WSJ Article on Monte Carlo simulators

Post questions about how to use the planner, user inputs, how the planner works, and comments and suggestions.
Post Reply
tyee1850
Posts: 7
Joined: Thu Oct 23, 2008 7:22 pm

WSJ Article on Monte Carlo simulators

Post by tyee1850 »

Hi,

I was wondering your perspective on this article appearing in today's WSJ Online.

http://online.wsj.com/article/SB1241218 ... outset-box

It would seem your tool is highly flexible, but just wondering on your perspective.

Thanks.
DJW
jimr
Posts: 821
Joined: Thu Feb 28, 2008 6:48 pm

Re: WSJ Article on Monte Carlo simulators

Post by jimr »

The article raises legitimate questions about the accuracy and reliability of Monte Carlo Simulations used for retirement planning. But IMO, retirement planning is an inherently unreliable exercise and that has nothing to do with MCS. Small errors in a plan's inputs compound over time and destroy the accuracy of the results. Monte Carlo based tools aren't the problem. The real problem is that we often can't nail down the inputs with enough accuracy for the outputs to be valid.

One of the early design decisions I made in the flexibleRetirementPlanner was to model only the high-level portfolio performance (return and standard deviation), rather than trying to model the performance of the assets inside the portfolio along with their cross-correlations. This simplification allows the tool's user to control how portfolio performance will be handled in the simulation, rather than embedding this important input into the simulation logic.

Another step I took in the design of the FlexibleRetirementPlanner was to allow users to change the parameters for their portfolio's performance over the life of the plan. This is done in the additional inputs window.

One key example of how this feature can be used is the ability to "override" the portfolio performance parameters for just a single year to simulate a market crash. For example, you can set the portfolio return to -25% and the std-deviation to 0 in the year after retirement starts. This will simulate a 25% portfolio loss early in retirement. This ability lets you to manually "poke in" black swan events and override the normal distribution that's assumed in the return/std-dev parameters in the main window.

Still, the flexibility that's built into this tool puts a huge burden on users to know what they're doing. Unfortunately, there's no easy way around this. Either the tool buries the assumptions (which may be wrong), or it leaves it up to the user to manage them for themselves.

My design philosophy was to expose everything to the users so they can experiment themselves. IMO, the experimentation that these tools allow is their most valuable feature. In many ways, the probability of success is only the beginning. The most useful output is what happens to that probability as the inputs are varied and which things that you can control will have the greatest impact on your plan's likelihood of success.

Jim
tyee1850
Posts: 7
Joined: Thu Oct 23, 2008 7:22 pm

Re: WSJ Article on Monte Carlo simulators

Post by tyee1850 »

Jim,

Thanks very much for the prompt and thorough reply.

I agree - your tool provides the best way to model through these scenarios.

Dave
jimr
Posts: 821
Joined: Thu Feb 28, 2008 6:48 pm

Re: WSJ Article on Monte Carlo simulators

Post by jimr »

Dave,

We've been having an interesting conversation over on bogleheads.org about this article as well.

You can follow it here.

Jim
Post Reply

Who is online

Users browsing this forum: No registered users and 26 guests