FireCalc =5% 40 years with 97.4% success rate

[Hydroman]

I just ran the advanced version of FireCalc and arrived at a 5% withdrawal rate for 40 years at a 97.4% success rate.
All entries were left at default except I changed 30 years to 40 years and entered the following investment allocation:

Us Micro: 10%
Us Small: 10%
Us Small Value : 10%
Us Large Value: 40%
US LT Bond: 10%
LT Corp. Bond: 10%
1 Month Treasury: 10%

When I run the above with ESR Bob's 95% withdrawal rule I get 100% success.

Since I thought the standard SWR advice is 4% even using FireCalc, I am wondering if this is some anomoly or is 5% really feasable based on historical data?

[sgeeeee]

Quote:

Originally Posted by Hydroman . . . Since I thought the standard SWR advice is 4% even using FireCalc, I am wondering if this is some anomoly or is 5% really feasable based on historical data?

The 4% rule is based on studies that considers only two asset classes: 1) S&P 500 index fund and 2) a bond index fund. The study also assumes a spending model that adjusts for inflation each year regardless of the performance of the portfolio.

Change the spending model and you change the results. The aditional asset classes included in the asset allocation can also have an impact on the SWR. You need to understand that the asset classes other than S&P 500 and bonds are modeled differently. Historical data going back 130 years is not available for other asset classes so these assets are modeled using performance distributions only from more recent history.

[Rich_by_the_Bay]

Quote:

Originally Posted by Hydroman Since I thought the standard SWR advice is 4% even using FireCalc, I am wondering if this is some anomoly or is 5% really feasable based on historical data?

Doesn't it depend on how much you have and how much you spend. You can make a 10% SWR work at 100% if those two parameters are right, no?

[youbet]

No. How much you withdraw and spend from your portfolio is your WR. The safety of that WR can be historically tested using Firecalc. Or Firecalc can tell you a SWR (probability of success and asset allocation being givens) for some portfolio value.

There does seem to be some confusion over the definition of SWR when other sources of income (SS, pension, etc.) are brought into play.

[rmark]

The historical 4% percent rule is generally based on something like this -

small total reits corp total bills *Past u.s. market averages
* * * * stock * * * * * * * *bond
* * * * * 12 * * * * * * * * * * * * * * * * * *average return
* * * * * -1 * * * * * * * * * * * * * * * * * * annualized return
* * * * * -1 * * * * * * * * * * * * * * * * * * survivor bias
* * * * * 10 * * * * * * * * * * 6 * * * * * * nominal return (tax deferred or paid from work)
* * * * * -3 * * * * * * * * * *-3 * * * * * * *inflation
* * * * * *7 * * * * * * * * * * *3 * * * * * * *u.s. domestic saver (u.s. only)
* * * * * -2 * * * * * * * * * *-1 * * * * * * *average down
* * * * * *5 * * * * * * * * * * *2 * * * * * * *u.s. domestic seller (u.s.only)

u.s. history only, no investing cost, retiree 60/40 stock bond = 4%, 40/60 stock bond = 3%


We can be relatively confident in the history of broad market averages, but detailed slice and dice in an attempt to raise withdrawal rates generally requires unsupported assumptions about more recent asset classes.

[Arif]

Quote:

Historical data going back 130 years is not available for other asset classes so these assets are modeled using performance distributions only from more recent history.

So how far back do the other asset classes (small cap. etc) go back when FIRECAL includes them in its model?

[bongo2]

You have two anomolies. One is that the 40 result is better than the 30 year result because you are not including the "failure" year of 1969. The second is that you are over-weighting asset classes that have done quite well in the last 80 years (small and value) and may or may not continue to out-perform in the future.

I'm a little curious where the data for small, micro, and value came from.

[haha]

Quote:

Originally Posted by bongo2 You have two anomolies.* One is that the 40 result is better than the 30 year result because you are not including the "failure" year of 1969.* *

IMO, a very important observation. This "impossible result" is an artifact of Firecalc, and obviously could have no relation to any rational expectation.

An historical calculator is path dependent.

Ha

[Sheryl]

I'm really confused.

I thought FIRECalc looked at every 30 or 40-year period from 1871 on, so how would the 30 year result not include 1969?

[Martha]

The 30 year result would, but the 40 year wouldn't.

[obryanjf]

The 2 anomolies above I agree with. FIREcalc throws out the last (in this case) 39 years of simulations, because the data set is not a complete 40 year history. *The previous version used the results of the shorter and shorter histroies. *This method is less of a negative than the previous version.

On the plus side, some sims I have done indicate a .5% to ?? improvement in the SWR by a better diversification than the standard S&P and Bonds portfolio. *My .5% improvement included 4 other asset classes and I surmise than by adding assets like commodities, real estate, precious metals, emerging markets, etc; in some proportion, that maybe a 1% improvement can be had.

Take my data with a grain of salt though as I used a different model than the typical here, fixed inflation adjusted model.

job

http://early-retirement.org/forums/i...p?topic=8049.0

[Sheryl]

Quote:

Originally Posted by Martha The 30 year result would, but the 40 year wouldn't.

Still confused.* *:P

It seems there would be several 40 year periods that include 1969:

1966 - 2006 or 1940 - 1980 for example.

What am I missing?

[obryanjf]

I think 1969 would be included in your examples above, but not 1968 -> 2007 nor 1969 -> 2008.* A couple of the years right after 1969 were not that good either with their high inflation.* All that would be eliminated from the 'start' of a 40 retirement sim in FIRECalc versus a 30 year sim.

job

[youbet]

Quote:

Originally Posted by Sheryl Still confused.* *:P It seems there would be several 40 year periods that include 1969: 1966 - 2006 or 1940 - 1980 for example. What am I missing?

Years of market losses or high inflation have more impact when they are at the beginning of a period. Having 1966, 67, 68 or 69 at or near the beginning of a period will have more impact than at the end.

[Sheryl]

I see. I misunderstood the post which said,

Quote:

You have two anomolies. One is that the 40 result is better than the 30 year result because you are not including the "failure" year of 1969.

The figures from the year are still included in some time periods, but retirements starting in that year are not. Sorry for the confusion.

[Hydroman]

Quote:

Originally Posted by bongo2 You have two anomolies. One is that the 40 result is better than the 30 year result because you are not including the "failure" year of 1969. The second is that you are over-weighting asset classes that have done quite well in the last 80 years (small and value) and may or may not continue to out-perform in the future. I'm a little curious where the data for small, micro, and value came from.

Just re-ran the numbers using 30 years. Success rate reduced to 93.9%. Increased 95.9% using ESR Bobs 95% rule. So not that much degredation when 1969 is included. Statistically insignificant since this is all historical and who knows what the future really holds for us.

I am not sure I would call the over-weighting asset classes an anomoly. That is what the advanced version of FireCalc is supposed to be testing. But I do have the same question as to what historical data the small, micro and value came from. With that knowledge we can reach some educated judgement as to the validity of the data. I would not see a problem if it was limited to 80 years worth of data ( takes us back to before the 1929 crash which is good enough for me. Not sure 1800s data has much relavence today or going forward.

[Hydroman]

Quote:

Originally Posted by rmark We can be relatively confident in the history of broad market averages, but detailed slice and dice in an attempt to raise withdrawal rates generally requires unsupported assumptions about more recent asset classes.

Do you know what assumptions FireCalc is using for the recent asset classes? On what basis do you believe they are not supported?