by William J. Klinger
Executive Summary
- Past research on retirement income has concentrated on defining real income profiles that are flat or increasing over time. This paper defines safe profiles in which real income decreases with age, called aggressive profiles. Such profiles may appeal to retirees desiring to spend earlier in retirement rather than later.
- Two types of aggressive profiles are defined, smooth and step. In smooth profiles, real retirement income decreases a constant amount each year. Step profiles maintain a constant real income level for five-year periods before decreasing.
- Monte Carlo simulations are used to test 30-year profiles and identify those that have 99 percent, 95 percent, and 90 percent success rates.
- Analysis of the simulations that result in failure identifies the low portfolio returns and high withdrawal rates that can provide early warnings of potential future problems.
- The aggressive profile definitions are extended by setting maximum and minimum annual real income levels that serve as guardrails for annual income.
- Three common-sense rules based on previous research are used to adjust a retiree's annual real income each year in retirement in response to economic conditions. The Negative Return and Capital Preservation Rules can help rescue troubled portfolios by dictating cuts in withdrawals when conditions warrant. Either rule can be combined with the Prosperity Rule, which increases income in good economic times. Adjustments in income from these rules are kept within the guardrails of the profile.
- Safe aggressive profiles providing real income reductions of 10 percent and 20 percent over the retirement period can be created and provide more total income than simple flat income profiles without incurring additional risk.
William J. Klinger, founder of B-K-Ind LLC, is a professor at Raritan Valley Community College. He holds an MBA in finance from the University of Chicago and has an MS in computer science from the University of Wisconsin—Madison. He can be reached at klinger@b-k-ind.com
Past research in retirement withdrawals has focused on finding a
percentage of a retirement portfolio, the initial withdrawal rate,
that will, after adjusting for inflation, sustain a retiree over a
lifetime of 30 or 35 years. Work by Bengen (1994) and subsequent
studies by Cooley, Hubbard, and Walz (1999) and others have shown
that an initial withdrawal rate of approximately 4 percent of a
retirement portfolio, with subsequent increases for inflation, can
be maintained over a 30-year period. The methodology and results of
the studies vary somewhat, but the consensus is that a 4 percent
initial withdrawal rate is generally considered safe. A retirement
income stream that remains roughly constant with increases for
inflation can be called a uniform profile. A
characteristic of a uniform profile is that there is, at least
statistically, the potential for a significant balance in the
retirement portfolio at the end of the retirement. This legacy
amount is potential lost earnings for the retiree.
Researchers such as Stein and DeMuth (2005) and Spitzer (2008) have
investigated the effect of increasing the retirement portfolio
withdrawal rate over a retiree's lifetime. This creates a
progressive profile where a retiree's real income
increases with age. Their research defines progressive profiles in
which the retirement portfolio withdrawal rate is adjusted upward
to a new safe level every five years and income is adjusted for
inflation during the intervening five-year periods.
While both lines of research are beneficial, they neglect retirees
who may want to spend more early in retirement rather than later.
One reason researchers have pursued the above strategies is that
they are conservative approaches. There are no do-overs in
retirement and one should rightly be cautious. Stolz (2009) points
out that planners and retirees are especially cautious given the
current economic climate.
To address the needs of retirees who want to spend more early in
retirement, Klinger (2007) proposed a method that resulted in
annual retirement incomes which decreased over time, aggressive
profiles. That method was based upon portfolio withdrawal
rates and rules that dynamically adjust a retiree's withdrawal rate
based upon economic conditions. While achieving the desired
aggressive profile, the rules can be difficult for the average
retiree to implement. Spitzer (2008) tried a hybrid approach that
calls for a relatively high withdrawal rate for the first five
years of retirement, followed by a real reduction the next five
years, and then a progressive profile for the remainder of
retirement. The problem remains, how can a retiree safely plan to
spend relatively more early in retirement and then less later? Is
there an easy-to-understand plan to achieve an aggressive income
profile? How much more can a retiree spend early and with what
risks? These are the questions addressed in this paper.
