3-2 Parameter specification
Various parameters have to be specified and could be adjusted to represent different
scenarios. In the preliminary model considered here, the parameters are either kept
constant or specified in advance in a lookup table. In a more advanced modelling
exercise many of these parameters could be modelled as stochastic variables,
increasing the complexity of the analysis.
The parameters specified in the preliminary model include the salary that the
employee will receive over their working life. This is specified in a lookup table, and
can be varied to represent different career profiles. It is also possible to consider the
retirement savings situation of an employee that only starts saving later in life by
specifying a salary of zero up to that age.
The investment returns received on the fund are also specified in a lookup table as a
variable that can change every year. In the simplified model used here, the varying
annual return is specified in advance and is therefore not treated as a true stochastic
variable.
Other variables that must be specified in advance are the normal contribution ratio
and the employee’s retirement age. The salary increases that the employee expects to
receive in the period up to retirement must be specified, because this determines the
rate at which the employee’s living standard can be expected to increase to
retirement. Together with the desired standard of living continuation ratio (the
proportion of their living expenses before retirement the employee would want to
receive during retirement) this determines the retirement income expectation.
The retirement income gap is the difference between the retirement income
expectation and the projected retirement income. The projected retirement income
depends on the projected retirement capital. The projected retirement capital has to
be estimated periodically by the pro-active employee, so that the appropriate
remedial action to close the retirement income gap can be undertaken when
required.
3-3 Estimating the projected retirement income
It is important to realise that the employee takes action based on their expectation of
the future. The future may play out differently from what they expect, and they then
have to adjust their actions to the new reality. Looking into the future, the employee
estimates the capital that they expect to have acquired at retirement, and use this to
estimate the expected income that they will derive from this during retirement. The
funds already saved for retirement are known when undertaking this estimate, but
the contributions that will be made in the future, how these contributions are likely
to grow due to salary increases, and the investment income that the employee will
likely receive on both the existing capital and the expected new contributions to
retirement. The model therefore requires a view on the expectations that the
employee will have about the future, so that the projected retirement capital (given
11
these expectations) can be determined. The retirement capital that the employee
expects to accumulate over their remaining working life (V,,) is calculated using
equation (1) below.
vom warsrate) +e (228) (1- (LE) seat
Te — Ge 1+7
..(1)
where:
Vay = Value of projected retirement savings at the time of retirement
Va = Value of actual retirement savings in year n
Cy = Annual retirement contributions in year n
a) = Year of retirement
n = Current year
15 = Expected annual return on investment in the period leading up to
retirement
Ge = Expected annual growth in the annual retirement contributions in
the period leading up to retirement
The projected retirement income is then calculated from the projected retirement
savings by means of the expected capital conversion rate. In the simplified model
used here, this is specified as a constant. An appropriate parameter would depend
upon the maximum drawdown rate for a living annuity, or the pay-out rate that can
be obtained for an inflation-linked life annuity.
12
3-4 Modelling the annual additional contributions
The model uses a second stock and flow to model the annual additional retirement
contributions that will be made by the employee. The policy outcome annual
additional retirement contributions is taken as a stock, to which adjustments derived
by the model is added.
The model first calculates the further additional contributions that would ideally be
required. This represents the further contributions that would close the retirement
payment gap at retirement.
If a positive retirement income gap is identified and further contributions required,
the changes in the additional contribution rate can be calculated. The actual
adjustment depends on two response policy variables that have to be specified. The
first is the strength of a response. The additional payments will require a decline in
living expenses, which will cause the employee some pain. The strength of the policy
response depends on how much pain the employee is prepared to accept. This is
specified as a percentage of living expenses in a lookup table. Having determined the
strength of the response, the second policy variable specifies how quickly the
adjustment is to be made.
Where the projected retirement provision is larger than the retirement income
expectation it can be appropriate to reduce the additional retirement rates. Two rates
of reduction are specified. The one specifies the speed with which contribution rates
can be reduced, the other the speed at which the accumulated surpluses can be
eliminated. The minimum of the two rates is then applied as a flow from the annual
additional retirement contribution stock variable.
The stock flow model shown in Figure 2 and described above is used in the section
that follows to determine the savings outcome for different scenarios.
13
4 Preliminary Results
The simulation model can be used to study the dynamic behaviour of the system of
an individual employee saving for their retirement. Various scenarios can be
investigated, and the results of some scenarios are presented below.
