New Features Wanted
Questions About Results
Questions on JFP paper
Q: How are the
red, yellow, and green dots in the analyses set?
A: The "traffic lights" are set according to the success rate of the
particular scenario simulated. If the success rate is greater
than 95%, the light is green. If the success rate is between 85%
and 94%, the light is yellow. Success rates below 85% are red.
Q: When I vary
an expense reduction I'm not seeing any change in projected values of income or
assets. Is this correct? Where should I look to see the effect of an expense
A: The Retirement Quant only worries about expenses prior to retirement.
That's because it is calculating how much you save towards retirement.
Once you are in retirement, the program is only concerned about your desired
retirement income level. That's because there are so many different ways
you can withdraw money from your accounts that have different tax consequences.
The assumption is
that you know what your retirement income target is and that it includes taxes.
See the Notes.
Q: How would I
model TIPS (Treasury Inflation Protected Securities)?
A: With Release 6, TIPS are now a pre-defined asset class. You do not
need to do anything special.
Q: Do I enter
an annuity in future dollars or today's dollars?
A: Enter all dollar amounts in today's dollars.
Q: I'm using
decision rules and am uncomfortable with the freezes and cuts the Retirement
Quant says I may get. What can I do?
Being comfortable with retirement is
important. Here is a suggested approach. Turn off all the rules and look for a level that gives you
99% or 100% success.
If you can live with that, you should be great. Levels above that amount are
possible but come with the risk you will run out of money. Decision rules can
reduce that risk but come with risks of their own, namely variable income over
your lifetime. The good news is, you will probably withdraw more in total but
you may have to take a cut(s) at some point. It is one of those "no free lunch"
deals. If you turn on decision rules, turn on the Modified Withdrawal
Rule first and see if that gets you to the level you want. If you are still
short, then add the CPR and PR. Play with their parameters to adjust the risk
level you can live with.
Q: Is there a difference between
"income" and "PV of final purchasing power"?
A: The income in constant dollars at the end of of a lifetime is the same as the
PV (present value) of final purchasing power.
Q: Are there strategies that can mitigate the risk of
early portfolio failures?
A: Try using decision rules.
Something like the Capital Preservation Rule can kick in and ‘rescue’ the
portfolio if things go south early in retirement. One thing a person can also
try is to start with a lower initial withdrawal rate and couple that with the
Prosperity Rule to give raises when the portfolio is really ‘safe’ and gradually
increase real withdrawals.
Q: Why do some retirement strategies
generate a retirement income profile that looks like a wave?
Here’s what I
think happens. In retirement strategies that allow for real raises, I can
believe that in early years the real income increases because on average the
markets do well. Not all scenarios will give raises, clearly, but on average I
can see it. These raises then continue until they become too large (again on
average) and need to be corrected back down. The degree of oscillation depends
upon how good the timing is of the corrections. Poor corrections will allow
Q: How exactly are taxes handled?
A: Before retirement, taxes are taken out of your income and the remainder is
applied to the expenses you have specified. This is because before
retirement, the simulator is calculating your asset growth and hopefully you are
saving something each year and adding it to your assets. The situation
changes when you reach retirement. Then the assumption is that you will
probably start to draw down your assets. Now asset withdrawals are
complicated when you consider tax-advantaged and taxable investments, different
asset allocations in each account, pre and post-tax contributions, minimum
required distributions, etc. The approach taken here is that in
retirement, you want to maintain a level of income. Yes, your net income
will certainly be different based upon how you've invested but that is for you
to work with your financial planner. To simulate all of that here would
too much; it would provide maybe more precision but not likely more accuracy.
Q: Are Monte Carlo simulations
really any good?
A: There are a number of ways to model the future. The simplest is to just
take some average returns and inflation rate and do arithmetic out into the
future. If in 1992, companies had forecasted the number of floppy disks
that way, we should be buried in them by now. Another approach is to use
actual history and look at different historical periods and see how a portfolio
would have done then. That's a reasonable approach but as we all know,
history is not a guarantee of the future. Read the book "Fooled by
Randomness." Monte Carlo simulators simulate thousands of possible
futures. Are they perfect? Certainly not. Not all simulated futures
are equally likely, for example. The approach one should take is that this
is one tool to use in evaluating your financial future. Do not take all
numbers as gospel but get an idea of the effect of different strategies on your
Q: Where do the income annuity
default rates come from?
A: They are quotes from
www.vanguard.com for $100,000 annuities
for a males with a spouse and 100% survivorship for the spouse. The rates were
retrieved in July 2007. Rates have quite likely changed and you should
obtain current quotes which match your situation and insert those values.
Q: Why does the constant, and
sometimes even the future, dollar income fall when using the strategy that
resets the withdrawal rate every five years to the safe (99%) rate?