Definitions, Methodology, and Assumptions
The research presented in this paper defines simple, aggressive
profiles that, when simulated, result in sustainable withdrawals.
The research then defines rules a retiree can use to dynamically
adjust the retirement income and improve either the confidence the
assets will not run out or further increase the retirement income.
The end result is straightforward criteria for producing safe,
aggressive retirement income profiles.
An aggressive profile is defined as the series of annual retirement
incomes where real (adjusted for inflation) income decreases over
time. A smooth aggressive profile is an aggressive profile
in which real retirement income decreases the same amount each
year. With a step aggressive profile, real income remains constant
for a specified number of years before dropping in a series of
steps throughout retirement. This paper uses the term "income" to
mean the amount that is withdrawn from a retirement portfolio in
any year. Tax implications are ignored in this research and,
depending upon the investment specifics, the amount withdrawn may
be subject to tax and not completely available for spending.
Withdrawal rates, a percentage of the retirement portfolio, are
useful in creating uniform and progressive profiles but can be
restrictive and confusing in defining aggressive profiles. The
research presented here uses explicit real income with smooth and
step aggressive portfolios defined by specifying the real income
target for each year in retirement. Because this research uses a $1
million initial retirement portfolio, the initial withdrawal
amounts can easily be converted to withdrawal rate percentages for
comparison with other research. After the first year of retirement,
however, the withdrawal amount is not based upon a percentage of
assets in that year but is specified in real income terms relative
to the initial withdrawal amount.
A Monte Carlo simulator (Ibbotson 2007) written in C++ was used to
test profiles and their sustainability through retirement. The
initial retirement portfolio is $1 million. The simulator assumes
that a retiree's income needs for a year are withdrawn from the
retirement portfolio on the first day of each simulated year. At
the end of each simulated year, the simulator calculates the return
on the portfolio's investments and the rate of inflation using a
random draw from a normal distribution based upon historical
returns (Ibbotson 2007). The correlation between stocks and bonds
is addressed by adjusting the values from the random draw by
multiplying them against a Cholesky Matrix (Gentle 2003) created
from the historical correlations. Real returns are calculated by
taking the simulated returns in each year and discounting them by
the rate of inflation. The portfolio is rebalanced to the target
asset allocation at the end of each simulated year. The simulation
does not take into account transaction costs or taxes. The
simulator repeats these calculations for every year of the
retirement period.
A simulation run is considered a success if there is a positive
portfolio balance at the end of the retirement period. Each
retirement profile is simulated 1,000 times and the percent of
simulations that result in a positive portfolio balance, the
success rate, is calculated. Portfolios that achieve 99 percent, 95
percent, and 90 percent success rates are used in the
analysis.
Key assumptions in the simulator are the asset allocation and
expected returns. While other researchers, such as Spitzer,
Strieter, and Singh (2007), have analyzed the effect of asset
allocation on portfolio sustainability in retirement, this research
focuses on the income profile using the relatively common 60/40
asset allocation. The asset allocation is 60 percent large company
stocks and 40 percent long-term corporate bonds. The simulator uses
data from 1926–2007 (Ibbotson 2007) and assumes that the
average annual return for large company stocks is 12.3 percent with
a standard deviation of 20.0, the average annual return for
long-term corporate bonds is 6.2 percent with a standard deviation
of 8.4, and average annual inflation is 3.1 percent. As an aside,
while these averages clearly do not resemble recent economic
conditions, one of the benefits of using Monte Carlo simulations is
that unlikely scenarios, not only historical results, are included
in the analysis.
