“He who lives by the crystal ball soon learns to eat ground glass.”
–Edgar R. Fiedler
Financial planners are asked to be, among other things, soothsayers, and determine what the future holds for their clients. Since we have not yet invented time travel to go into the future to be able to accurately report on what will happen, we must use forms of prognostication to try to predict what will happen and how our clients’ plans will prepare them for that future.
There are many tools used in making those projections. Most people use simple averages. The stock market averaged a total return (capital appreciation plus dividends) of 9.4% from 1900 through 2011, and inflation was 3%, so planners expect a real return of 6.4% and use that as an average real return going forward. The problem with such simplistic calculations is that the market fluctuates, often wildly. Rarely does it yield 6.4% real return, and it certainly won’t do that every year for the next 30 years.
Others, who realize the shortcomings of the simple average method like to use what is called aftcasting. Aftcasting means acting as if you were dropped at some time period in history and lived for the next X number of years, and in the end, if you still have money, then you succeeded. They look at time periods, usually from 75-90 years back, run each year as if you were plopped down there, and if you succeed a given percentage of the time – usually 80% – 90% – then they declare your plan a success and move on.
While aftcasting is better than simple averaging, I do not believe it to be better than accurately derived Monte Carlo projections. Monte Carlo projections take a random number in a provided range and use that number for a given return. So, let’s say that I want to project 5 years of real returns. The first year might be -1.5%, the second year +4.6%, the third year +3.3%, and so on. I then repeat that exercise repeatedly to generate an average and an expectation of what best and worst cases could be.
I use Monte Carlo simulations because there are simply too many outcomes to mathematically model, so simulation is the best proxy available to get a good idea of how those outcomes are distributed. They also allow us to project other things in your life which may have an effect. What if you get laid off? What if a spouse dies? What if there’s an oil shock and inflation spikes for a couple of years? What if cancer is cured and people live to be 120 years old? Aftcasting can’t account of all of these things because some have never happened before, and the ones that have might not be adequately covered in an aftcast.
Firstly, let me state that I am no shaman, and I do not believe that Monte Carlo models are perfect. They are not. I’ll run a million Monte Carlo simulations to project what might happen in your life, and I can almost guarantee that your life won’t fit any of the projections. However, a Monte Carlo simulation will provide many more samples than an aftcast will (1 million futures versus 75-90 futures), and it will account for outcomes that have never happened before. The market has never gone up 100% in a year. It’s highly unlikely, but it’s possible, and a well-designed Monte Carlo simulation accounts for that unlikely event.
Humans do very well at accounting for what they’ve seen before. They know how to pattern-match and how to make analogies. That’s why aftcasting is so appealing; it simply replays history and tells you how you would have done given that history. While it’s a reasonable approach to determining what might happen in the future, it’s limited in that it relies on past performance to make future predictions, and if the future doesn’t look like the past, then knowing how you would have done in the past won’t provide much of a guidepost for you in an uncertain and a new future.