Anyone withdrawing more than 4%....?

[obryanjf]

Quote:

Originally Posted by FIRE'd@51 Let's assume that the future will be no worse than the past. With a 75/25 equity/fixed mix and a 4% withdrawal rate there is a 5% chance of failure over a 30-year horizon. Alternatively, the success rate is 95%. This means after N resets (assuming fixed 30-year horizons), the probability of success (not running out of money) would be

The math might be right but after that 1 year the new end of the plan is 29 years, in theory. My point is the 'no other income' SWR for 30 to 40 years is around 4%. But not much discussion on 25, 20, 15 yrs SWR rates. Of course we need to be conservative with the terminal age for the calcs, but the SWR is age dependant. As an 80 year old, I doubt I will be using 4%.

job

[pdhenry]

I'm struggling with the mental math on this as well.

On the one hand, you should have a 95% probability of lasting 30 years beginning in year 2 because probability has no memory - the fact that you had a portfolio increase in year 1 doesn't factor into the future.

On the other hand, it's reasonable to assume that many of the scenarios begin with an increase in portfolio value, only to need that increased value to make it through a lean time later on (e.g., 5% of the periods fail. Keep restarting and eventually you'll choose a failing one.) I think this is what FIRE'd@51's math shows.

[Coach]

Quote:

Originally Posted by FIRE'd@51 With respect to the above discussion of resetting the 4% withdrawal rate each year based on the growth/decline of the portfolio: Let's assume that the future will be no worse than the past. With a 75/25 equity/fixed mix and a 4% withdrawal rate there is a 5% chance of failure over a 30-year horizon. Alternatively, the success rate is 95%. This means after N resets (assuming fixed 30-year horizons), the probability of success (not running out of money) would be (0.95)^N N=5 => a success rate of 77% N=10 => a success rate of 60% N=13 => a success rate of 51% I think this illustrates that resetting to 4% of the current portfolio every year does indeed increase the risk of failure, and that increase is larger than one might think. Intuitively, what one is doing is "searching" for that "bad time" to start one's retirement cycle.

Fire'd, I took Probability and Stats a long time ago, but I don't think your logic is right. If there were a 5% chance of failure in each year, your math would hold.

The odds of four heads in a row in a coin toss is (.5)^4, but this is a different case.

Coach

[Sam]

Quote:

Originally Posted by SecondCor521 However, the much more likely scenario is that you won't retire into that worst case scenario, ...

Agreed.

But let's take a look at one of the worst case scenarios: You started with 1MM in 1973. After the first year, your nest egg was reduced to only 781K. You believed history is on your side, and continued with 4% inflation adjusted withdrawal. After the second year, the balance was reduced once again to 565K. Two consecutive very bad years now. Do you have the courage to continue with your original plan?

Most likely, you would reset your plan and start with 4% of the new 565K balance. It sounds bad, but I don't think it is. Everything is relative, and everybody is a little (may be a lot) poorer. So relative to people around you, you are still at the same financial level.

All the plannings we do with FIRECalc, spreadsheet, Bernicke, Guyton, etc... are just that, planning. When it's time to put the plan into action, we follow it if and when the condition is right. If not, we modify it to fit the present situation, just like everything we do everyday.

[SecondCor521]

I should clarify a few things.

Whenever I referred to 4% in this thread, I was using that as shorthand for "the 100% safe SWR in FIREcalc using my actual asset allocation, actual (assumed/projected/guessed) longevity, expenses, etc." I can believe y'all didn't think I meant that since that's not what I wrote. :-P

Given that and the premise that the worst I'll face in my retirement is at least as good as the history FIREcalc covers, then I think that using max(last year's withdrawal + CPI, 4% of current balance) is a strategy that works.

If you're using something less than 100%, then FIRE'd@51's statement about increasing your risk of failure with resets makes sense, even if his math is a little off.

Sam, I'm one of those people who wouldn't reset. But I'll also make darn sure that I investigate the 4% research, data, calculations, etc. before I follow the plan. It's probably moot in my case for reasons I won't get into here.