The research establishes both smooth and step aggressive profiles
that result in 99 percent, 95 percent, and 90 percent success
rates. To find such profiles, first a single profile is created
with the desired percentage income reduction. For example, a real
initial income level of $50,000 can be defined with a desired
percentage reduction of 10 percent. The final income target would
be $45,000 ($50,000 × 0.9). The real drop in income is $5,000
($50,000 – $45,000). The annual real decline in income is
the total drop divided by the number of years in retirement minus
one (there is no reduction for the first year of retirement), which
in this case rounds to $172 ($5,000 ÷ 29). Taking that starting
profile, the income level in every year is increased and decreased
in 1 percent increments to create other profiles, which are tested
to find the maximum income level that meets the success rate
criteria. The profiles meeting the criteria are compared to uniform
profiles with the same success rates. In the first phase of the
research, the annual income defined by a profile is not allowed to
vary. In the second phase, the income level is allowed to vary
within limits according to predefined rules, and the results are
compared to the simple profiles.
Aggressive Retirement Profiles
To create a baseline for comparison, three uniform profiles are
identified. These profiles have the characteristics that real
income does not change throughout retirement, and they have 99
percent, 95 percent, and 90 percent success rates. The income
levels for uniform profiles meeting these requirements are $28,657,
$35,822, and $42,031, respectively, for 30-year periods. These
levels are consistent, if not conservative, compared to past
research, which has suggested a safe initial withdrawal rate is
approximately 4 percent. The initial withdrawal rates for uniform
profiles in this research are 2.9 percent, 3.6 percent, and 4.2
percent for success rates of 99 percent, 95 percent, and 90
percent, respectively. Some variation in results is to be expected
as the assumptions in asset allocation, asset returns, allowed
variations in annual income, and the methodology used vary across
different studies.
Next, smooth aggressive profiles are defined and tested. The
characteristic of these profiles is that the real income drop is 10
percent from the first year of retirement to the last year of
retirement. The 10 percent drop was chosen as a conservative,
realistic change in income that one might consider. The 30-year
smooth aggressive profiles that have success rates of 99 percent,
95 percent, and 90 percent and the uniform profiles with the same
success rates are shown in Figure 1.
The initial income levels for the 99 percent, 95 percent, and 90
percent smooth aggressive profiles are $30,317, $37,576, and
$43,981, respectively. Aggressive profiles begin with 4.6 percent
to 5.8 percent higher incomes than uniform profiles with the same
success rate, providing retirees with the additional income they
desire early in retirement. A critical requirement in implementing
a smooth aggressive strategy is taking real income cuts each year.
In the aggressive profiles in Figure 1, the annual real income
reductions range from $461 for a 99 percent success rate to $650
for a 90 percent success rate.
An alternative to annual decreases in real income is a step
aggressive profile. This research uses step profiles that hold real
income constant for five years followed by a real decrease in
income. The step profiles that resulted in simulations meeting the
success rate criteria and the smooth aggressive profiles meeting
the same success rate criteria are shown in 
.
The initial incomes for the step profiles for 99 percent, 95
percent, and 90 percent success rates are $30,317, $37,576, and
$43,981, respectively, and identical to the corresponding rates for
a smooth profile. In the step profiles, the real drop in income
every five years ranges from $606 for the 99 percent success rate
to $880 for the 90 percent success rate case. The final incomes are
4.8 percent to 5.8 percent below the corresponding uniform profile
final incomes.
Two other useful measures in comparing profiles are the real total
income during retirement and the assets remaining at the end of the
retirement period, the legacy. Table 1 compares the uniform,
smooth, and step profiles using these measures rounded to the
nearest thousand dollars. The values shown are the median real
values from all the simulation runs, including simulation failures,
with total income including a final payment at the start of the
31st year. The aggressive profiles have similar total incomes and
legacy amounts to the uniform profiles and in no case is there more
than a 2.4 percent difference. The healthy legacy amounts shown in
Table 1 might lead one to ask whether there are strategies that
could further increase the income in retirement at the expense of a
smaller legacy. Before looking at increasing the income and
potentially increasing risk, it is important to understand what
causes profiles to fail.
