Author Topic: Pfau on Variable Withdrawal Strategies  (Read 16232 times)

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Pfau on Variable Withdrawal Strategies
« on: March 17, 2015, 09:56:22 AM »
But still, we'll continue to reference that hoary old research. Why? Because the guys that did that hoary old research had fancy letters before or after their names, and wrote their stuff in a serif font on two-columned dead-tree paper (now PDFs) rather than on the wild-and-crazy World Wide Web. It's just dumb how a forum like this is filled with people every bit as smart or even smarter than those guys, who have done as much or more work and research and insight-creation, and we fail to recognize that.

Pfau has a new research paper out surveying the literature on variable withdrawal strategies, as summarized and linked to in his latest post on his blog.

The "literature" surveyed in his academic article includes the bogleheads forum thread on the "variable percentage withdrawal" strategy.  Pfau often gives due acknowledgement to the good work being done in the early retirement blogosphere, but seeing a citation to a bogleheads forum thread thrown in among the litany of italicized bluebook citations to various Journals of Hoary Academic Research on Financial Planning in the article's endnotes reminded me of this thread and made me smile.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #1 on: March 17, 2015, 10:19:04 AM »
But still, we'll continue to reference that hoary old research. Why? Because the guys that did that hoary old research had fancy letters before or after their names, and wrote their stuff in a serif font on two-columned dead-tree paper (now PDFs) rather than on the wild-and-crazy World Wide Web. It's just dumb how a forum like this is filled with people every bit as smart or even smarter than those guys, who have done as much or more work and research and insight-creation, and we fail to recognize that.

Pfau has a new research paper out surveying the literature on variable withdrawal strategies, as summarized and linked to in his latest post on his blog.

The "literature" surveyed in his academic article includes the bogleheads forum thread on the "variable percentage withdrawal" strategy.  Pfau often gives due acknowledgement to the good work being done in the early retirement blogosphere, but seeing a citation to a bogleheads forum thread thrown in among the litany of italicized bluebook citations to various Journals of Hoary Academic Research on Financial Planning in the article's endnotes reminded me of this thread and made me smile.

Thanks for the link!  Just downloaded the paper, will dive into it later.
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

skyrefuge

  • Handlebar Stache
  • *****
  • Posts: 1015
  • Location: Suburban Chicago, IL
Re: Pfau on Variable Withdrawal Strategies
« Reply #2 on: March 17, 2015, 10:28:25 AM »
The "literature" surveyed in his academic article includes the bogleheads forum thread on the "variable percentage withdrawal" strategy.  Pfau often gives due acknowledgement to the good work being done in the early retirement blogosphere, but seeing a citation to a bogleheads forum thread thrown in among the litany of italicized bluebook citations to various Journals of Hoary Academic Research on Financial Planning in the article's endnotes reminded me of this thread and made me smile.

Ah, yeah, that is cool and good!

Of course such acknowledgement now might just increase the fawning, finger-down-my-throat vomit-inducing references that Bogleheads make to "Dr. Pfau" (and similarly "Dr. Bernstein"). *gag* *hack* *gag*

From the Not-Exactly-Rigorous-Science-Department (don't worry, that's ok since I don't have "Dr." in front of my name):

Google's ratio of "Dr. Pfau" to "Pfau" references at bogleheads.org: 45/270 = 17%
Google's ratio of "Dr. Pfau" to "Pfau" references at mrmoneymustache.com: 4/117 = 3%

Given MMM's lack of respect for fancy titles and institutions, it's not surprising that his followers show the same inclinations, but it's fun to see it supported with data!

And suddenly I realize that's why there are so many MMM-skeptics at the Bogleheads forum: it's because MMM explicitly reveals himself to be an unworthy "Mr.", right in his URL! If he had simply declared himself Dr. Money Mustache, he would be their undisputed hero!

