Wonky Results - Thoughts?

[Dumello]

Hi. New member here posting a question about my FIRECalc results. I'm running multiple tests with nothing changed but the years of retirement and I'm getting results opposite what I expect with respect to success rate.

My scenario:
Spending - $250k
Portfolio - $3.6M
Years - 45, 40, 35, 30

I have other entries for 2 SS payouts, a pension that does NOT adjust for inflation, a few ins (inheritance, 1 year contributions) and a few outs (weddings), but all remain constant for each run. I run a future retirement year of 2020 in each scenario. Constant spending model with AA of roughly 60/40.

Here are my results:

45 years: 3 failures, 91.2% success rate.
40 years: 3 failures, 92.3% success rate.
35 years: 6 failures, 86.4% success rate.
30 years: 9 failures, 81.6% success rate.

Why is that? I would expect there to be more failures the longer I want my money to last, all other things equal. Especially given my spend rate.

Is this because there are fewer cycles of historical data that will last 45 and 40 years? That, or that the WR is lower when SS kicks in (modeled at 70). I don't know what the portfolio balance will be at that time, but it represents about a 3.5% WR (from the original portfolio), so that doesn't seem like it either.

Thoughts?

 

[SecondCor521]

The years that usually fail are the 1966 through 1973 start years, partly because there was high inflation and lousy market returns around that time.

I suspect it has something to do with the fact that when running for 45 years, you have fewer runs that span that bad period. But it doesn't make complete sense.

What I'd suggest doing is asking FIREcalc to give you the starting years that fail for 45, 40, 35, and 30 years. I think FIREcalc still gives you the option of downloading an XLS which will tell you this. Then you can see which years fail in one case and not in the other, and then you can look at the data to see why.

 

[Dtail]

I have noticed this scenario too. Not sure why either.
However when using some other calculators such as Fidelity, which runs Monte Carlo simulations, the more retirement years, the lower the success rate.

 

[marko]

What year are you starting SS?
Not sure if this is connected but unless I enter my SS start date at/beyond the current year I get false results. IOW, if I enter my true year of starting SS (2014) the numbers go wrong...I have to enter 2019 even though I've taken it much earlier.

 

[ziggy29]

I suspect that by the time you get to 40-45 years or more, the market's trend toward mean-reversion is so strong that the expected return band is tighter. So you would be less likely to see fantastic annual returns over a much longer time, but also less likely to see annual returns that are low enough to bust retirements. In other words, with a longer period of time, the bell curve for expected returns tightens and standard deviations are lower. Expected 10 year returns can easily be +/- 8-10% (annualized) of what is expected over a long period of time, and expected 40 year returns will usually be +/- (say) 3-4% of expectations. Thus, if only the worst outcomes will bust you, it's less likely with a longer time horizon.

 

[Independent]

Here are my results:

45 years: 3 failures, 91.2% success rate.
40 years: 3 failures, 92.3% success rate.
35 years: 6 failures, 86.4% success rate.
30 years: 9 failures, 81.6% success rate.

When I run FireCalc for 30 years, it tells me that it runs "118 cycles". Are you getting that number?

If so, 9 failures in 118 cycles => 109 successes. That is a 92.4% success rate. For 9 failures to equal an 81.6% success rate, I'd only have 49 cycles.

I have the same problem with the other durations. For example, 3 failures is consistent with a 91.2% success rate if the number of cycles is 36. FireCalc should be running about 103.

 

[Dumello]

When I run FireCalc for 30 years, it tells me that it runs "118 cycles". Are you getting that number?

If so, 9 failures in 118 cycles => 109 successes. That is a 92.4% success rate. For 9 failures to equal an 81.6% success rate, I'd only have 49 cycles.

I have the same problem with the other durations. For example, 3 failures is consistent with a 91.2% success rate if the number of cycles is 36. FireCalc should be running about 103.

For me, I only have 49 cycles in 30 years with retirement starting in 2020. It's 34 cycles in 45 years. Not sure why we would be so different.

 

[Dumello]

What year are you starting SS?
Not sure if this is connected but unless I enter my SS start date at/beyond the current year I get false results. IOW, if I enter my true year of starting SS (2014) the numbers go wrong...I have to enter 2019 even though I've taken it much earlier.

For mine I modeled SS starting many years from now.

 

[Dumello]

I suspect that by the time you get to 40-45 years or more, the market's trend toward mean-reversion is so strong that the expected return band is tighter. So you would be less likely to see fantastic annual returns over a much longer time, but also less likely to see annual returns that are low enough to bust retirements. In other words, with a longer period of time, the bell curve for expected returns tightens and standard deviations are lower. Expected 10 year returns can easily be +/- 8-10% (annualized) of what is expected over a long period of time, and expected 40 year returns will usually be +/- (say) 3-4% of expectations. Thus, if only the worst outcomes will bust you, it's less likely with a longer time horizon.

The results, however, don't appear consistent with that. For the 45 year the high outcome is 83M and the low is -2.7M, but for the 30 year the high is 44M and the low is -11M. I would think the range would tighten on the 45 year if it was reverting to the mean.

 

[Dumello]

As I think about this, I'm more confused. How can I have less failures over 45 years compared to 30 years with everything else being equal? I'd be trying to make my money last 15 years beyond the period where I've already had 9 failures, but somehow in the next 15 years I have -6 failures? It doesn't make sense.

 

[REWahoo]

OP, you don't provide enough information for us to attempt to duplicate your four FIRECalc runs. We need all your dates, amounts and inputs to check your results.

 

[cathy63]

For me, I only have 49 cycles in 30 years with retirement starting in 2020. It's 34 cycles in 45 years. Not sure why we would be so different.

This is because you chose the portfolio option of "A mixed portfolio consisting of the following assets (based on performance since 1927)". I don't know why, but when you make this choice, the last year that it will use as a starting year is 1975.

If you use the total market option, then it goes from 1871 through 1988.

 

[ERD50]

OP, you don't provide enough information for us to attempt to duplicate your four FIRECalc runs. We need all your dates, amounts and inputs to check your results.

+1. If OP provides a link to the data, we could provide more feedback.

Link to this set of data (Right-click and either "Copy Shortcut"
to paste into an email, for example, or "Add to Favorites")

is near upper right corner on the results page.

This is because you chose the portfolio option of "A mixed portfolio consisting of the following assets (based on performance since 1927)". I don't know why, but when you make this choice, the last year that it will use as a starting year is 1975.

If you use the total market option, then it goes from 1871 through 1988.

I believe we have a winner! Makes sense.

-ERD50

 

[Dumello]

This is because you chose the portfolio option of "A mixed portfolio consisting of the following assets (based on performance since 1927)". I don't know why, but when you make this choice, the last year that it will use as a starting year is 1975.

If you use the total market option, then it goes from 1871 through 1988.

Yes! This worked. I chose the total market option and the results are much like I expected.

30 years: 34 failures - 71.2% success rate
35 years: 39 failures - 65.5% success rate
40 years: 45 failures - 58.8% success rate
45 years: 46 failures - 58.3% success rate

Thank you cathy63 and thank you everybody for your help!

 

[Independent]

For me, I only have 49 cycles in 30 years with retirement starting in 2020. It's 34 cycles in 45 years. Not sure why we would be so different.

I'd suggest loading FireCalc and running it with the default assumptions. I'll bet you'l get 118 cycles for 30 years.

Then start inputting your own assumptions one at a time and see which one kicks the number of cycles down.

If you are only getting 34 cycles, you're looking at historic results for retirements beginning in 1871 through 1905.