4.1 Could the normal contributions be sufficient?
The system models the additional contributions that an employee makes. A pertinent
question is whether these additional contributions are required at all. If the employee
belongs to a retirement fund that requires sufficient “normal” contributions, then it
is quite possible that no additional contribution is required. The first set of results
therefore investigates this possibility by studying the behaviour of the system for
different “normal” contribution rates.
Figure 3 shows the additional contributions required with normal contributions rates
of 15, 20 and 25 percent. The additional contribution is expressed as a percentage of
the living expenses of the employee. Figure 4 shows the development of the
retirement income for the same scenarios.
14
Additional contributions as a proportion of living expenses
2
15
zo
05
0
18 30 41 53 64
Time (Y ear)
Additional contributions as a proportion of living expenses : Normal ion 15 percent
Additional contributions as a proportion of living expenses : Normal ion 20 percent
Additional contributions as a proportion of living expenses : Normal ion 25 percent
Figure 3. Additional contributions as a proportion of the living expenses with
different levels of normal contributions (contributions start at age 25, the
standard of living continuation ratio is 0.75, the contributions can be adjusted
by up to 5 per cent of living expenses per year, adjustment takes place in 1
year)
It is evident from Figure 3 that all the contribution rates considered here are
insufficient to yield enough retirement capital so that a gap does not develop along
the way which the model then tries to close with additional contributions. The model
detects a retirement income gap and calls for a steep increase in additional
contributions to close that gap. In about year 40 there is another increase in
contributions because the specific scenarios modelled here showed a decline in
investment returns in that period. When the anticipated investment returns do not
materialise the gap widens and the model calls for further additional contributions.
The gap is closed in about year 50, and then additional contributions are no longer
required.
The different scenarios show the results for different normal contribution rates. As
can be expected, higher normal contributions mean that less additional contributions
are required.
15
Projected retirement income gap
70,000
35,000
8
: 0
-35,000
-70,000
18 30 41 53 64
Time (Y ear)
Projected retirement income gap : Normal 15 percent
Projected retirement income gap : Normal 20 percent
Projected retirement income gap : Normal 25 percent
Figure 4. Projected retirement income gap with different levels of normal
contributions (contributions start at age 25, the standard of living continuation
ratio is 0.75, the contributions can be adjusted by up to 5 per cent of living
expenses per year, adjustment takes place in 1 year)
Figure 4 shows the projected retirement gap for the three normal contribution rates.
It is clear that a large gap develops initially. This gap is closed through additional
savings. The gap is close to zero until the age of 40 years. The savings then
overshoots because high investment income is attained in the later years, and the
model does not allow for a reduction in the rate of normal contributions. The extent
of overshooting is much larger for the higher contribution rate than for the lower
contribution rate.
4.2. What happ when start ing for retir t when
they are older?
The example above looks at employees starting to work and starting to make
contributions to retirement savings at the age of 25. The example below will consider
what happens if employees start to work at a higher age, or if they withdraw their
retirement savings and then have to start afresh saving for retirement when they are
older. Figure 5 shows the additional contributions required for employees starting at
16
different times when they are older, and with the normal retirement contribution set
at 20 percent in all cases.
Additional contributions as a proportion of living expenses
3
225
075
0
18 30 41 53 64
Time (Y ear)
A dditional contributions as a proportion of living expenses : Starting ions at 35
Additional contributions as a proportion of living expenses : Starting ions at 30
Additional contributions as a proportion of living expenses : Starting ions at 25
Figure 5. Additional contributions as a proportion of the living expenses with
different starting ages (level of normal contributions is 20 per cent throughout,
the standard of living continuation ratio is 0.75, the contributions can be
adjusted by up to 5 per cent of living expenses per year, adjustment takes place
in 1 year)
Figure 5 shows the importance of starting to save for retirement early. Much larger
additional contributions are required for someone starting to save later. If the
starting point of saving for retirement is delayed by only ten years from the age of 25
to the age of 35, this requires additional savings of more than 20 percent of the living
expenses to try and make up the backlog. Even with that high level of additional
expenses, the backlog is not completely reduced when the employee retires, as an
additional payment is still required right up to retirement.
This is also evident from Figure 6 below, which shows how the retirement income
gap develops over the employee’s working life. The savings plan no longer overshoots
for someone starting to work at age 35, and the additional payments are only
sufficient to keep the income gap at close to zero.