Failure Analysis
While it is interesting to look at the successes when doing
research on retirement income, it is arguably more important to
look at the conditions under which particular strategies fail. It
is certainly important when one implements a strategy to look for
indications of potential future problems. To get a deeper look at
potential failure scenarios, the number of retirement simulations
is increased to 15,000. Increasing the number of simulations
increases the absolute number of failure scenarios.
In the case of the 30-year smooth aggressive profile with the 90
percent success rate, out of 15,000 simulations, the earliest
failure occurs eight years into retirement. In that earliest
failure case, the portfolio return in the first three years is
–30.8 percent, –41.0 percent, and –27.1
percent. This causes the withdrawal rate from the retirement
portfolio to go from 4.39 percent to 12.5 percent with the real
portfolio value dropping to $299,000. When looking at all failures,
the average failure occurs 23 years into retirement. The average
portfolio return in the first three years is 1.5 percent, 2.1
percent, and 2.5 percent. On average, this causes withdrawal rates
to go from 4.3 percent to 5.0 percent and exceed 6 percent in six
years. After three years, the average real portfolio value is
$845,000 and after six years, it is only $693,000.
The 30-year step aggressive portfolio with a 90 percent success
rate has similar failure statistics. The earliest failure scenario
is also caused by poor early returns and occurs eight years into
retirement. The returns for the first three years are
–14.6 percent, –22.5 percent, and –7.3
percent. The withdrawal rate, which starts at 4.39 percent, goes to
7.6 percent in three years, and by year six leaps to 11.1 percent,
with a portfolio value of only $335,000. The average failure occurs
23 years into retirement. The average portfolio returns of the
failures in the first three years are 0.8 percent, 0.7 percent, and
3.4 percent, causing withdrawal rates to go from 4.3 percent to 5.1
percent. By year six, the average withdrawal rate is 6.3 percent
with a portfolio value that has dropped in real terms to
$681,000.
One thing to observe from this analysis is that, although the
earliest failure comes eight years into retirement, there are clear
early warning signs. As one might expect, negative portfolio
returns in the early years of retirement are clearly an indication
of potential trouble. Problems can also be seen in the rising
withdrawal rates. No matter what the research, a prudent retiree
will not keep retirement on autopilot when faced with such
conditions, but will cut back on expenditures. Fortunately, it is
possible to simulate the effects of such actions.
Dynamic Aggressive Profiles
Several researchers have looked at the effect of implementing
dynamic rules that adjust portfolio withdrawals in response to
economic conditions. Clyatt (2005) modeled a common sense rule
which states that if a portfolio loses value in any year, income
for the next year is cut to 95 percent of the prior year's level.
Income in subsequent years is increased only to keep pace with
inflation. Stein and DeMuth (2005) specify that if the market falls
10 percent any year in the first 10 years, the portfolio withdrawal
rate should be set to 4 percent. Guyton and Klinger (2006)
recognized that market returns may be bad in a particular year but
it might not be necessary to cut income if the portfolio were still
large. They proposed a rule triggered by a rise in withdrawal
rates. When withdrawal rates exceed a threshold, the withdrawal
rate is cut a specified amount. Their work also analyzed conditions
in which the markets were favorable and proposed a rule to give
retirees a raise when withdrawal rates become low.
These ideas can be combined with the simple aggressive profiles to
either improve the success rate for a particular profile or,
keeping the same success rate, increase the income profile. The
research here takes the approach that the income profile will be
modified to allow for a limited variation in income each year.
Smooth and step aggressive profiles are defined and tested as above
with the enhancement that, instead of there being a fixed income
amount for each retirement year, there is a maximum and minimum
income level for each year. The real income in any simulated year
is allowed to fluctuate according to the dictates of dynamic rules;
however the real income is not allowed to fall outside the bounds
of the maximum and minimum levels in any year. In effect, the
specified maximums and minimums will serve as guardrails
for the actual real income profile throughout retirement. These
types of profiles are called guardrail profiles.