Oh, and ha, I like what you wrote in post #15 on this thread. I hope we all remember that nothing is absolutely safe, and that one of the risks is working too long.

2Cor521

[pdhenry]

A slightly more conservative method than max(last year's withdrawal + CPI, 4% of current balance) might be to use max(first year's withdrawal adjusted to current year CPI, 4% of current balance).

[Sam]

Quote:

Originally Posted by SecondCor521 Sam, I'm one of those people who wouldn't reset. But I'll also make darn sure that I investigate the 4% research, data, calculations, etc. before I follow the plan.

Good for you. I know now I won't be that brave. I know my plan would change if I were to be hit by 1973 and 1974, consecutively.

BTW, what else is there to research/investigate as far as the 4% is concerned? Isn't it already black and white?

[Webzter]

Quote:

Originally Posted by Sam BTW, what else is there to research/investigate as far as the 4% is concerned? Isn't it already black and white?

Depends on if you're h0cus or not.

[SecondCor521]

Quote:

Originally Posted by Sam Good for you. I know now I won't be that brave. I know my plan would change if I were to be hit by 1973 and 1974, consecutively. BTW, what else is there to research/investigate as far as the 4% is concerned? Isn't it already black and white?

I'm not brave. Far from it in fact. But I do respect my own intelligence and math abilities and have confidence in history "rhyming", even if it doesn't repeat itself. I can also be pig-headedly stubborn.

As far as research goes: I know that people make mistakes when they develop software, so mainly I would be checking out all the data to make sure that Dory correctly applied the inflation adjustments, did all the math right, copied the inflation numbers from the original Schiller data correctly, etc. I honestly think his work is extremely solid and most likely defect-free, but I am sort of a Missoura "show-me" kind of person.

2Cor521

[Kwirk]

Quote:

Originally Posted by SecondCor521 At that point, you can simply pretend that you're just starting out your retirement, and begin taking 4% of your larger portfolio.

Increasing your withdrawal rate after 1 or more good years and assuming that your chance of success is unchanged may not be correct. To get a meaningful estimate of your probable success you might run a firecalc-type estimate but starting only with years that follow one (or more) good years. 4% withdrawal starting after good years may not be as safe (historically) as choosing any random year. I wouldn't be surprised if you find that the chance of success increases if you retire following one or more bad years.

Flipping a coin may have no memory of previous trials but I don't think that we can easily assume the same about markets.

[REWahoo]

Quote:

Originally Posted by Kwirk Increasing your withdrawal rate after 1 or more good years and assuming that your chance of success is unchanged may not be correct.

If the logic behind FIRECalc and the 4% withdrawal rate works when starting your withdrawals in year X it should also work for year X+1.

Theoretically (and if you have the nerve), you should be able to reset your withdrawal amount after any and all "up" years (based on a larger portfolio amount) and have the same statistical result/chance of success that you had when you began withdrawing. FIRECalc doesn't know what year it is when you run the program, your first year of withdrawal, your second, third or your 15th...

And don't forget that each year you withdraw from your nest egg is one less year it will need to last...

[Kwirk]

Quote:

Originally Posted by REWahoo If the logic behind FIRECalc and the 4% withdrawal rate works when starting your withdrawals in year X it should also work for year X+1.

This may be true only as long as you run it for all years X and all years X+1, I think. A FIRECalc type analysis may not give the same success frequency when including only years that follow notably good years (or notably bad years) rather than all years.

Quote:

Originally Posted by REWahoo And don't forget that each year you withdraw from your nest egg is one less year it will need to last...

This is true but does it necessarily make up for the possible ill effect of choosing only certain years?


The test is to do a FIRECalc-type analysis by determining the success rate of 4% withdrawals for one less year beginning in only years that follow a year (or 2) with, say, a 20% increase in the S&P. My gut tells me that the success rate will be less than for all years. I don't have the energy to actually do this test but I'm hopeful that this forum does have somebody energetic enough, compulsive enough,....

