To the Editor:
In the November Registered Rep. magazine, two articles espousing the merits of Monte Carlo Simulations (MCS) were alarming to me. In the first article, “Retirement Roulette” by Brad Zigler, the scenario illustrated by Michael Kitces equates to a garbage-in/garbage-out type of scenario. The flawed assumption is that the returns an investor will achieve are symmetrical (for every 10 percent drop, an investor will immediately receive a 10 percent gain). From a mathematical standpoint, this will not compute (-10 percent + 10 percent won't equal 0 percent). Rather, research by Brian Rom at Investment Technologies has found investment returns from an annual standpoint are 65 percent asymmetrical. We can expect that number to be higher using monthly returns, the real way in which annualized returns are realized by an investor.
Because Mr. Kitces uses standard deviation in the equation, he wrongly implies symmetrical returns will be achieved over the life of the investor. The reality is that Registered Rep. ran a Monte Carlo simulation (MCS) on something that has 0 percent probability of occurring.
In Mr. Luxenburg's article, “Needlessly Scrimping?” he writes: Some planners prefer to run simulations every year. Say a portfolio has a 90 percent chance of success. Then, after a bad year in the markets, the chance of success drops to 80 percent. The client can elect to continue withdrawals — and hope that a market recovery improves the portfolio's odds. Some clients may insist on immediately reducing withdrawals, maintaining the 90 percent figure every year. If MCS does what it claims, why would the second year analysis show such a varied outcome? Were ‘all’ distributions not considered? The answer is effectively no, you cannot substitute mean and standard deviation for real returns because extreme outliers like today's market environment, while not new, are considered outside the range of standard deviation and ignored.
Monte Carlo Simulations, better known as re-sampling in the statistics community, are very powerful and highly accurate when done correctly. It is imperative that the calculation use the original data, not the average and its standard deviation. More critically, in order to infer what could happen instead of the probability of the past repeating itself, use bootstrapping. Bootstrapping takes the original data and replaces it on a random basis to provide a more robust estimate of what the real mean and standard deviation could be.
One of the most critical concepts that advisors need to understand, especially at this time, is that if your MCS is unable to show the client still on track versus the simulation run this time last year, it is based on incorrect assumptions and calculations. Verification of the model should allow the result to repeat itself in subsequent trials. At the end of the day, done correctly, the MCS should have illustrated not only the current market pullback, but the advisor should have recognized it and set a plan in place to make sure the client could sustain the event.
Brent E. Bentrim
Carolopolis Fiduciary Counsel Inc.