This research uses the initial income level as the initial maximum
income level. The minimum income level will be 10 percent below the
maximum level in each simulated year. As in the case of no
guardrails, the maximum and minimum income levels are defined in
real income amounts, not as withdrawal rates, and are determined at
the time of retirement, not dynamically set. This produces the
aggressive income profiles desired. There is nothing special about
10 percent; it was chosen as an amount retirees might reasonably
tolerate in income fluctuation. Figure 3 shows examples of smooth
and step guardrail profiles. A simple smooth or step profile is
just the special case of setting the allowed variation to be 0
percent.
Three rules based upon past research and designed to be triggered
by the kinds of events observed in the failure analysis are defined
and tested. The first rule, the Negative Return Rule, based upon
the work by Clyatt (2005), states that if the portfolio has a
negative nominal return in any year, real retirement income is
reduced by 10 percent. The second rule is a modified version of the
Capital Preservation Rule from Guyton and Klinger (2006), which, as
used here, dictates that if the withdrawal rate in any year exceeds
6 percent, real income in the following year is reduced by 10
percent. The third rule is a modified version of the Guyton and
Klinger Prosperity Rule, which, as used here, states that if the
withdrawal rate in any year falls below 3.8 percent, the retiree
will take a 10 percent raise in income. The 3.8 percent threshold
for the Prosperity Rule was selected as a value conservatively
under the safe 4 percent rule of thumb. The objective of these
three rules is to mimic what a retiree might reasonably do when
faced with a shrinking portfolio and the types of conditions found
in the failure analysis or in good economic times. In this
research, either the Negative Return Rule or the Capital
Preservation Rule, but not both, will be applied to a given
retirement strategy. The maximum and minimum limits of the
guardrail profile may not be exceeded when applying the dynamic
rules. Conceptually, the actual retirement income profile a retiree
will see when adopting this profile with these rules will be a
zigzag path through the guardrail profile.
Using these rules, a retiree can achieve either an increase in the
success rate for a given profile or, keeping the success rate
constant, increase the retirement income profile. The initial
income level for a smooth profile without dynamic rules and a 90
percent success rate is $43,981, a 4.4 percent initial withdrawal
rate. Using the smooth profile shown in Figure 3 with the 90
percent success rate and with guardrails and the Negative Return
and Prosperity Rules, the same initial income level of $43,891 now
has a 95 percent success rate. The real annual income in this case
has a standard deviation of $2,694. The original 95 percent success
rate scenario improves to a 98 percent success rate with a standard
deviation of $2,106 in real annual income. Using guardrails with
the Negative Return and Prosperity Rules allows for an increase in
income while halving the number of failures. Applying the Capital
Preservation and Prosperity Rules, the 90 percent success rate
profile now has a 92 percent success rate with a real annual
standard deviation of $1,446 and the original 95 percent success
scenario improves to 96 percent with a standard deviation of $947.
In both cases, the increase in the success rate comes at the cost
of fluctuating income. This is because using guardrails and
decision rules allows income to fall, if needed, to rescue the
portfolio.
Alternatively, if one is comfortable with a particular success
rate, using guardrails and dynamic rules can increase the income
profile. Table 2 shows the effects of using the above dynamic rules
on the simple 90 percent smooth aggressive profile. Using the
Negative Return and Prosperity Rules, one can achieve a 17 percent
increase in the initial income and a 9 percent increase in total
income over not applying any rules. These increases come at the
expense of a standard deviation of real annual income of $3,261 and
a 13 percent reduction in the legacy. The Capital Preservation and
Prosperity Rules result in an 8 percent increase in the initial
income and total income, less than is achieved by using the
Negative Return and Prosperity Rules but with a standard deviation
approximately half thereof. Again, increases in income come at the
cost of variability in annual income. In addition, we see that one
can increase income early in retirement at the cost of a smaller
legacy. It is interesting to note that using guardrails, the 90
percent success rate scenario ends with a real income
greater than what one would start with using the simple 4
percent rule of thumb initial income and increasing it by
inflation.
