There are a number of programs designed to simulate market behaviour
for retirement planning purposes. Two of the better known programs are
RRIFmetric and RetireWare. There are three main methodologies for
doing this:

1. straight line projection,

2. Monte Carlo simulation,

3. major market cycle simulation.

The "flavour of the month" in retirement planning has become Monte
Carlo simulation, which simulates random variability in returns of
various categories of securities depending on the particular portfolio
asset mix, then runs thousands of scenarios over a specified time
horizon to evaluate the probability of one's retirement goals being
achieved based on specified monetary inputs. However, as Otar
(www.retirementoptimizer.com) and others have pointed out, Monte Carlo
simulation only simulates short term market cycles - long term or
major market cycles or corrections are entirely ignored. Otar (q.v.)
developed a program that uses the actual historical data of the 20th
century instead of assumptions about volatility and return to test
retirement planning projections. He found that Monte Carlo simulations
are invariably far too optimistic and actually fail when tested
historically, a claim that I too have confirmed using his software.
However, my objection to Otar's approach is that the 20th century
itself is only one possible scenario and that it should be possible to
elaborate Monte Carlo simulation into a more comprehensive approach
that uses simular principles to superimpose long term variations onto
the short term fluctations, using historical corrections, durations,
recovery times, and intervening periodicities to allow one to test a
retirement projection against a simulation that incorporates thousands
of scenarios based on both minor AND major market cycles.

I am trying to find a retirement planning program similar in principle
to RRIFmetric, RetireWare, and Otar that goes beyond Monte Carlo
simulation in just the manner that I have described above, and is not
limited to historical market data, but despite several intriguing
leads I have been unable to find such a program. Yet such a program
should not be that much harder to develop than the existing Monte
Carlo simulations and I am hopeful that it exists. Your mission,
should you decide to accept it, is to find such a program available
online for free or for purchase.

Two comments:
First, looking to historical data is largely not used anymore,
especially data that would cover the entire 20th century.  Many
financiers see historical data as irrelevant and prefer to use the
random walk notion.

Second, it is very difficult to design a program of the type you are
asking for, not because the variable are difficult to understand, but
because it could be a violation of securities laws to recommend
securities to individuals.  (That requires a securities license)

Hi! thanks for your comments. Just to clarify, I hold an FMA
professional financial designaton so do have some knowledge in this
area.

With respect to your first point, while it is true that historical
returns cannot predict future results, markets do exhibit some
regularities within broad limits with a high level of probability. For
example, it is highly improbable that tomorrow morning the markets
will be 50% lower. Over the long term market returns are about 11%.
There are regular bull and bear market cycles of varying severity,
durations, recovery times, and intermittent intervals, within broad
limits, superimposed on short term volatility that also exhibits
certain characteristics. The notion that markets are totally random is
an academic dogma that is becoming increasingly doubtful, especially
amongst professional investors. Buckminster Fuller has stated this as
long ago as 1938. That said, there are high levels of variability and
that is why sophisticated modelling programs are so desirable. One
cannot make any plan at all without making some sort of assumptions.
The point is that Monte Carlo simulation is simplistic, and that
tested against historical data, it fails. Consequently a more
sophisticated model is desirable.

With respect to your second point, such a program as I describe does
not violate securities legislation, and such programs already exist. I
instanced RRIFmetric, RetireWare, and Otar, and there are others. For
one thing, the Securities Act prohibits the recommendation of specific
securities. The program that I describe does not do this, or indeed
"recommend" anything. Second, educators, writers and publishers are
exempt from this restricton, provided any such "recommendation" is
secondary to their educational, writing, or publishing activity. To
apply this restriction to any of these areas would violate fundamental
Constitutional liberties of freedom of speech and self expression. A
Disclaimer such as the following would be sufficient to obviate any
such restriction: "This software is intended for research and
educational purposes only. Nothing herein shall be construed as
advising others as to the selling or buying of specific securities."

With regard to your first rebuttal; well.. sure there is an
expectation of return... and in equity markets historically it has
been 11%.  But to go by secotr or individual security, you'll find the
deviations are too massive to be useful other than, yeah, things
should go up over time.

In your original post, i was unable to glean your purpose.  I assumed
it was to help people to choose investments thatthey would actually
invest in.

Thanks for your continuing interest. All such comments are more than
welcome. Hopefully they will be useful to others too.

Well, retirement programs don't go by sector or individual security.
Actuaries use asset class, typically cash, real estate, bonds,
Canadian stocks, U.S. stocks, and international stocks, and there are
meaningful average differences between these categories over the long
term (all retirement plannng should be long term, you really should
have an explicit retirement plan in place by age 40 at the latest -
the shorter your time horizon the more uncertain your plan). A
retirement plan is typically built up based on percentage allocations
to different asset classes based on risk tolerance and need for return
(also based on one's general philosophy of markets and investing and
what one is willing or able to do). ULTIMATELY, of course, one has to
choose specific securities, but this goes beyond retirement planning
per se, as you pointed out. Good financial planning works from the
"outside in," and is synthetic or holistic - SYNERGY. Specific
securities selection is actually the last, not the first, step. Many
people err in this regard, however.

An interesting way of introducing long term cycles to retirement
planning in a programming environment might be emphasizing more recent
severities, durations, recovery times, and intervening intervals more
than older periods, like the difference between a moving average and a
weighted moving average (used by technical analysts). Whether this
would make for better planning who can say? I don't have corrections
data for longer than the period from 1960 to the present. I'm not sure
reliable numbers are available for much further back than the early
20th century. This whole area is still extremely new, and there are
opportunities for far more refined analysis than what exists. I
recommend you take a look at RetireWare (www.retireware.com) and Otar
(www.retirementoptimizer.com) to see what's been done. Then you will
have a much deeper understanding of what is still possible. I wish I
had the programming skill myself to develop these ideas further. I
would like to develop these skills myself, or perhaps work on a
collaborative basis with someone who would like to explore these
possibilities, in which we could educate each other, but there are
only so many hours in the day. Hence my interest in finding a program
that already does this.