Mississippi Mudstache

  • Handlebar Stache
  • *****
  • Posts: 2170
  • Age: 40
  • Location: Danielsville, GA
    • A Riving Home - Ramblings of a Recusant Woodworker
Re: Pfau on Variable Withdrawal Strategies
« Reply #3 on: March 17, 2015, 10:37:45 AM »
And suddenly I realize that's why there are so many MMM-skeptics at the Bogleheads forum: it's because MMM explicitly reveals himself to be an unworthy "Mr.", right in his URL! If he had simply declared himself Dr. Money Mustache, he would be their undisputed hero!

Dr. Dollar Van Dyke? Kinda catchy.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #4 on: March 17, 2015, 11:07:54 AM »
The "literature" surveyed in his academic article includes the bogleheads forum thread on the "variable percentage withdrawal" strategy.  Pfau often gives due acknowledgement to the good work being done in the early retirement blogosphere, but seeing a citation to a bogleheads forum thread thrown in among the litany of italicized bluebook citations to various Journals of Hoary Academic Research on Financial Planning in the article's endnotes reminded me of this thread and made me smile.

Ah, yeah, that is cool and good!

Of course such acknowledgement now might just increase the fawning, finger-down-my-throat vomit-inducing references that Bogleheads make to "Dr. Pfau" (and similarly "Dr. Bernstein"). *gag* *hack* *gag*

From the Not-Exactly-Rigorous-Science-Department (don't worry, that's ok since I don't have "Dr." in front of my name):

Google's ratio of "Dr. Pfau" to "Pfau" references at bogleheads.org: 45/270 = 17%
Google's ratio of "Dr. Pfau" to "Pfau" references at mrmoneymustache.com: 4/117 = 3%

Given MMM's lack of respect for fancy titles and institutions, it's not surprising that his followers show the same inclinations, but it's fun to see it supported with data!

And suddenly I realize that's why there are so many MMM-skeptics at the Bogleheads forum: it's because MMM explicitly reveals himself to be an unworthy "Mr.", right in his URL! If he had simply declared himself Dr. Money Mustache, he would be their undisputed hero!

I think the bogleheads people are just more academic than here, percentage-wise.  We have just as smart of posters, but we also have a huge chunk of young, non-academic posters.  So when you do a percent, they skew higher.  I could be wrong, but that's my impression on both.
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #5 on: March 17, 2015, 11:14:03 AM »
Given MMM's lack of respect for fancy titles and institutions, it's not surprising that his followers show the same inclinations, but it's fun to see it supported with data!

Great observation, which leads me to question why the bogleheads forum has no poster-ranking system while our own, with its more democratic ethos, does?

- Senior Mustachian Brooklynguy, Esq.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #6 on: March 17, 2015, 12:12:32 PM »
Given MMM's lack of respect for fancy titles and institutions, it's not surprising that his followers show the same inclinations, but it's fun to see it supported with data!

Great observation, which leads me to question why the bogleheads forum has no poster-ranking system while our own, with its more democratic ethos, does?

For funsies!

This is something I can now tweak as an admin (i.e. add more to the scale).  Post suggestions here: http://forum.mrmoneymustache.com/off-topic/mustache-levels/

:)
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #7 on: March 17, 2015, 12:37:13 PM »
But still, we'll continue to reference that hoary old research. Why? Because the guys that did that hoary old research had fancy letters before or after their names, and wrote their stuff in a serif font on two-columned dead-tree paper (now PDFs) rather than on the wild-and-crazy World Wide Web. It's just dumb how a forum like this is filled with people every bit as smart or even smarter than those guys, who have done as much or more work and research and insight-creation, and we fail to recognize that.

Pfau has a new research paper out surveying the literature on variable withdrawal strategies, as summarized and linked to in his latest post on his blog.

The "literature" surveyed in his academic article includes the bogleheads forum thread on the "variable percentage withdrawal" strategy.  Pfau often gives due acknowledgement to the good work being done in the early retirement blogosphere, but seeing a citation to a bogleheads forum thread thrown in among the litany of italicized bluebook citations to various Journals of Hoary Academic Research on Financial Planning in the article's endnotes reminded me of this thread and made me smile.