A: It is because
the increase in safe withdrawal rates between five year periods does not
necessarily keep up with inflation. For example, if you retire at age 65
and have a life expectancy of 99, then you have 40 years remaining in retirement
and will use the 4.7% as your withdrawal rate. At age 70, you have 35
years remaining and will set your withdrawal rate to 5.0%. 5.0/4.7 =
1.064, which means you would get roughly a 6% raise over your original
withdrawal rate at age 65. But in the intervening 5 years you got raises
for inflation. If inflation averaged only 3%, you would have seen more
than a 15% raise in your nominal income, But when the withdrawal
rate got reset with a 6% increase, you could very well see a decline. It
all depends upon how well your portfolio did relative to inflation.
Q: The paper in the Journal of
Financial Planning talks about values for Fall and Exceeds but they don't appear
in the Retirement Quant decision rules. What happened?
A: The original Fall and Exceeds inputs were defined in terms of a percentage
change from the original withdrawal percent. This was, in retrospect, a
sort of clumsy way to express them. Retirement Quant uses specific
percentages to determine when to give a Raise or Cut. For example, if one
wants to start with an initial withdrawal rate of 5% and would have defined
Exceeds as 20%, that would mean that if the withdrawal rate reached 6%, a cut
would be made. Rather than put that Exceeds value of 20% into the
simulations, it is more obvious and probably easier for a person to implement,
if the actual 6% was put in as the limit.
Retirement Quant do any market timing for withdrawals?
A: It does allow for two different withdrawal
strategies based upon asset class returns. One is the portfolio
management rule (PMR) which determines the source(s) of each year's
- Following years where an asset class has
a positive return that produced a weighting exceeding its target
allocation, the excess allocation is sold and the proceeds invested in
cash to meet future withdrawal requirements.
- Portfolio withdrawals are funded in the
following order: (1) overweighting in equity asset classes from the
prior year-end, (2) overweighting in fixed income from the prior
year-end, (3) cash, (4) withdrawals from remaining fixed-income assets,
(5) withdrawals from remaining equity assets in order of the prior
- No withdrawals are taken from any equity
following a year with a negative return if cash or fixed-income assets
are sufficient to fund the required withdrawal.
The other is called the Fastest Horse Rule. It
basically says one should withdraw first from the asset classes which have the
lowest historical returns. I.e. keep your money on the fastest horse. The PMR
was proposed by Guyton in his original paper. The Fastest Horse Rule was made up because
there are other withdrawal strategies that might help and this gives one
comparison. We are not believers in market timing. Check how the PMR and FHR work for
Q: On the "Retirement Income
Analysis" page, something is wrong with PV of final retirement income. The
value drops to $XXX at some income level and stays there.
A: This is working correctly. Notice
that the probability of success has dropped below 95%. Therefore, the only
income you can have with 95% certainty is your retirement income, such as a
pension. Your assets have run out at the 95% confidence level.
"Variance Limit" also apply only to investment assets and not account for an annuity
in the percentage calculation?
A: The Variance Limit is compared to whatever you expressed as your retirement income
need and therefore includes an annuity.
Q: Is there a way to mimic TIPS? And why aren't they included anyway?
A: TIPS are now included with Release 6.
New Feature Requests
Q: Why doesn't
Retirement Quant do XXX?
The answer depends upon the specifics of XXX. The answer is one of: (1) it
requires access to data to which we do not have the rights and some of the
owners want a whole lot of money for it, (2) it is really difficult and
not likely to be done, (3) we would like to do it but have not gotten to it yet.
If you have ideas,
them to us.
Q: Why doesn't
Retirement Quant have XXX asset class?
A: See answer to above question. In addition
to getting the rights to the data, the data needs to have a reasonable history.
For example, it is not reasonable to project 30 or 40 years in the future based
upon 4 years of data. Revisit what happened to Long Term Capital
Management. However, if you know a ticker symbol for an index or other
asset that represents what you want, starting with Release 6 you can include
it yourself. Create your own asset classes.
Questions About Results
Q: When I run
Retirement Quant, I get different results than what I get from running the
calculator at Vanguard, Fidelity, etc.. Why?
A: Retirement analysis is driven by the
assumptions going into the program The assumptions for the returns of
different asset classes and especially their standard deviations will have a big
effect on the results.. Some programs, for example, assume that inflation
will increase at a constant rate such as 3% (i.e. the standard deviation is 0).
Requirement Quant models a fluctuating inflation rate. If you can match
all the assumptions, the results should be similar but may vary due to other
factors. In Retirement Quant, go to the Assumptions dialog and change the
assumptions to match those in the other simulator and re-run the analyses.
One should not get caught up in the difference between say 96% and 98%.
What is most important is which strategy gives a higher value and why. It is good
practice to try to validate your analyses with other tools.
