Safe Withdrawal Rate Timing Problem

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
dspencer
Posts: 225
Joined: Wed Jul 06, 2016 11:29 am

Safe Withdrawal Rate Timing Problem

Post by dspencer »

I am wondering if there is a bias built into planning retirement based on historical success probabilities. For example, let's say that you run the numbers and determine that at a 4% WR you will run out of money in only 5% of historical 30 year periods. You decide that this is an acceptable risk and that you need $1 million to retire. Once you hit that number you retire. Even assuming that historical returns are a perfect way of predicting the future, do you really have a 95% chance of success?

My thought is that you don't, because you are unlikely to retire after a period of poor market performance. Assuming your savings contributions are small relative to your overall investment goal, you will only hit a benchmark like $1 million if your investment returns have been positive. Wouldn't this skew your odds so that your real chance of success is lower than a randomly selected 30 year period? For example, you're more likely to hit $1MM in 2007 than in 2009, but your odds of success with the same amount of money are better if you retire in 2009.
livesoft
Posts: 74574
Joined: Thu Mar 01, 2007 8:00 pm

Re: Safe Withdrawal Rate Timing Problem

Post by livesoft »

You are right and everybody agrees with you. So what are you gonna do?
Wiki This signature message sponsored by sscritic: Learn to fish.
User avatar
siamond
Posts: 5732
Joined: Mon May 28, 2012 5:50 am

Re: Safe Withdrawal Rate Timing Problem

Post by siamond »

For early retirees, what you describe is probably especially true. For more regular retirees, they may not choose their date of retirement, so maybe this is less true.

Anyhoo, do you really think it is acceptable to play roulette and have a 5% (or 10% or whatever) chance of ruin? I don't.

Nobody in their right mind should use a fixed withdrawal method when retired. This is simply not adaptive enough (downsides AND upsides). And once you start exploring the idea of variable withdrawal methods, then you start to understand that this SWR metric isn't as significant as many researchers make of it... It is informative, mind you, it but just shouldn't be the cornerstone of a retirement plan. It's simple to compute and to model, but real life is more complicated and human being are more adaptive creatures than that.
sport
Posts: 9857
Joined: Tue Feb 27, 2007 3:26 pm
Location: Cleveland, OH

Re: Safe Withdrawal Rate Timing Problem

Post by sport »

siamond is exactly correct. Some years, things will happen and you will need more than 4%. Therefore, you must plan on using less than 4% in other years.
grok87
Posts: 9169
Joined: Tue Feb 27, 2007 9:00 pm

Re: Safe Withdrawal Rate Timing Problem

Post by grok87 »

dspencer wrote:I am wondering if there is a bias built into planning retirement based on historical success probabilities. For example, let's say that you run the numbers and determine that at a 4% WR you will run out of money in only 5% of historical 30 year periods. You decide that this is an acceptable risk and that you need $1 million to retire. Once you hit that number you retire. Even assuming that historical returns are a perfect way of predicting the future, do you really have a 95% chance of success?

My thought is that you don't, because you are unlikely to retire after a period of poor market performance. Assuming your savings contributions are small relative to your overall investment goal, you will only hit a benchmark like $1 million if your investment returns have been positive. Wouldn't this skew your odds so that your real chance of success is lower than a randomly selected 30 year period? For example, you're more likely to hit $1MM in 2007 than in 2009, but your odds of success with the same amount of money are better if you retire in 2009.
Another good reason to follow bernstein and divide your retirement portfolio into 2 pieces:

I) A liability matching portfolio (LMP) in duration matched tips to guarantee a minimum floor of retirement income.

2) a risk portfolio (rp) that is mostly in stocks, real estate, etc with some dry powder for rebalancing.

Cheers,
Grok
RIP Mr. Bogle.
User avatar
TheTimeLord
Posts: 8781
Joined: Fri Jul 26, 2013 2:05 pm

Re: Safe Withdrawal Rate Timing Problem

Post by TheTimeLord »

dspencer wrote:I am wondering if there is a bias built into planning retirement based on historical success probabilities. For example, let's say that you run the numbers and determine that at a 4% WR you will run out of money in only 5% of historical 30 year periods. You decide that this is an acceptable risk and that you need $1 million to retire. Once you hit that number you retire.
I hit my number 2 years ago, that is not the sole determinant of when to retire. Personally, if we enter a Bear market I will probably work several more months until things are working again, checking out at the top of the market would worry me. Thus the reason I am so focused on risk management.
IMHO, Investing should be about living the life you want, not avoiding the life you fear. | Run, You Clever Boy! [9085]
User avatar
Watty
Posts: 21395
Joined: Wed Oct 10, 2007 3:55 pm

Re: Safe Withdrawal Rate Timing Problem

Post by Watty »

The "4% rule" is more of a guideline than a firm rule.

