Monte Carlo vs. Asset Allocation

[al4trade]

Can the group please share with me the thoughts on running FireCalc with actual asset allocation vs. the Monte Carlo method.

In what ways is one superior/inferior to the other?

Is one method considered any more accurate than the other?

Thanks

[sgeeeee]

Neither will predict the future.

Stock, bond and inflation results are correlated to each other and to the recent past in complex ways that cannot be captured by mathematical formula. Historical simulation gets around this problem by simply using real data. Whatever the correlation was, historical simulation captures it. Monte carlo simulation typically ignores the correlations or, at best, provides a poor approximation of some of the relationships. Historical simulation is limited to asset classes that have been accurately tracked and tabulated for 130 years. This is a serious limitation if you want to invest in real-estate, commodities, small-caps, . . . or anything other than the S&P 500 and bond indexes. Monte Carlo simulation can take whatever historical returns are currently available, assume the future will provide a similar distribution of returns and examine portfolios that include that asset class.

The best answer: Use both and try to understand the strengths and weaknesses of each. If they do not provide answers that are at least in the same ballpark, you or the simulation is doing something wrong. And getting in the ballpark is all either simulator is going to do as far as predicting your future.

[al4trade]

Here's an article from Marketwatch comparing historical-based projections vs Monte Carlo.
http://tinyurl.com/gwgo7




Al4

(moderator shrunk URL)

[dory36]

The article compares monte carlo simulation with dependence on average rates of returns. The author says:

Way back then, many of those Americans probably calculated how much they could withdraw from their retirement accounts using historical rates of return as their guide, the ones that show that large-company stocks gaining on average 10% per year since 1926, small-company stocks gaining 12%, bonds gaining nearly 6%, and U.S. Treasury bills gaining 3.7%.

In fact, historical simulation (instead of averages) with a start year of 1926, as he suggests, gets a safe withdrawal rate of about 3.5% (depending on your allocation), not the much higher 6-8% he uses.

His point is good -- a simulation that addresses volatility beats using averages. But monte carlo misses the correlations that sgeeee pointed out.

[Nords]

Dory,

This article was also posted by FundAlarm's "Ted" and I made some comments over there (before seeing it again over here). One of the other posters responded at length and in part: "Thanks particularly for your update on FIRECalc. I examined it many years ago and was unimpressed with its sophistication or methodology. Based on your comments I plan to revisit that tool."

FundAlarm's owner, Roy Weitz, was impressed with the exchange. If you send him an e-mail about the new version then he may write it up in his "Highlights & Commentary" article that he publishes at the beginning of each month. It's widely read and he's always looking for fresh material...