[REWahoo]

Here is the point I'm trying to make. From the guy behind FIRECalc:

Quote:

Originally Posted by dory36 Here's another way of looking at it, boiled down to a simple case: Say a buddy retired at the end of 2004 with $1.2 mil. We would all agree that the 4% rule says he can take $48k/year. If you retired in 2000 with $1 mil and it grew to $1.2 mil in 2004 as of the same time that your buddy retired, then you are both starting 2005 with $1.2 mil, so your withdrawals can be the same. The fact that you had some prior history doesn't somehow taint your returns for the same investments that pay 4% of $1.2 mil for your buddy.

[Kwirk]

I certainly respect Dory's contributions to the ER community (far greater than I ever hope to make) but there may be other ways of looking at this.

I agree that both friends have the same chance going forward. However, this is no longer a random year. It's now a year following a very good year (20% gain the previous year) and the odds of success may be less (or, heaven forbid, more) than for all years. Additional information lets you make a more informed estimate. My argument here would not apply in a Monte Carlo simulation but I doubt that we can look at market results as completely random from year to following year.

A different way of understanding that this may be a bad approach is to think of Russian roulette (where the results from one trial to the next are presumably random). My chances on the second round may be the same as on the first round but that doesn't imply that it's a healthy idea to plan to take a second or third shot if I don't kill myself with the first.

[jdw_fire]

Quote:

Originally Posted by Kwirk I certainly respect Dory's contributions to the ER community (far greater than I ever hope to make) but there may be other ways of looking at this. I agree that both friends have the same chance going forward. However, this is no longer a random year. It's now a year following a very good year (20% gain the previous year) and the odds of success may be less (or, heaven forbid, more) than for all years. Additional information lets you make a more informed estimate. My argument here would not apply in a Monte Carlo simulation but I doubt that we can look at market results as completely random from year to following year. A different way of understanding that this may be a bad approach is to think of Russian roulette (where the results from one trial to the next are presumably random). My chances on the second round may be the same as on the first round but that doesn't imply that it's a healthy idea to plan to take a second or third shot if I don't kill myself with the first.

Kwirk, you need to recognize that FIRECalc isn't looking a random events, it is looking at worst case scenerios. So if we are looking a 100% success FIRECalc scenerios and we are assuming that the future will be no worse than the past you can restart every year if you wish and have complete success. The key is the 100% success scenerios.

[SecondCor521]

I think what ReWahoo and Dory have said in #31 and #33 on this thread is correct as long as one is considering a 100% safe withdrawal rate. I haven't thought it through for cases for a <100% safe rate.

2Cor521

[Kwirk]

If 4% inflation-adjusted withdrawals represents a 100% success rate using the historical data then the model allows you to increase your withdrawals to a new 4% level ANY year that your inflation adjusted base increases. I would really like that. I don't recall that the success rate is quite 100%. I'm too lazy to look it up (which probably makes me a perfect candidate for ER).

It isn't really Russian roulette if there aren't any bullets.

[unclemick]

Norwegian widow in hard times - previous year's current yield of portfolio and peanut butter sandwiches/el cheap'o travel/trips.

5% variable in good times and party til you puke.

Skip the fancy math.

Heck in my 14 th year of ER - still experimenting(varying spending) - maybe someday I'll get the hang of it. The 4% neatly spreadsheeted didn't work - because of all my messy things - temp work, pension and early SS kicking in a various times, selling and eating some real estate, DRIP plans with mergers, spin off's, cash buyouts, etc. Probably under 4% smeared out/averaged over the years but I never attempted a SWAG and spending had spikes totally independant from inflation.

In theory - having a plan is good. Being a cheap bastard is even better IMHO.

heh heh heh

[Sam]

Quote:

Originally Posted by unclemick 5% variable in good times and party til you puke. Skip the fancy math.

Sounds like an excellent plan to me.

[laurence]

I'm going to be dealing with a tainted scenario, as I'll have a pretty substantial pension that should cover minimum expenses starting at age 55. So I plan on taking 5% (perhaps a little more!) of the total portfolio each year and resetting each year. If it was a good year, vacation to Hawaii, bad year, camping across the west....