In looking at it as a firm rule you also have to consider people that retire a year apart but in that year their portfolios go up or down by 10%. It it was a hard rule then one person would get to spend 10% more than the other
User avatar
DG99999
Posts: 205
Joined: Fri May 30, 2014 12:30 pm

Re: Safe Withdrawal Rate Timing Problem

Post by DG99999 »

I favor the original definition of SWR which would represent a maximum rate that did not fail during any previous historic period (100% success). A definition from Kitces reads:

"The origin of the safe withdrawal rate was actually rather straightforward – it’s simply the initial withdrawal rate that would have sustained inflation-adjusted spending in the worst case scenario in (US) history."

So I guess I would quibble with your title, since by introducing a 5% probability of failure, I would not consider the result a SWR (though I know it is common in the literature and simulations). The upshot is that if you wish to retire at the earliest possible moment, then it might be best to make the calculation using the old/100% success rate.
I am not a financial professional. My posts are only my opinion on the topic. You need to do your own due diligence and consult with a professional when addressing your financial questions.
User avatar
snowshoes
Posts: 348
Joined: Tue Feb 03, 2015 12:33 pm

Re: Safe Withdrawal Rate Timing Problem

Post by snowshoes »

A 0%-4% VWR is certainly recognized and seems to come out on top again and again if one has their base costs covered.
kolea
Posts: 1322
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Safe Withdrawal Rate Timing Problem

Post by kolea »

dspencer wrote:I am wondering if there is a bias built into planning retirement based on historical success probabilities. For example, let's say that you run the numbers and determine that at a 4% WR you will run out of money in only 5% of historical 30 year periods. You decide that this is an acceptable risk and that you need $1 million to retire. Once you hit that number you retire. Even assuming that historical returns are a perfect way of predicting the future, do you really have a 95% chance of success?
You pose an interesting question, if I am understanding it correctly. I believe you are thinking the market is roughly cyclical and that most people will tend to retire in a bull market, but due to the cyclical nature, a bear is more likely to follow within a short period of the time of your retirement. And if so, we are more likely to encounter sequence-of-return risk than if we had just picked a random year to retire. I think that is a very valid concern.

One resolution, on the testing front, is to run a Trinity (or Bengen) style analysis using historical data, but rather than start it at year-0 of retirement, start it at year-0 of saving and run it all the way through accumulation and then through retirement. The back test would have to trigger when retirement took place by some combination of age and whether the magic number has been reached. But since the retirement criteria would more likely be reached in a bull market, the back-test would be testing how we actually do tend to retire and should remove the bias.

The difficulty with this back-test is that now we are talking 30 years of accumulation + 30 years of retirement. It is hard enough to get good stats on just 30 years periods and it will be worse with 60 year test periods. But I believe that would be the proper way to do it. Maybe I will dust off my back-test app and try it. :)

The resolution on the actual retirement front is, as others have said above, have a flexible plan.
Kolea (pron. ko-lay-uh). Golden plover.
AlohaJoe
Posts: 5672
Joined: Mon Nov 26, 2007 2:00 pm
Location: Saigon, Vietnam

Re: Safe Withdrawal Rate Timing Problem

Post by AlohaJoe »

kolea wrote:One resolution, on the testing front, is to run a Trinity (or Bengen) style analysis using historical data, but rather than start it at year-0 of retirement, start it at year-0 of saving and run it all the way through accumulation and then through retirement. The back test would have to trigger when retirement took place by some combination of age and whether the magic number has been reached. But since the retirement criteria would more likely be reached in a bull market, the back-test would be testing how we actually do tend to retire and should remove the bias.
It isn't exactly what you're suggesting by a 2011 article in the Journal of Financial Planning did something much like this by looking at a 30-year accumulation period followed by a 30-year retirement period.

Image

Moving right along the chart means "we needed to save more every year in order to reach our retirement goal". You can see that it goes from needing to save around 11% of income up to around 37% of income every single year for one's entire career. You have to save a lot because the market isn't doing great so a lot of the growth of your assets comes from savings rather than capital appreciation.

Moving up along the chart means "once you retire, you're able to withdraw this much money every year". You are able to withdraw more income because the market does well and capital appreciation means your nest egg grows even during retirement.