17
Projected retirement income gap
70,000
35,000
: N
; 0
-35,000
-70,000
18 30 41 53 64
Time (Y ear)
Projected retirement income gap : Starting contri at 35
Projected retirement income gap : Starting conti at 30
Projected retirement income gap : Starting contri at 25
Projected retirement income gap : Normal 20 percent
Figure 6. Projected retirement income gap with different starting ages (level of
normal contributions is 20 per cent throughout, the standard of living
continuation ratio is 0.75, the contributions can be adjusted by up to 5 per cent
of living expenses per year, adjustment takes place in 1 year)
4.3 What happ when employ respond more slowly?
The model can also be used to study different response behaviours in trying to close
the retirement income gap. As an example, Figure 7 presents the pattern of
additional contributions if the adjustment is slower. Up till now, the model assumed
that the adjustment will take place in a year. The simulation runs in Figure 7 also
show the results when the adjustment takes place over five years and over ten years.
Additional contributions as a proportion of living expenses
09
.0675
EB 045
0225
0
18 30 41 53 64
Time (Y ear)
Additional contributions as a proportion of living expenses : A djust living standard in 1 year
Additional contributions as a proportion of living expenses : Adjust living standard in 5 years
Additional contributions as a proportion of living expenses : Adjust living standard in 10 years
Figure 7. Additional contributions as a proportion of the living expenses with
different adjustment rates (contributions start at age 25, level of normal
contributions is 20 per cent throughout, the standard of living continuation
ratio is 0.75, the contributions can be adjusted by up to 5 per cent of living
expenses per year)
Figure 7 shows that the adjustment is slower, reaching a lower proportion of living
expenses. From Figure 7 it would also appear as if the total level of additional
contributions required is lower with the slower adjustment rate. This is so for the
particular set of parameters assumed here. With the slower rate of adjustment the
savings do not overshoot as much as with a quicker adjustment rate.
This is also evident from Figure 8 below, which plots the retirement income gap for
different adjustment rates. With a slower adjustment rate the projected income gap
does not close so quickly, and then does not overshoot to the same extent as
experienced with the quicker adjustment rate. The additional amount saved with the
quick adjustment rate is not lost, though. The retiree can enjoy a higher retirement
income which is created because the process overshoots the target.
Projected retirement income gap
50,000
25,000
g
: 0
-25,000
-50,000
18 30 41 53 64
Time (Y ear)
Projected retirement income gap : A djust living standard in 1 year
Projected retirement income gap : A djust living standard in 5 years
Projected retirement income gap : A djust living standard in 10 years
Figure 8. Projected retirement income gap with different adjustment rates
(contributions start at age 25, level of normal contributions is 20 per cent
throughout, the standard of living continuation ratio is 0.75, the contributions
can be adjusted by up to 5 per cent of living expenses per year)
4.4 What happ when empl start late and have high retirement
y
income expectations?
The importance of starting early to save for retirement is clear from the simulations
presented above. If an employee starts to save later in life, it becomes necessary to
make large additional savings in order to reach a retirement income goal. To
illustrate this point, the next set of simulations considers an employee starting at the
age of 45, aiming to maintain their standard of living in retirement.
The additional contributions this employee makes to retirement savings is
constrained by the amount by which the employee is prepared to adjust their
standard of living. The set of simulations consider an employee that will adjust in a
year, and can adjust by three different amounts in that year (one per cent of living
expenses, two per cent of living expenses, and five per cent of living expenses
respectively).
20
Additional contributions as a proportion of living expenses
2
15
ol
5
0
18 30 41 53 64
Time (Y ear)
Additional contributions as a proportion of living expenses : Starting at 45 increase adjust by 5 percent peryear
Additional contributions as a proportion of living expenses : Starting at 45 increase adjust by 2 per cent peryear
Additional contributions as a proportion of living expenses : Starting at 45 increase adjust by 1 percent peryear
Figure 9. Additional contributions as a proportion of the living expenses for an
employee starting contributions at the age of 45 and aiming for a standard of
living continuation ratio of 1 and with different adjustment responses (level of
normal contributions is 20 per cent thr t the at takes
place over 1 year)
Figure 9 shows that the additional contributions required in this instance ramps up
against this adjustment constraint for close to the full 20 year working period. Only
in the case of a five per cent adjustment per year do we reach a point where the
calculations do not show that an even higher level of additional contributions is
required. At that point, the additional contributions required are larger than the
amount that the employee retains for living expenses.
This result shows that people who start saving for retirement late in their careers
cannot rely on a slow or a weak response if they want to reach a realistic retirement
income target.
