Percent Success better for 45 year period vs. 40?

[tuffshed]

This make no sense to me, maybe someone can explain it to me. I entered my data into FireCalc, set the timeline at 45 years. My success rate was 92.9%. I then changed the timeline to 40 years, checked to see that no other entries had changed, hit calculate and now my success rate is 91.3%?

How can the success rate be better for a longer period? Doesn't it make sense that with the exact same assets that trying to make them last 5 more years would give you LESS success not more?

What am I missing here? And yes, I did run through them several times, the only thing changing was the time period.

 

[Cb]

FIREcalc includes data though 2012. By selecting the longer period you eliminate failures beginning in the late 60's that were included in the shorter retirement duration.

 

[tuffshed]

CB,

Let me see if I understand you.

As I understand FireCalc ran through all scenarios from data taken at all years since 1871 and projected it forward X years. In other words, I have X number of dollars, etc, used the growth figures for 1871, 1872, 1873... for 45 years and chart a line, next use the same dollars used with growth figures for 1872, 1873, 1874... for 45 years and chart another line....

If I chose 45 years then the last line charted is a calculation based from 1967 through 2012. If I chose 40 years then the last line charted is from 1972 through 2012. There should be 5 less lines then, true?

Now comes the part I don't understand from what you said, the longer period ELIMINATES failures in the late 60's? Wouldn't the longer period include lines from 1967 through 1971 that the shorter period does not include? Did you mean to say that from 1967 through 1971 that things were going good so these extra good periods (lines) skew the percent success to a better figure?

 

[Cb]

Mike, I don't know the particulars of your FIREcalc runs, but retirements beginning in 1968 and 1969 are frequently shown as failures. My guess is by shortening the duration of the retirement period your 40 year runs include 1968 & 1969 (with one of both shown as failures), while the 45 year run did not include them, boosting the success % slightly.

I think a better approach is to do FIREcalc runs of 15, 20, 25, 30, 35, 40 years, plot a curve of the resultant SWR's, and extrapolate your curve for a SWAG of 45+ year periods.

I'd also recommend reading a thread just down the page, regarding FIREcalc's accuracy & future. There were some disturbing (to me, anyway) observations in that thread and others with respect to the way FIREcalc uses historical fixed income data.

 

[ERD50]

CB,

Let me see if I understand you.

As I understand FireCalc ran through all scenarios from data taken at all years since 1871 and projected it forward X years. In other words, I have X number of dollars, etc, used the growth figures for 1871, 1872, 1873... for 45 years and chart a line, next use the same dollars used with growth figures for 1872, 1873, 1874... for 45 years and chart another line....

If I chose 45 years then the last line charted is a calculation based from 1967 through 2012. If I chose 40 years then the last line charted is from 1972 through 2012. There should be 5 less lines then, true?

Now comes the part I don't understand from what you said, the longer period ELIMINATES failures in the late 60's? Wouldn't the longer period include lines from 1967 through 1971 that the shorter period does not include? Did you mean to say that from 1967 through 1971 that things were going good so these extra good periods (lines) skew the percent success to a better figure?

I think I know where you are getting confused. It's not about which years are included, it is what starting years are included.

So a 45 year run can include starting in 1967 and ending in 2012.

But a 45 year run can not include starting in 1968 and ending in 2012 because then you would only have a 44 year run.

So lets just say 1968 was a great year in the market and it went up 40%, and then the market collapsed in later years. Starting with $X in 1967 would be better than starting with the same $X in 1968. Your portfolio would grow from 1967-1968 and be better prepared to weather the storm. But if you had the same $X go in at 1968, you missed the run up, and you hit the downturn with ~ 40% less money.

So in terms of doing a historical run, longer periods can filter out bad shorter sequences.

-ERD50

 

[photoguy]

How can the success rate be better for a longer period? Doesn't it make sense that with the exact same assets that trying to make them last 5 more years would give you LESS success not more?

As others have mentioned, you're basically dealing with the imperfections of finite data.

What I would do is run it for 30 years and look at the actual distribution of final portfolio values. You could consider runs with say less than X% of portfolio value to be failures for the 45 year run.

 

[Ready]

I've been thinking about sequence of returns risk as I read through this thread. For those who retired about five years ago, do you think it's too soon to declare that retiring around 2009 will likely put you on a path toward the higher end of the firecalc success runs? Here are the S&P Figures since 2009:

2009 23.45%
2010 12.78%
2011 0%
2012 13.41%
2013 29.6%

If not, how many years do you need to see positive results before you can begin to draw conclusions on how sequence of returns will affect your retirement?

 

[Animorph]

I've been thinking about sequence of returns risk as I read through this thread. For those who retired about five years ago, do you think it's too soon to declare that retiring around 2009 will likely put you on a path toward the higher end of the firecalc success runs? Here are the S&P Figures since 2009:

2009 23.45%
2010 12.78%
2011 0%
2012 13.41%
2013 29.6%

If not, how many years do you need to see positive results before you can begin to draw conclusions on how sequence of returns will affect your retirement?

If the US pulled something like Japan, even those numbers might not be enough.

Assuming the future is somewhat like the past, I'd say if you start at some WR that doesn't have a 100% success ratio (like 4% for 30 years) and then have a series of good years that lowers your WR to something that does show a 100% success ratio (like 3% for 25 years I think) then you are in good shape. Just keep running FIRECalc each year and checking for that 100%.

 

[robbridg]

I posted a similar thread a while ago. The time periods selected can be misleading. It's all about historically what time periods are used to evaluate your scenario. Also, a longer time hoizon gives Firecalc less periods on which to evaluate. What I do is see what time period gives me the worst results and go with that. For instance, I might get 100% success at 40 or 45 years but 30 years gives me worse results and 35 years gives me the worst results. I'll go by the 35.