The line slopes up and the to right showing: when you are in accumulation phase you have to save a lot because the market sucks. But once you retire you're able to withdraw a lot because the market rebounds. And the opposite is also true: if you are able to meet your retirement goals easily, thanks to good returns, then your retirement is more likely to have lower returns.
lazyday
Posts: 3798
Joined: Wed Mar 14, 2007 10:27 pm

Re: Safe Withdrawal Rate Timing Problem

Post by lazyday »

Probably not enough independent data to be very useful, but it might be interesting to see a chart with X axis = real expected return of a 60/40 portfolio, Y axis = SWR, each retirement year is a point on the chart.

If data is available for other countries, show them on the same chart using different colored points.

Expected return of stocks could be something like 1/CAPE, expected return of bonds… maybe something like nominal rate - prior 5 year annualized inflation, or whatever’s been found to make sense for a time before TIPS.
aristotelian
Posts: 8634
Joined: Wed Jan 11, 2017 8:05 pm

Re: Safe Withdrawal Rate Timing Problem

Post by aristotelian »

If you hit your number but then the market drops such that you would be forced to work more, then you never really hit your number.

Your number should be big enough that you can sustain yourself with a conservative portfolio that isn't affected by a stock market drop, or such that a stock market drop doesn't concern you even with an aggressive allocation.
wolf359
Posts: 2253
Joined: Sun Mar 15, 2015 8:47 am

Re: Safe Withdrawal Rate Timing Problem

Post by wolf359 »

The solution is simple. If the market is going to crash after you retire, don't retire. :confused

More seriously, build flexibility into your plan.

- If the market crashes, be able to cut back on your spending.
- Be willing to earn income after retiring. Earning as little as $10,000 in a year can prevent you from selling equities in a down year.
- Plan your expenses such that you can survive on a 3.5% SWR.

The 4% Rule is just that -- a Rule of Thumb. It's a guideline for knowing when you have enough. The math of probabilities is false precision. All the probabilities are based on backtesting. What happened in the past might not happen in the future.

If you have Social Security, or pensions, or other assets, they're not accounted for by the 4% Rule (and they improve your chances.)

It takes a long time to build up the assets to retire. If you get into the ballpark and can early retire, I'd suggest building up a little slack to allow for the unknown before pulling the trigger. If you're forced to retire, then you just have to deal with the hand you're dealt.
Topic Author
dspencer
Posts: 225
Joined: Wed Jul 06, 2016 11:29 am

Re: Safe Withdrawal Rate Timing Problem

Post by dspencer »

kolea wrote:
dspencer wrote:I am wondering if there is a bias built into planning retirement based on historical success probabilities. For example, let's say that you run the numbers and determine that at a 4% WR you will run out of money in only 5% of historical 30 year periods. You decide that this is an acceptable risk and that you need $1 million to retire. Once you hit that number you retire. Even assuming that historical returns are a perfect way of predicting the future, do you really have a 95% chance of success?
You pose an interesting question, if I am understanding it correctly. I believe you are thinking the market is roughly cyclical and that most people will tend to retire in a bull market, but due to the cyclical nature, a bear is more likely to follow within a short period of the time of your retirement. And if so, we are more likely to encounter sequence-of-return risk than if we had just picked a random year to retire. I think that is a very valid concern.

One resolution, on the testing front, is to run a Trinity (or Bengen) style analysis using historical data, but rather than start it at year-0 of retirement, start it at year-0 of saving and run it all the way through accumulation and then through retirement. The back test would have to trigger when retirement took place by some combination of age and whether the magic number has been reached. But since the retirement criteria would more likely be reached in a bull market, the back-test would be testing how we actually do tend to retire and should remove the bias.

The difficulty with this back-test is that now we are talking 30 years of accumulation + 30 years of retirement. It is hard enough to get good stats on just 30 years periods and it will be worse with 60 year test periods. But I believe that would be the proper way to do it. Maybe I will dust off my back-test app and try it. :)

The resolution on the actual retirement front is, as others have said above, have a flexible plan.
Would it be necessary to include 30 years of accumulation? Obviously it depends on savings rates but most people would not be close until the last few years. If you are able to figure anything out let me know.

For me personally it is more curiosity than necessity. My plan is to be flexible but also to have a large buffer of luxury between my planned lifestyle and eating cat food. It seems that some people really do pull the trigger with say $800k invested for an early retirement.
Topic Author
dspencer
Posts: 225
Joined: Wed Jul 06, 2016 11:29 am

Re: Safe Withdrawal Rate Timing Problem

Post by dspencer »

aristotelian wrote:If you hit your number but then the market drops such that you would be forced to work more, then you never really hit your number.