Although the profiles are designed based upon a 10 percent drop in
real income over retirement, Table 2 shows final income levels that
are more than 10 percent less than the original income level when
guardrail rules are used. The drops are 18 percent–19
percent. This is because the guardrails permit an additional 10
percent fluctuation in income each year to protect the retirement
portfolio assets. In the 90 percent success rate case, these rules
clearly kick in and the final income is at the lower end of the
guardrail range. Simulations using a 95 percent success rate, with
a lower initial income level of $44,408, result in a final income
level of $39,860, or a 10.2 percent drop. There is a direct
relationship between taking greater risk and having a lower final
income level. This relationship holds for all guardrail
profiles.
The same approach is applied to a step profile. Allowing a minimum
income guardrail level of 10 percent below the step profile with a
90 percent success rate and applying the Negative Return and
Prosperity Rules results in an increase of the success rate to 95
percent while introducing a standard deviation of real annual
income of $2,682. Using these rules, the 95 percent success rate
scenario improves to 98 percent with a standard deviation of annual
income of $2,100. Applying the Capital Preservation and Prosperity
Rules increases the success rate of the 90 percent scenario to 92
percent with a standard deviation of real annual income of $1,483
while the 95 percent scenario improves to 96 percent with a
standard deviation of real annual income of $943.
Using the same decision rules and success criteria as in the
previous scenario, the entire income profile can be shifted upward.
The results of these actions are summarized in Table 3. Applying
the Negative Return and Prosperity Rules increases the initial
income by over 16 percent and the median total income 9
percent.
As with the simple smooth and step profiles, it is important to
analyze the failures when guardrail profiles and dynamic rules are
introduced. The smooth guardrail profile with a 90 percent success
rate applying the Negative Return and Prosperity Rules begins with
an initial income level of $51,240. The earliest failure scenario
out of the 15,000 simulations is the same as when no rules are
applied, and the earliest failure still occurs eight years into
retirement. When looking at all failure scenarios, the average
failure now occurs 24 years into retirement, a year later. The
average returns over the first three years are 1.3 percent, 1.9
percent, and 2.3 percent, causing the portfolio withdrawal rate to
go to 5.0 percent and then over 6 percent in six years.
The step profile with a 90 percent success rate using the Negative
Return and Prosperity Rules has an initial income of $47,397. Using
these rules, the earliest failure now occurs 11 years into
retirement and is caused by the same early failure scenario as in
the smooth portfolio case. The average failure occurs 24 years into
retirement with returns of –0.5 percent, –0.9
percent, and 1.1 percent in the first three years. The withdrawal
rate is 4.5 percent after three years and reaches 6 percent in
seven years.
The use of guardrail profiles and dynamic rules in these two cases
does not appear to significantly alter the circumstances that cause
a profile to fail. This is to be expected, because we did not
increase the success rate. However, for the same risk, a retiree
may receive greater income by using dynamic rules.
It is reasonable to ask whether the results given here are
scalable. In other words, will a portfolio of $2 million and a
profile twice the aggressive profiles defined above produce
proportionate results? The answer is yes. Starting with a $2
million portfolio and using the 30-year smooth profile with 10
percent guardrails, applying the Negative Return and Prosperity
Rules gives results which are twice that for a $1 million starting
portfolio. For a 90 percent success rate criterion, the maximum
initial income level is $102,480, the standard deviation of real
annual income is $6,523, the total real income is $2,815,000 and
the real legacy is $3,130,000. To apply this research to portfolios
of sizes other than $1 million, the numbers can be scaled
proportionately.
