FREE NEWSLETTER
Go to main Forum page »
Trusts are said to be a tool for its grantor to control from the grave. The first things I look at every morning when I open HD are Jonathan’s quips above the “Latest Posts”, and the most recent thoughts in the “Get Educated” section.
I sort of think of “Get Educated” as the legacy that Jonathan has granted to us; his way of continuing to guide and educate.
One of today’s topics is Monte Carlo analysis, put another way, is your money going to last as long as you? You can pay an advisor big bucks to do an analysis for you, or you can click on the link that Jonathan provides and get a general idea in about one minute.
Here is Jonathan’s Get Educated thought for the day (06/22/2026):
“MONTE CARLO analysis. Suppose you wrote down all annual historical stock market returns on index cards and randomly selected 30 cards—to get a hypothetical 30-year return—and did so 10,000 times. You’d have a sense of the range of possible 30-year returns and their likelihood. To see Monte Carlo analysis in action, try playing with Fi Calc’s calculator.”
Just a clarification – It is my understanding that FiCalc does not perform Monte Carlo analysis. The About section on the site says this:
FI Calc is a calculator that utilizes backtesting to evaluate different withdrawal strategies.
It uses historical asset return data to backtest a retirement plan. This is a fundamentally different technique than Monte Carlo analysis, but both have value in evaluating a retirement plan.
Rick, same understanding here — and I was actually corrected on this exact point in a comment I made here a while back.
I have done some work with simulations. Hope this explanation helps folks understand the method better. Such a dive is key because like all models, Monte Carlo Simulations (MCS) have inbuilt assumptions.
The theory is the following: The return for any given year is random, with a probability attached to every possible return. We use the term “probability distribution” to represent this idea. We make the (questionable) assumption that returns are independent across periods. Then, we draw 30 (or some arbitrary number to represent the planning horizon) values from this distribution. These draws obey the idea that the chance of a given return being drawn equals its probability. Then, these 30 random draws are a sequence of returns over 30 periods. For this given sequence of returns, we can compute how the initial portfolio would grow over 30 periods. We repeat this process of drawing a sequence of returns N times, where N is a large number (say 10,000).to get 10,000 end values. Plotting these end values gives us the distribution of ending portfolio values.
In practice, we do not (and cannot) know the distribution of returns. Hence, we make the assumption that the returns over the past periods (months, quarters or years) approximates the underlying distribution. How many past periods? The usual answer is “the more the merrier,” which answer assumes that the return distribution is stable and does not change. We continue to assert that returns are independent. We can then implement the simulation.
Of course, the validity of the simulation depends on how well the empirical distribution (actual returns) mirror the true return distribution over the (future) planning horizon. That is, there is the implicit belief that past patterns will hold in the future, and that our sample of past returns is large enough to effectively model the true distribution The assumption of independence across periods could also be questioned, particularly if we consider business cycles.
Thus, while simulations are an extremely useful tool, they are not fool proof. For model simplicity, they also tend to leave out things like fees, expenses and taxes which can drag returns down. Even so, simulations provide a great estimate of how things are likely to turn out and, more important, an estimate of the distribution over the range of possible outcomes.
Hope this helps!
Thus, while simulations are an extremely useful tool, they are not fool proof.
They can’t predict a lengthy stay in long term care either. Man plans, God laughs.
Thanks, Ram, it does help.
Dan, have you somehow been spying on what I’ve been writing?
Mark, an expanded conversation on the subject could be very helpful to people.