Isn't FireCalc and Monte Carlo simulations based on higher gains?

[modhatter]

Isn't Firecalc and Monte Carlo simulations based on an average stock yield of around 10% or better. (historical numbers) I guess you could also include bonds in thqt same scenerio, as they also had a great run for a while. Also, stock losses were tempered somewhat due to over all much higher dividends paid out historically.

If it is, and if all predictions that the FUTURE ten years will see returns not higher than 7% at best, then those simulations would be very flawed for people trying to calculate their withdrawel rate. Is that not true?

If I am correct, wouldn't a new adjusted formula have to be applied for future sustainabillity of a portfolio?

 

[neo]

With a Monte Carlo simulation, you should be able to set the future average stock market return to whatever you want. You set the basic assumptions about the future, not the program. I use 7% myself, but who knows? Personally I think 10% is way too optimistic.

 

[mathjak107]

Long term certain asset mixes always return very close to the same average return, From about 12 years out almost any time period i pulled out since the 1960's magically came out to around 7% or so for my 50% stock mix . The longer the time frame the smoother the returns get.

Its wild how each year i can be up 30 or down 15% or go 3 years with no growth but when all is said and done bingo that long term average holds true.

I found 14 years was almost perfection to the average and was right on the mark.

 

[neo]

Long term certain asset mixes always return very close to the same average return, From about 12 years out almost any time period i pulled out since the 1960's magically came out to around 7% or so for my 50% stock mix . The longer the time frame the smoother the returns get.

Have you tried 1969 - 1981?

 

[mathjak107]

dont forget we arent just talking stock but a diversified mix of multi asset classes. gold and commodities sored thru that time period, real estate took off, bonds sucked ,stocks sucked

 

[youbet]

Isn't Firecalc and Monte Carlo simulations based on an average stock yield of around 10% or better.

No. Neither Firecalc or Monte Carlo simulations are based on averages. Both specifically use actually historical data (Firecalc) or assumed future data (Monte Carlo). The whole point of Firecalc and Monte Carlo analysis is to NOT use average data.

 

[youbet]

If I am correct, wouldn't a new adjusted formula have to be applied for future sustainabillity of a portfolio?

For Firecalc, if the past is not a reasonable predictor of the future, results will not be accurate. Firecalc does not use a "formula" with constants such as average returns or inflation rates. It uses actual historical data of yearly periods.

 

[mathjak107]

Monte carlo simulations are the actual returns for that particular year combined with other returns from random years and the long term averages calculated for each random selection of years.

 

[neo]

(a) A Monte Carlo simulation does uses an average; it assumes an average for the random distribution that it generates as a basis for future "histories". You (the user) should be able to specify this average. (b) A Monte Carlo simulation is not based on actual, historical, returns, except insofar as it may (or may not) use average past returns, type of distribution curve, standard deviation, etc. in its generation of future random histories. Using actual yearly returns in a Monte Carlo simulation would not make sense, the whole point is to average over many possible future histories.

 

[youbet]

; it assumes an average for the random distribution that it generates as a basis for future "histories".

Agreed. The important part is that a random distribution of data is generated. The "average" is not used for each period.

I'm afraid I wasn't very clear. What I am saying is that folks who use some average as the data point for each time period are making a huge mistake. Using, say, 7% annual returns as an assumption for each year going forward will be misleading. Monte Carlo will generate a group of data points averaging 7%, very different from saying each time period is 7%. That is, that sigma = zero.

 

[New Thinking]

The poster could be implying that since Monte Carlo or FireCalc tend to base modeling based on historical returns, that it is possible that we will see returns over the next 20 years that have never been seen before..A possibility in the world markets since we have never seen a demographic like the baby boomers before..Also, the whole world has moved away from pensions to individually managed accounts and we don't know how retirees will react.

 

[mathjak107]

TRue we might not see those returns but thats why you want at least a 95% success rate. That means at least 95% of all the historical combo's yielded that amount over x amount of years but theres all ways the small chance "THIS TIME ITS DIFFERENT"

 

[modhatter]

Yes, that is exactly what I mean. The historical data in probability will not coincide with the next 20 years. I don't like the remedy as the last poster suggested in accepting numbers knowing that the 95% will probably not hold true. It seems that there would be some mathematical correction that could be added based on probability of lower returns.

Maybe something as simple as: if the last 50 years produced returns averaging 10% and current probability is more like 7%. That is a 30% decrease. if the 4% withdrawel rate were to be reduced by 30%, we would be left with a 2.8% withdrawel rate as opposed to the 4%.
Sound feisable

 

[ats5g]

Yes, that is exactly what I mean. The historical data in probability will not coincide with the next 20 years. I don't like the remedy as the last poster suggested in accepting numbers knowing that the 95% will probably not hold true. It seems that there would be some mathematical correction that could be added based on probability of lower returns.

Maybe something as simple as: if the last 50 years produced returns averaging 10% and current probability is more like 7%. That is a 30% decrease. if the 4% withdrawel rate were to be reduced by 30%, we would be left with a 2.8% withdrawel rate as opposed to the 4%.
Sound feisable

Bill Bernstein did a similar thingee in The Retirement Calculator From Hell - Part II. You could also game FIREcalc by just upping the expenses of your portfolio by 1-5%, depending on how much you think future returns will be below past returns.

- Alec

 

[modhatter]

Now that's good thinking. Why couldn't I have thought of that>

 

[dory36]

Yes, that is exactly what I mean. The historical data in probability will not coincide with the next 20 years. I don't like the remedy as the last poster suggested in accepting numbers knowing that the 95% will probably not hold true. It seems that there would be some mathematical correction that could be added based on probability of lower returns.

Maybe something as simple as: if the last 50 years produced returns averaging 10% and current probability is more like 7%. That is a 30% decrease. if the 4% withdrawel rate were to be reduced by 30%, we would be left with a 2.8% withdrawel rate as opposed to the 4%.
Sound feisable

The basis of FIRECalc is "worst case", not average. The average return could be 30% for all of history, and if there was one single sequence of years that failed when the withdrawal rate was 4%, the SWR will decrease.

So saying that future averages are different than past averages, while undoubtedly true, doesn't affect what FIRECalc will report. Only if the next XX years are worse than the worst XX year period anytime in the last 135 or so years will you be be over-optimistic if you rely on the historical SWR methodology that FIRECalc implements.

An analogy... if you are trying to avoid getting too cold, and you investigate and find that in your location it has never in all recorded history dipped below 65 degrees F in August, you might feel reasonably safe leaving your winter coat in storage. So we're talking about what the weather likely WON'T do, just as in FIRECalc we are talking about what a portfolio likely won't do. That is entirely different than betting on what the weather (or the portfolio) WILL do.

 

[Nords]

Now that's good thinking. Why couldn't I have thought of that>

While you're at it, I should point out that FIRECalc uses the CPI or the PPI for inflation.

Over the last century, Dimson & Marsh calculated U.S. inflation at about 3%.

Over the last 30 years they've determined it was about 5%.

Which of the above numbers do you want to run against your hypothetical max 7% future returns?

 

[mathjak107]

My own belief in the "laws of large numbers " says go with the 3%