SWR is 4% of what?

[Exflyboy]

Just need a clarification here..

I think I am ready to retire with about $1.2M (although only $475k in non retirement accounts.. But there seems to be enough ways to get early payouts without the penalty) at age 52. Planning to wait till April 2014. Am currently working on $10 cell phone plans etc.

The Wife will keep working and we'll make about $45k with her salary and rent income combined.. Plus she gets great health bennies. As $1.2M is suspiciously close to the $45k we will be making it seems prudent to live off the income and keep the stash growing until she is ready to quit herself.

All well and good but the $1.2M was last week.. before the market started to pull back.. Hmm.. Well what if it pulls back 30%... In other words, if 4% is the SWR.. what number does that 4% represent?

Is it meant to be 4% of the smallest value of the stash in any one year?.. I.e if the markets tanked to 50%.. should it be 4% of $600k?

Thanks

Frank

[Louisville]

Good question.

[dragoncar]

This one is a bit of a head trip, but it's the amount you have at time of retirement.

So if you retire 1/1/2014 with $1 million, it's 40k.
If 2014 sucks and you retire 1/1/2015 with $750k, it's 30k.

Historically, the 4% works even in the worst case- retiring right before a crash.  So practically speaking, scenario 2 above is really safe whereas scenario 1 is right at the edge of historically workable withdrawal rates.

But just remember that these are rules of thumb and can't guarantee future results.

[Half-Borg]

Yes 4% is at the time you retire.
If you retire at market tops and the market tanks right away for serveral years, well you're screwed, these are the few percent where the 4% rule did not work. That why some people head for 3% SWR or plan on cutting back in rough times.

[Kira]

The idea is that you set the 4% on the portfolio you have when you retire.

Then hopefully over time you are still taking out the same $, but your total portfolio value has risen, so that you can weather the ups and downs your stash will take over the years. Some years the 4% will actually be 5% of value and some years it will be 3%, etc.

[dragoncar]

Yes 4% is at the time you retire.
If you retire at market tops and the market tanks right away for serveral years, well you're screwed, these are the few percent where the 4% rule did not work. That why some people head for 3% SWR or plan on cutting back in rough times.

I thought 4% always worked within the boundaries of the trinity study (any 30 years of withdrawals from 1926-1995, under 75% equities)

This includes the Great Depression

[SnackDog]

4% of your intended retirement cashpile may be withdrawn the first year. Every year thereafter, you may increase the withdrawal dollar amount by inflation.   Studies of the US stock market and inflation rates the last 100 years or so suggest you have a high chance of your money lasting 30 years with a broad indexed investment of about 70/30 stocks and bonds.

Future market and economic behavior could be different. In particular, if the market tanks or lags inflation rates severely your first five years or so you may need to reset the 4% to match your new lower principle. Or go back to work until conditions improve.

Good luck!

[pom]

I use 3% for my planning myself.

The trinity study is only a sample of 70 years, those are not independent trials (for exemple 1926-1956 is hugely correlated with 1927-1957)

http://en.wikipedia.org/wiki/Trinity_study

There are similar studies in other countries that result in much lower safe withdrawal rates.

That being said, 4% is probably safe enough.

[dragoncar]

I use 3% for my planning myself.

The trinity study is only a sample of 70 years, those are not independent trials (for exemple 1926-1956 is hugely correlated with 1927-1957)

http://en.wikipedia.org/wiki/Trinity_study

There are similar studies in other countries that result in much lower safe withdrawal rates.

That being said, 4% is probably safe enough.

Where are you getting "not independent trials from?"  You may be right, it just doesn't coincide with anything I've read so far.

Of course fire calc will be a more accurate tool for back testing any strategy, but I still think 4% is a good rule of thumb

[pom]

Sorry for the late answer.

Take my exemple: the pattern for the years 26-56 is hugely correlated with 27-57. Obviously 29 of the years are exactly the same.

Lets assume that you find the single withdrawal rate that give you 50-50 chances of failing and you test it on the samples of the trinity study.

If the samples have no correlations, over a 70 trials you should see exactly 6 fails in a row (this serie: "...PFFFFFFP...") only one time (70x0,5^6 = 1), then you should see five fails in a row one time as well (70x0,5^5 - 70x0,5^6 = 1), 4 times should appear twice (70x0,5^4 - 70x0,5^5 = 2) ... of course this being statistics, it will never happen exacly as predicted but it should not be too far.

That is not what you will see in this study because if 26-56 fails, the chances of 27-57 failing is more than 50% because it is not independant. If you have only 70 years and you want 30 years samples, you can only create 2 truly independant samples (and lets not get into serial-correlation because that is another problem with most studies of the. stock market)

I hope this helps.

[grantmeaname]

But why do you think that matters? If you retire in 2021 and I retire in 2022 I don't get my returns because then our retirement portfolio is partially correlated? No? Then why does the statistical behavior of the sample matter?

(Also, the trinity study has been corroborated over longer periods, so it's not like it's just 70 years anymore.)

[dragoncar]

Thanks, I now understand what you are saying but agree with grant that it doesn't matter.  At least it doesn't matter any more than the fact that future returns could be terrible and are not obligated to mimic the past.

In other words, the value of the trinity study is just to say that 4% has always worked in the past (for certain asset allocations during the long period studied).

[pom]

Agreed that there is value in saying that it worked in the past.

Why it matters? Correlation matters lot to a statistician, it means that the distribution of sample results is unlikely to follow a normal curve and it greatly reduces the predicting power of the study results.

Anyway, I was just explaining why there is correlation. Correlation does not imply that the results are necessarily wrong, just that one should take extra care in assessing the results.

[grantmeaname]

Why it matters? Correlation matters lot to a statistician, it means that the distribution of sample results is unlikely to follow a normal curve and it greatly reduces the predicting power of the study results.

Right, I understand statistical inference. But it doesn't qualitatively change anything about the results and the general result of the study has been upheld in an expansion to a longer period (e.g., FIREcalc/cFIREsim), so I'm asking why it matters.