This effect is also evident from the projected retirement income gap as presented in
Figure 10. With a weak response (which would increase the additional contributions
by one or two per cent of living expenses per year) the income gap never closes. It is
only with the aggressive five per cent adjustment that the income gap is eliminated
21
shortly before retirement. The results show that it may have been preferable to adjust
the contributions even quicker. The five percent per year increase in contributions
with a two per cent per year salary increase means a constant three per cent per year
living standard decline over close to twenty years. A large initial adjustment effecting
a quantum jump (“biting the bullet once”) might have been preferred by an employee
instead of the slow decline in living standards modelled here.
Projected retirement income gap
200,000
150,000
g
: 99,990
49,980
-30
18 30 41 53 64
Time (Y ear)
Projected retirement income gap : Starting at 45 increase adjust by 5 percent peryear
Projected retirement income gap : Starting at 45 increase adjust by 2 percent per year
Projected retirement income gap : Starting at 45 increase adjust by 1 percent per year
Figure 10. Projected retirement income gap for an employee starting contributions at
the age of 45 and aiming for a Jard of living conti ion ratio of 1 and
with different adjustment responses (level of normal contributions is 20 per
cent throughout, the adjustment response takes place over 1 year)
5 Conclusions
This paper analysed the dynamic behaviour of the decisions of a salaried employee
saving for retirement. In a preliminary and simplified model of this decision, various
scenarios were considered and presented. There are four conclusions that can be
drawn from the simulation results and the various scenarios that were considered:
22
The results show the importance of starting the process of saving early if a
reasonable retirement goal is to be met.
The results also show that it is important to evaluate the projected retirement
income periodically and to make strong and swift adjustments to ensure a
reasonable retirement income. Any additional contribution to retirement
income has to be financed by a cut in living expenses. This will come with
some pain. But if an employee has fallen behind it becomes more and more
difficult to catch up again. It may be better to make a quantum adjustment
once than to experience a slow decline in living standards as more and more
funds are required to meet the retirement targets.
The alternative, sadly, is to be forced to take a drop in living standards during
retirement. This is a predicament into which many retirees are eventually
forced if they are not pro-active in saving for their retirement.
The model also shows how the savings plan can overshoot. This usually
happens when the realised investment income is higher than the expected
investment income on which the savings plan is based. This result was
experienced in many of the simulations present here. But the opposite is also
possible. If the expected investment income is not realised this could lead to
an unexpected shortfall. An extended model that considers investment as a
stochastic variable could help to devise an optimal response which will ensure
that sufficient funds are accumulated while limiting the extent to which the
plan overshoots with the varying investment returns.
The model used in this study is a very simplified model of the real decision, and the
research could be expanded to investigate other aspects of the system.
The first aspect already referred to is the stochastic nature of the market returns. An
interesting dynamic develops when the expected investment returns used to estimate
the retirement income gap are not realised. What was considered a surplus in one
year could, after very bad investment returns, turn into a shortfall. It is also possible
that good investment returns can turn a shortfall into a surplus. It is possible that
23
this dynamic could encourage a slower response than we have advocated here. It may
be optimal to wait longer to see how things pan out before making adjustments. To
study this would require a better understanding of the stochastic process generating
investment returns, how the decision maker could respond, and what the decision
criterion would be to select an optimal strategy.
A similar dynamic exists in the capital conversion rate that is used to convert the
retirement capital into retirement income. This conversion rate depends very
strongly on long term interest rates, especially if life annuity rates are used as the
basis for this conversion. The model can therefore be extended to include a
consideration of the stochastic nature of the conversion rate so that the influence of
this variation on the dynamic behaviour of the system can be observed.
The present model can also be extended by allowing an individual to bring their
retirement date forward or to postpone retirement. An interesting dynamic will likely
develop between the capital conversion rate and the retirement age, and an extension
of the model can be used to observe how this influences the behaviour of the rest of
the model.
Future extension of the model will include investigation of the dynamics between the
investment returns and the capital conversion rate. The latter depends on long term
interest rates, so it is really the relationship between equity returns and interest rates
that will be underlying this relationship. The system as defined at present potentially
suffers from what Chu, Strand and Fjelland (2003) refer to as radical openness. The
investment returns and the interest rates underlying the capital conversion rates
from part of the broader investment environment of the system. But these two
variables may be influencing each other in a broader system, causing a dynamic
relationship between the two. This may increase or reduce the risk of the system we
are studying.
Without any extensions to the simplified model used here, the importance of the
central message emerging from the interactions is evident. Pro-active employees
should position them for their retirement, should periodically determine their
24
retirement income gap, and should take the necessary steps to close this gap to
ensure an acceptable income in retirement.
References
10X Investments. 2013. Epic miss — Part 1: What annuity income will your savings
buy? The 10X Blog 25 September 2013. Available online at
http://www.10x.co.za/blog/epic-miss-part-1-what-annuity-income-will-your-
savings-will-buy/ Accessed 7 October 2014.
