FIRECalc Question

[ERReady]

Maybe someone can help with a problem I'm having with FIRECalc.

--Start with all the defaults, except for one thing - select 5 year treasuries.
--Look at the result: "A withdrawal of about $29,770 is the highest withdrawal that would have survived 95% of the periods tested"
--Now submit the exact same input except for one change - change the TIPS return to 1% instead of the 2.5% default. Don't check the TIPS Button, choose 5 year treasuries again; just change the return in the TIPS box. In other words, leave everything exactly the same as the first trial, except for the change in the TIPS return box.
--Now check the results again: "A withdrawal of about $29,185 is the highest withdrawal that would have survived 95% of the periods tested"

If I'm selecting the exact same mix of five year treasuries & stocks (25%/75%) both times, why would changing the return in the TIPS field change the historical safe withdrawal amount? Am I missing something? What is the correct withdrawal rate that would have survived 95% of the periods tested when selecting 5 year treasuries? Is it possible to know using FIRECalc?

 

[cc]

Hmm- you're right, the results do seem to change based on an unselected parameter. Instead of retiring maybe you should consider a career in software quality control!

 

[cute fuzzy bunny]

Firecalc takes a few dollars away if you fiddle with stuff and then dont use it. That just makes it mad.

 

[ERReady]

OK, another problem. In FIRECalc go with the defaults except enter these two changes:

1) Stock Allocation 35%
2) Select 5 Year Treasuries

Now review the output: "A withdrawal of about $28,145.00 is the highest withdrawal that would have survived 95% of the periods tested, using the equity split and other criteria you entered. This would be a withdrawal starting at about 4.33% of your starting portfolio, with adjustments for inflation...".

Now enter the same into REHP spreadsheet (4.33% initial withdrawal rate, $650,000, 30 years, 35% stocks/65% five year treasuries, etc.).

Depending on what month you choose in the "Schiller Data Selected" field, and whether you select front-end or back end, the range of survivability is between 48% and 64%, NOT 95% as in FIRECalc. Now we're no longer talking about just a few dollars.

So, which is the correct historical output? I doubt that a portfolio consisting of 35% stocks and 65% 5 year treasures would have supported a 4.33% withdrawal rate 95% of the time in the past. I don't trust those results, and I'm beginning to question whether FIRECalc can be relied upon to produce accurate results. Am I missing something here? This software appears to have been around awhile. Have others noted these problems?

 

[intercst]

OK, another problem. In FIRECalc go with the defaults except enter these two changes:

1) Stock Allocation 35%
2) Select 5 Year Treasuries

Now review the output: "A withdrawal of about $28,145.00 is the highest withdrawal that would have survived 95% of the periods tested, using the equity split and other criteria you entered. This would be a withdrawal starting at about 4.33% of your starting portfolio, with adjustments for inflation...".

Now enter the same into REHP spreadsheet (4.33% initial withdrawal rate, $650,000, 30 years, 35% stocks/65% five year treasuries, etc.).

Depending on what month you choose in the "Schiller Data Selected" field, and whether you select front-end or back end, the range of survivability is between 48% and 64%, NOT 95% as in FIRECalc. Now we're no longer talking about just a few dollars.

So, which is the correct historical output? I doubt that a portfolio consisting of 35% stocks and 65% 5 year treasures would have supported a 4.33% withdrawal rate 95% of the time in the past. I don't trust those results, and I'm beginning to question whether FIRECalc can be relied upon to produce accurate results. Am I missing something here? This software appears to have been around awhile. Have others noted these problems?

The problem is that 5-year Treasuries data doesn't exist for the entire 1871-2002 timespan. The only asset categories that have complete data sets are s&p500, commercial paper, and TIPS.

intercst

 

[ERReady]

Thanks. I understand that the data set for 5 year treasuries doesn't exist for the entire timespan. But it's my understanding that FIRECalc uses the same model as the REHP spreadsheet. What I don't understand is why the results are so dramatically different?

I like FIRECalc because it enables input of SS, pension, etc. But when I see problems such as these I must ask whether I can rely upon FIRECalc to do what it claims to do accurately. I tend to adhere to the "rat theory" - for every rat you see, there are at least 100 more you can't see. Perhaps there is a good explanation I am missing, but I think I have seen two rats already in a very short time. I hope I'm wrong, but I don't think that the limited data set explains the difference if both programs are using the same data. And I still don't know why the results change based upon an unselected parameter.

 

[cute fuzzy bunny]

I tried it too. Definitely something wrong. By changing the tips rate while 5 year treasury is selected produces a slightly different withdrawal rate depending on what tips return you put in.

Apparently some calculation is factoring something from that field in, albeit in a small manner, regardless of whether tips are chosen or not.

 

[Beau86]

When I input the varialbles into Fire, it says that the first withdrawal is at the end of the year - example: at the end of year 1

If I do an analysis on November 29, 2004, is the end of year 1 on December 31, 2004 or 12 months from now on November 29, 2005?

Thanks

 

[Beau86]

When I input the varialbles into Fire, it says that the first withdrawal is at the end of the year - example: at the end of year 1

If I do an analysis on November 29, 2004, is the end of year 1 on December 31, 2004 or 12 months from now on November 29, 2005?

Thanks

 

[Hyperborea]

When I input the varialbles into Fire, it says that the first withdrawal is at the end of the year - example: at the end of year 1

If I do an analysis on November 29, 2004, is the end of year 1 on December 31, 2004 or 12 months from now on November 29, 2005?

FIREcalc has no idea what day it is today. It uses year start and ends on calendar years. The end of year 1 is the end of the calendar year.

 

[JWR1945]

ERReady

Thanks. I understand that the data set for 5 year treasuries doesn't exist for the entire timespan. But it's my understanding that FIRECalc uses the same model as the REHP spreadsheet. What I don't understand is why the results are so dramatically different?

I have a modified version of the Retire Early Safe Withdrawal Calculator that gives me good visibility into what is going on with its calculations.

There are no entries for 5-year treasuries in the REHP calculator prior to 1953. The calculator assigns all 5-year treasuries an interest rate of zero percent from 1871-1952. The calculator assigns 5-year treasuries an interest rate of 5% from 2003-2010.

The REHP calculator tallies only completed sequences when reporting survival percentages. Since its data stops at 2002 (it uses dummy data afterward), there are 102 completed sequences. This is used in the denominator. (FIRECalc includes partially completed sequences when calculating survival percentages, but I do not know the details.)

There are 20 completed 30-year sequences starting from 1953 and going through 2002.

Looking at completed sequences with 30-years of actual data (start years of 1953-1972), here are the results:
a) When I selected 50% front/back withdrawals, there were 2 failures (1965 and 1966).
b) When I selected 100% withdrawals at the beginning, there were 7 failures.
c) When I selected 0% front-end withdrawals, there were no failures.

[The following results are with 50% front- and back-end withdrawals.] Some sequences starting before 1953 would have survived even with the zero percent [dummy] interest rate through 1952. Sequences from 1942-1952 all survived. The 1941 sequence failed. From 1871-1972, there were 45 failed sequences. The reported statistics were 45 failures out of 102 sequences.

It is a good idea to look at individual starting years when using FIRECalc and the Retire Early Safe Withdrawal Calculators. When using the Retire Early Safe Withdrawal Calculator, scroll over to columns S through AK. You can see the ending balances for the partially completed sequences.

Have fun.

John R.