Other Aggressive Profiles
Figure 4 shows smooth guardrail profiles with incomes that decline 10 percent and 20 percent, over a 30-year retirement period with a 90 percent success rate. The income in each year is subject to the Negative Return and Prosperity Rules. The total real income produced by both profiles is approximately $1.4 million and both profiles leave legacies of $1.5 million. The standard deviations are $3,261 for the 10 percent drop and $3,174 for the 20 percent drop. There is little real difference between the profiles other than the annual income stream they produce. Of course, a retiree may have a preference due to needs, willingness to make income reductions, and perceived risk.
A range of dynamic step profiles for the 90 percent success
criterion can be created in the same manner as the smooth profiles.
The profiles corresponding to 10 percent and 20 percent real income
drops over retirement have initial incomes of $51,240 and $53,375.
The total income for both profiles is approximately $1.4 million
with real annual income standard deviations of $3,242 for the 10
percent drop profile and $3,135 for the 20 percent drop profile.
The legacy amounts are both $1.5 million.
Summary and Conclusions
If a retiree wants a higher level of real income early in
retirement rather than later, it is important to create a plan that
guides the retiree with a safe strategy for withdrawals from a
retirement portfolio that starts at a level higher than a uniform
strategy would specify, and decreases over time.
A smooth aggressive profile with a 10 percent drop in income over a
30-year retirement period that starts with an initial withdrawal
amount of $30,317 has a 99 percent success rate in the simulations
and represents an initial income 6 percent higher than a uniform
profile with the same success rate. At a 90 percent success rate,
the initial income grows to $43,981. Retirees desiring somewhat
more consistency could select a step aggressive profile in which
the real income remains constant over five-year periods before
dropping to a new, lower level. To get a high success rate, these
profiles need a reasonable asset cushion to fall back on in
difficult market conditions. This means that there is a sizeable
legacy left at the end of retirement that represents lost potential
income.
Analysis of the simulation scenarios in which a retiree runs out of
assets before the end of retirement shows that negative or very low
portfolio returns early in retirement can be an indicator of
potential future problems. This reinforces what retirees
instinctively know. Another indicator of future problems is that
the withdrawal rate from the portfolio increases, sometimes
dramatically. These early warning signs can be used to improve
either the success rate of a profile or increase the income at the
same success rate.
To protect portfolios or potentially increase the income levels,
the retirement profile is changed to be a band within which the
actual annual retirement income is allowed to vary. The width of
the band, that is, the difference between the maximum and minimum
income in a year, can be set to accommodate the willingness of the
retiree to have the income vary. The research in this paper uses a
10 percent difference. Such a profile can be considered a guardrail
profile.
Rules are defined to dynamically adjust the annual income according
to economic conditions but stay within the guardrail profile. The
Negative Return and Capital Preservation Rules can help rescue
troubled portfolios by dictating cuts in withdrawals when
conditions warrant. The Negative Return Rule produces marginally
better results at the expense of a higher variability in annual
income. Either rule can be combined with the Prosperity Rule, which
increases income in good economic times. The result is that the
actual income a retiree will experience will fluctuate with
economic conditions but always remain within the profile's
guardrails.
Tables 4 and 5 summarize these results and can be used as the basis
for defining a retiree's retirement profile, incorporating their
preferences for income and risk. For simplicity and because the
Negative Return and Capital Preservation Rules produce similar
results, only the Negative Return Rule results are shown. The
Capital Preservation Rule produces approximately 10 percent less
total income but with half the standard deviation in annual income.
The choice of one rule over another is a matter of preference in
the total income/risk trade-off.
The specification of a profile and an estimation of the results
follow these steps:
- Determine whether a smooth or step profile is desired. For a smooth profile, use Table 4; for a step profile, use Table 5.
- In the appropriate table, look up the Profile Specification for the desired decrease and success rate. For example, if a step profile with a decrease of 20 percent over retirement is desired with a 90 percent success rate, look at Table 5 under the column 20 percent Income Reduction and 90 percent Success Rate. The two rows below specify the step aggressive profile meeting those criteria. In this case, the initial income level is $53,375. The real level of income will be constant for five-year periods, after which a $2,135 decrease in real income is made. This specifies the maximum values for the annual income and gives a profile which looks like those in Figure 2.