Q: I made a run using 100%
large cap value stocks, no bonds or cash and didn't turn on decision rules, no
pension or annuity. This simulation had better performance than a diversified
portfolio of multiple asset classes, with better success rates etc..
A: Try doing this:
first test using just the S&P500 and then perform analysis using the Asset
Allocation function (1)
Using the S&P500 only, if you look at equity/bond trade-offs, there is no
improvement in the success rate over about 50/50 but the earliest failure does
get earlier at the % equities goes up. That is another indicator of risk
and shows that the risk does increase as you increase the % equities. If
you do a variety of equity/bond trade-offs (use the radio button), you’ll see
that 60/30/10 equities/bond/cash has the highest success rate (with 70/20/10)
with the same behavior that the earliest failure gets earlier the higher the
percent equities. So, with S&P500 only equities, you can reduce the risk
by diversifying and adding cash to the mix makes a noticeable difference.
All this is what one might expect. (2) Using 8 asset classes and using
only large cap value and bond trade-offs, we get the same behavior as with the
S&P500. Adding more asset classes and comparing them to 100% large cap
value or 100% REITS shows that the 100% in those two asset classes gives better
results than wide diversification. Why? (3) Compare using 100% large cap
value with 50% large cap value and 50% REITS. Adding REITS does improve
the success rate. Adding REITS and International Equities also increases the
success rate and lowers the average return. Why? (4) Go into the Assumptions dialog and go to the bottom and click
on “Treat asset returns independently” . This should show us how ‘pure’
diversification might work. Go back to Asset Allocation and run the tests. Now
you will see that adding more asset classes makes a noticeable difference in the
success rate. Uncorrelated diversification really helps.
the equity asset classes are correlated.
Adding uncorrelated assets is what reduces risk and
that's what many hedge funds are after. Yes, the large cap
value looks better than the other equities because of historical returns - which
may not hold going forward. That is always a function of the input data and one
reason that it is unwise to try to model some of the newer asset classes with
little historical data.
Q: When I compare
strategies, the strategy that resets the withdrawal rate every five years always
has a 100% success rate even though the final income is lower than I need.
A: Whether a strategy
succeeds or fails depends upon whether you have the desired income throughout
retirement. If you select this strategy, you are willing to
live with the rule which includes buying an annuity at age 85, even though
this may cause your income to drop. The rule is designed to have some income to last your lifetime. By definition, this
strategy is successful. However, should you create your own custom strategy
trying to mimic this strategy, it will very likely have a low success rate; because at age 85 there will probably be a big drop in income as you use
up the original 85% of your assets and purchase an annuity with the remainder.
Questions on JFP Paper
We have a couple of questions about how
to apply the model.
Q: You indicated
in your article that when you tested the model for a 40 year time horizon,
5.3% IWR, and the following decision rules, you had a 99% success rate:
= 20%, Cut = 10%
Fall = 20%, Raise = 10%
understood that this particular model yielded the greatest PV total of all
the different scenarios.
A: Right on both accounts.
client wants to be the 1% that failed. So, either you tweak the knobs above
to generate a 100% success rate, or alternatively, you extend the time
horizon to build in a cushion (e.g. for a 60 year old, assume a 45 year
horizon instead of 40 years). Also, different clients have different time
horizons based on starting age, etc. This all leads to a question about IWR
as a function of time horizon. Have you done any work determining
appropriate values of Exceeds, Cut, Fall, and Raise for a given time horizon
such that you maximize IWR while still achieving a 99% or 100% success
A: In my paper, I show what works for a 30 year time horizon. I haven't
done the work, but it would be a rather straightforward exercise with
Retirement Quant to look at any time horizon. One would first set a
retirement age and expected lifetime. For example, you could test 50 years
by retiring at 50 with a lifetime of 100. Then it would require a few runs
of setting the decision rules to see the effect to come up with appropriate
alternatively if that is too complicated, setting Exceeds, Cut, Fall, and
Raise to the values above and then determining the maximum IWR as a function
of time horizon that achieves the 99% or 100% success rate?
A: That works too and would only require the click of one
button - that is one of my built-in functions. It may not give you the
'optimal' strategy in terms of maximizing total withdrawals but it should
get you pretty close.
possible that some clients would want to maximize PVtotal rather than IWR.
Have you done any analyses as above to determine IWR as a function of
time horizon where you are instead maximizing PVtotal (while still achieving
the 99% or 100% success rate)?
A: In the work I did, you will see that for a given time horizion (in my
paper 30 or 40 years) there are a number of different IWRs that will achieve
the same maximum PVtotal. The different IWRs will produce the different
income profiles (uniform, aggressive or progressive). So your question
needs to be refined a bit to something like: for a uniform withdrawal
profile, what IWR will maximize the PVtotal and achieve a 99% success rate?
Again, I've got the numbers for 30 and 40 years and I don't think it would
be too much trouble to do it for other time horizons