If you’ve ever spoken with a financial adviser about retirement, they probably suggested running a Monte Carlo simulation program to help determine how financially prepared you are. These programs randomly combine historical outcomes (annual market returns for the most part) with personal financial data to arrive at a probability of success (i.e. that you won’t run out of money in retirement). Telling clients that the program runs thousands of scenarios to arrive at this information sounds like it is a very thorough, and, therefore, accurate, process. But there are a few problems. To begin with, not all of these programs are created equal. Some allow more inputs than others, some run more iterations than others, and some use different methodology than others, clearly making some far more accurate than others. There really is no way for the end user (you) to know exactly what you are getting, no matter how wonderful your adviser tells you their program may be.

Let’s put the quality issue aside and assume that the program being used is one of the better ones. What is it really doing? Think of it like this – a handful of marbles are tossed into the air all at once to see how many will fall into a single cup below them. Because of the variability of how the marbles are tossed (the number of marbles used, the amount of force used each time, the weight and size of each marble, etc.) this needs to be done thousands, if not hundreds of thousands, of times to project a reasonably accurate probability of a specific number of marbles landing in the cup for any given throw. Now apply those results to project your own success rate for landing a given number of marbles in the cup, tossing them yourself, and using your own marbles. Monte Carlo uses past performance to project future results. But the past is not a good predictor of the future if everything is not perfectly equal, and none of your throws will be perfectly equal no matter how hard you try. There's also no way to know if any of your throws will be perfectly equal to any of the test throws. To make this even more complex, what if you used a different number of marbles then the test throws, even once? How would that affect the projections since that occurrence was never tested? One of the failings of Monte Carlo simulations is that they cannot factor in variations of the user, especially regarding saving, investment, and spending habits. They use historical market results and the assumption of 20 - 30 years of perfect consistency on your part to generate a probability of success. That’s not how things work in real life.

While Monte Carlo simulations might be a nice “double-check”, they do not help you to reduce risk, retire earlier,

or keep you from outliving your savings/investments. If you run 5,000 simulations, and your money lasts through 4,000

of those cases (80% probability of success), will you feel secure? That leaves 1,000 cases where you end up broke.

All of a sudden, an 80% probability of success isn't so reassuring.

The following are a few of the issues when considering using Monte Carlo for retirement planning:

• Monte Carlo relies on a probability outcome based on historical situations that have very little predictive ability. There are no individual, or unique situations considered. How adequately then, can it capture your unique circumstances and future actions? How closely do the outcomes really mimic what you might experience or how you intend to handle different situations? Human nature is unpredictable, and Monte Carlo cannot account for any changes or inconsistency in future behavior.

• Monte Carlo simulators consider thousands of iterations, leading you to believe that they have captured all (even most?) of the possible outcomes that should be considered. But do you really know if the programmers have adequately captured the most relevant outcomes? All Monte Carlo simulators are not created equal and one can be quite different from another. The margin of error in these simulations can be anywhere from +/- 5% to 40%. Knowing that, how would you feel if you were told that you have an 85% chance of success? Even if that was accurate, are you satisfied knowing that you have a 15% chance of failure, especially if that is likely to occur when you are in your 80’s? If there’s a chance that I’m going to run out of money, I want to know exactly when it will happen, why it will happen, and what I can do to keep that from happening. How much more money do I need – a year’s worth? 5 years? 10 years? Can I reduce expenses to fix the problem? By how much and by when? I want to see specifics in black and white so that I can take specific action. A hardy pat on the back and an 85% likelihood that things will work out based primarily on the stock markets past performance doesn't exactly inspire confidence.

• The last point brings up another question. Even if your investments have consistently achieved a rate of return equal, or better, than the general market (exceedingly rare), it’s likely that you will take a more conservative stance as you enter, or continue in, retirement. Do you intend to maintain exactly the same exposure to equities through every year of retirement? Monte Carlo does not account for shifts in investment strategy.

• Monte Carlo doesn't recognize that the overall performance of your investments is highly dependent on the sequence of returns, not just on the average rate of return. This is a critical flaw.

(See our Blog post from 6/1/23, "Sequence Of Returns Risk", And Why It Matters".)

• Monte Carlo simulations don’t account for bear markets or recessions very well (directly related to sequence

of returns) but this is likely to happen at some point during retirement.

• The simulations do not account for the direct relationship between expected investment returns and the need for realistic and fluid cash flows. There is no allowance for unexpected spending variations, or a need/desire to unexpectedly sell investments to raise cash (for any reason).

• It considers multiple starting positions for the market rather than the current, actual starting point. Why assume that the market currently reflects a price to earnings ratio well below the historical norm, for example, increasing the likelihood for near term appreciation and a greater rate of return, if that isn’t the case? Including an inaccurate starting scenario significantly affects the end result and increases the margin of error.

• Human nature cannot be factored in. When the S&P 500 fell more than 18% in 2022, how many people sold investments to raise cash rather than hold all their positions? How many bought back in (at just the right time), added to their existing positions as the market rose, or stayed out of the market as it regained it's high's? Monte Carlo cannot factor in special circumstances that may require additional spending/distributions at random times and in random amounts. While no planning tool can predict the future, one that requires perfect consistency over a 20-30 year time frame is among the least realistic. The ability to plan “what-if” scenarios on the part of the user is far more effective at illustrating possible outcomes.

All things being considered, Monte Carlo simulation may be a more useful tool for advisers than for their clients. It helps the advisor to quickly revise inputs to increase the odds of reaching a certain result. They can then present those inputs to the client as necessary steps, basically turning chances into choices. This may be a good way to demonstrate a strategy, but the client needs to be aware of the inadequacies of the simulations in arriving at those steps, and that their future actions will need to be perfectly consistent to give them any chance of the probability being correct. Not likely. Monte Carlo may be a good sales tool for advisers but be careful if you are going to rely on an inherently flawed probability % to assure you of not outliving your savings.