- The minimum annual real income limits are set by taking the maximum income levels defined in Step 2 and subtracting 10 percent. The result is a guardrail profile similar to those in Figure 3.
- During retirement, a retiree's income will be bounded by the guardrail profile defined in Steps 1–3. As economic conditions warrant, the Negative Return and Prosperity Rules will trigger an income adjustment of –10 percent in the case of negative portfolio returns or +10 percent if the annual portfolio withdrawal rate falls below 3.8 percent. In no case should the annual real income fall outside the guardrail profile.
- The results of using the strategy defined in Steps 1–4 are shown in the last four rows of Tables 4 and 5. In the example case, simulations show that real annual income has a standard deviation of $3,135, the median final income in retirement is $38,430, median total real income is $1,383,000, and the median legacy is $1,555,000. All figures are in real dollars.
- For portfolios other than $1 million, the numbers in the tables can be scaled proportionately. For example, using a step aggressive profile and starting with a portfolio of $1.5 million, if a 10 percent decrease with a 95 percent success rate is desired, the initial income would be 1.5 times $44,408 (from the table), which is equal to $66,612.
Using an aggressive guardrail profile and dynamic rules, a
retiree can live a retirement that matches his or her desired
lifestyle without incurring excessive risk. An important result of
this research is that, by applying the dynamic rules during
retirement, an individual can not only start with a relatively high
level of income but even after 30 years of decreasing real income,
the median final real income is comparable to what the individual
would have had using only a uniform profile. As a result, the total
real income over retirement is also greater without incurring
additional risk.
Hopefully, this research will serve as a catalyst for other
research in tailoring withdrawals and investment directions to
achieve desired retirement profiles. For example, decision rules
could be defined to change asset allocations during retirement
depending upon certain conditions. As the level of sophistication
increases, one could investigate dynamic decision rules that change
over the retirement period. Any future research, however, should
start with retirees' objectives and then define decision rules, not
the other way around.
References
Bengen, William P. 1994. "Determining Withdrawal Rates Using
Historical Data." Journal of Financial Planning
(January): 14–24.
Clyatt, Bob. 2005. Work Less, Live More, The New Way to Retire
Early. Berkeley, California: Nolo Press.
Cooley, P. L., C. M. Hubbard, and D. T. Walz. 1999. "Sustainable
Withdrawal Rates from Your Retirement Portfolio." Financial
Counseling and Planning 10: 39–47.
Gentle, James E. 2003. Random Number Generation and Monte Carlo
Methods. New York: Springer-Verlag.
Guyton, J.T., and W. J. Klinger. 2006. "Decision Rules and Maximum
Initial Withdrawal Rates." Journal of Financial Planning
(March): 48–58.
Ibbotson & Associates. 2007. Stocks, Bonds, Bills, and
Inflation 2007 Yearbook. Chicago: Morningstar Inc. 31.
Klinger, William J. 2007. "Using Decision Rules to Create
Retirement Withdrawal Profiles." Journal of Financial
Planning (August): 60–67.
Spitzer, John J. 2008. "Retirement Withdrawals: An Analysis of the
Benefits of Periodic 'Midcourse' Adjustments." Financial
Services Review 17: 17–29.
Spitzer, J.J., J.C. Strieter, and S. Singh. 2007. "Guidelines for
Withdrawal Rates and Portfolio Safety During Retirement."
Journal of Financial Planning (October):
52–59.
Stein, Ben, and Phil DeMuth. 2005. Yes You Can Still Retire
Comfortably! Carlsbad, California: New Beginnings Press.
Stolz, Richard F. 2009. "Withdrawal Rules of Thumb Seductively
Simple, Yet with Limitations." Journal of Financial
Planning (March).