Cameron, B. 2013. An age-old problem. Personal Finance 29 January 2013.
Available online at http://www.iol.co.za/business/personal-
finance/retirement/an-age-old-problem-1.1460734#.VC3uGE2KBD8 Accessed
3 October 2014.
Chaim, R.M. 2006. Combining ALM and system dynamics in pension funds.
Proceedings of the 24th International Refereed Conference of the System
Dynamics Society, Netherlands. Available online at
http://www.systemdynamics.org/conferences/2006/proceed/index/htm
Accessed 10 February 2015
Chaim, R.M. 2007. Dynamic stochasticity in the control of liquidity in asset and
liability management (ALM) for pension funds. Proceedings of the 25th
International Refereed Conference of the System Dynamics Society, Boston.
Available online at
http://www.systemdynamics.org/conferences/2007/proceed/papers/CHAIM3,
56.pdf Accessed 10 February 2015.
Gharakhani (2009). Towards a better understanding of pension systems.
Proceedings of the 27th International Refereed Conference of the System
Dynamics Society, Albuquerque, NM, USA. Available online at
http://www.systemdynamics.org/conferences/2009/proceed/papers/P1311.pdf
Accessed 10 February 2015
Chu, D., Strand, R., Fjelland, R. 2003. Theories of Complexity. Common
Denominators of Complex Systems. Complexity, 8(3): 19-30.
De Villiers-Strydom, J. and Krige, N. 2014. Comparing South African annuity options
at retirement. Journal of Economic and Financial Sciences 7 (2) (July): 433-
450.
Finke, M., Pfau, W.D. and Blanchett, D.M. 2013. The 4 Percent Rule Is Not Safe ina
Low-Yield World. Journal of Financial Planning 26 (6): 46-55.
Forrester, J. W., 1961. Industrial Dynamics. Cambridge, MA: Productivity Press.
Leach, J. 2014. Can you afford to retire at 65? Nedgroup Investments Insights 9
January 2014. Available from
www.nedgroupinvestments.co.za/Files/Fetch/8766 Accessed 1 October 2014.
25
Maani, K. E, and Cavana, R. Y., 2007. Systems thinking & modeling: understanding
change and complexity. New Zealand: Pearson.
Pickworth, E. 2013. Most pensioners in SA face a grim future. BDLive 16 August
2013. Available online
http://www.bdlive.co.za/business/financial/2013/08/16/most-pensioners-in-
sa-face-a-grim-future Accessed 3 October 2014.
Sapiri, H., Kamil, A.A., Tahar, R.M. and Tumin, H. 2010. Dynamics simulation
approach in analyzing pension expenditure. International Journal of
Mathematical, Computational, Physical and Quantum Engineering 4 (10): 21-
27.
Sapiri, H., Kamil, A.A., Tahar, R.M. and Tumin, H. 2011. System dynamics approach
in analyzing impact of demographic and salary risks on pension expenditure.
African Journal of Business Management 5(3): 902-909.
Sapiri, H., Kamil, A.A. and Tahar, R.M. 2012. Sensitivity Analysis in Pension
Expenditure Model. Paper presented at the International Conference on
Management, Behavioral Sciences and Economics Issues (ICMBSE'2012)
Penang, Malaysia. Available online at
http://psrcentre.org/images/extraimages/212126.pdf Accessed 10 February
2015
Sexauer, S.C., Peskin, M.W. and Cassidy, D. 2012. Making Retirement Income Last a
Lifetime. Financial Analysts Journal 68 (1): 74-84
Shimada, T., Kameyama, S., Uchino, A., Machida, K. and Watanabe, M. 1990.
Simulation model of Japanese welfare annuity system. Proceedings of the 8th
International Refereed Conference of System Dynamics Society, Chestnut Hill,
USA. Available online at
http://www.systemdynamics.org/conferences/1990/proceed/pdfs/shima1038.
pdf Accessed 10 February 2015.
Sterman, J. D., 2000. Business dynamics: systems thinking and modelling for a
complex world. New York: McGraw-Hill/Irwin.
Viehweger, B. and Jagalski, T. 2003. The reformed pension system in Germany, a
system dynamics model for the next 50 years. Proceedings of the 21st
International Refereed Conference of System Dynamics Society, New York,
USA. July 20-24. Avaialble online at
http://www.systemdynamics.org/conferences/2003/proceed/PAPERS/191.pdf
Accessed 13 February 2015.
26