Your number should be big enough that you can sustain yourself with a conservative portfolio that isn't affected by a stock market drop, or such that a stock market drop doesn't concern you even with an aggressive allocation.
Doesn't that require knowing the future or at least being extremely wealthy? If your number is $1 million then does it mean that if you think your portfolio could decrease by 1/3 in one year then you really need $1.5 million?

This is the train of thought that led me to question how people are using the simulations. It seems like even a modest drop, for example 10%, would lead to a lot of people realizing that if they re-ran the calculation that led them to retire with a 95% success rate, they would now see a much lower number. If you now think the chance is 80%, do you go back to work? As many have said, the best approach is being flexible.
aristotelian
Posts: 8634
Joined: Wed Jan 11, 2017 8:05 pm

Re: Safe Withdrawal Rate Timing Problem

Post by aristotelian »

dspencer wrote:
Doesn't that require knowing the future or at least being extremely wealthy? If your number is $1 million then does it mean that if you think your portfolio could decrease by 1/3 in one year then you really need $1.5 million?

This is the train of thought that led me to question how people are using the simulations. It seems like even a modest drop, for example 10%, would lead to a lot of people realizing that if they re-ran the calculation that led them to retire with a 95% success rate, they would now see a much lower number. If you now think the chance is 80%, do you go back to work? As many have said, the best approach is being flexible.
You don't have to know the future necessarily, but you need to be conservative when estimating your number. If you are wrong, then what you thought was your number was not really your number. The point of a financial independence number is to get to a point that you have at least a reasonable degree of certainty that you are now set for life. If you could not handle a 50% crash, then you are not truly financially independent, and you are taking a calculated risk that the worst case scenario won't happen.

I would say that if you need to invest $1M aggressively in order to sustain it, then your true number is probably higher.
Last edited by aristotelian on Wed Mar 22, 2017 2:16 pm, edited 1 time in total.
sport
Posts: 9857
Joined: Tue Feb 27, 2007 3:26 pm
Location: Cleveland, OH

Re: Safe Withdrawal Rate Timing Problem

Post by sport »

dspencer wrote:Doesn't that require knowing the future or at least being extremely wealthy? If your number is $1 million then does it mean that if you think your portfolio could decrease by 1/3 in one year then you really need $1.5 million?
If you have a 40/60 allocation, it is most unlikely that your portfolio could decrease by 1/3 in one year. If you have a 30/70 allocation, it is even less likely.
Topic Author
dspencer
Posts: 225
Joined: Wed Jul 06, 2016 11:29 am

Re: Safe Withdrawal Rate Timing Problem

Post by dspencer »

sport wrote:
dspencer wrote:Doesn't that require knowing the future or at least being extremely wealthy? If your number is $1 million then does it mean that if you think your portfolio could decrease by 1/3 in one year then you really need $1.5 million?
If you have a 40/60 allocation, it is most unlikely that your portfolio could decrease by 1/3 in one year. If you have a 30/70 allocation, it is even less likely.
True, but if your expected returns are lower due to a conservative AA then you need more money to start with anyway.
aristotelian
Posts: 8634
Joined: Wed Jan 11, 2017 8:05 pm

Re: Safe Withdrawal Rate Timing Problem

Post by aristotelian »

dspencer wrote: True, but if your expected returns are lower due to a conservative AA then you need more money to start with anyway.
Exactly. If you can't achieve your income with a conservative allocation, then you have not really hit your number. You have hit a minimum, somewhat arbitrary, number that gives you a good chance.
itstoomuch
Posts: 5343
Joined: Mon Dec 15, 2014 12:17 pm
Location: midValley OR

Re: Safe Withdrawal Rate Timing Problem

Post by itstoomuch »

IMO, if a person has considerable IRA assets, the issues for RMD become larger than a SWR.
Your RMD will exceed a 4% swr in your early 70s and there after. You have to "time" your RMD into a taxable account when your IPS/Otar/Firecal/Monte Carlo says you should only do a SWR amount.

YMMV :beer
Rev012718; 4 Incm stream buckets: SS+pension; dfr'd GLWB VA & FI anntys, by time & $$ laddered; Discretionary; Rentals. LTCi. Own, not asset. Tax TBT%. Early SS. FundRatio (FR) >1.1 67/70yo
Post Reply