Thanks for the link!  Just downloaded the paper, will dive into it later.

Okay, that was neat, but tough to sum up.

I think my favorite way to approach it is decide what you want your floor to be in real spending at the 10th percentile, and look at that column.

I.e. starting w/ 1MM, you can withdraw 28.8k - 2.88% SWR - on the typical 4% SWR rule of inflation adjusted, and that's your constant spending.  If you're willing to have a low of 10k real, Zolt may be acceptable, etc.

For me, I think my floor is 15k annual, or 1500 on this chart that uses 1/10th of those numbers, at the 10% level.  Under that, the RMD strategy is the highest initial WR that hits that, at only 3.23%

Scary.  I'd have thought variable WRs would have added a half percent or percent to one's WR.

The climate is so tough, I suppose.

We always say though "well 4% flat may not work, but if you adjust your spending you should be fine" -- here it shows you'd need a 2.88% SWR, and even if you're willing to adjust spending you'll probably need to be sub-4%, unless you're okay with the spending going REALLY low (i.e. under 1% of the initial stache, or less than 1/4th of your initial spend level).

Bengen's Floor and Ceiling, at 3.33% WR, actually look the most appealing to me.   But yikes if you're in the bottom 10% and your real wealth is cut from 1MM to 5k after 30 years.  GG.  Those of us with significantly longer timelines than 30 years (perhaps even double that) may need a lower WR than we've been thinking.  The "be flexible" we mention may not be quite enough, it may be "start with a lower WR and be flexible"

Of course, I'll probably have a more aggressive AA than 50/50 also.
« Last Edit: March 17, 2015, 12:39:55 PM by arebelspy »
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #8 on: March 17, 2015, 01:05:36 PM »
Of course, I'll probably have a more aggressive AA than 50/50 also.

Yep, this was the key takeaway for me.  It's not that easy to determine whether you've accumulated enough to have "won the game," so most of us will need a pretty aggressive allocation unless we are willing to work much longer than what should be necessary or accept the possibility of having to drastically cut spending (or seek alternative income) at certain points in retirement.

I'd like to see the figures for these tables with something closer to a 90/10 stock/bond allocation.

EDIT:  Maybe we should break this discussion out into a new thread, because there's plenty of meat to chew on in the Pfau article.
« Last Edit: March 17, 2015, 01:07:28 PM by brooklynguy »

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #9 on: March 17, 2015, 02:24:49 PM »
Bengen's Floor and Ceiling, at 3.33% WR, actually look the most appealing to me.   But yikes if you're in the bottom 10% and your real wealth is cut from 1MM to 5k after 30 years.  GG.  Those of us with significantly longer timelines than 30 years (perhaps even double that) may need a lower WR than we've been thinking.  The "be flexible" we mention may not be quite enough, it may be "start with a lower WR and be flexible"

It's important to note that the methodology Pfau used for the paper (Monte Carlo simulations with the capital market expectations described in the appendix to the article) produced significantly more pessimistic results than actual historical performance would predict.

If you run simulations using the same parameters in cfiresim based on historical data, the results are substantially better.  For the Bengen Floor and Ceiling spending plan referencedin your quote above (using the same parameters as Pfau's paper -- 3.33% initial WR, 50/50 allocation, 0.50% expense ratio, spending band of 15% below real value of first year's withdrawal and 20% above real value of first year's withdrawal), the median ending value after 30 years using historical data is $78,091 (as opposed to $69,270 using Pfau's methodology).  Cfiresim doesn't tell you the 10th percentile figure, but the absolute worst case ending value (i.e., bottom <1st percentile) is $7,823, which is more than fifteen times higher than Pfau's $500 figure for the bottom 10th percentile.

For the straight constant inflation-adjusted spending plan (the normal non-variable SWR spending approach we usually talk about), the history-based figures are even more proportionally optimistic -- median ending value of $116,850 (compared to Pfau's $78,070) and worst-case (bottom <1st percentile) of $22,438 (compared to $510 for Pfau's bottom 10th percentile).

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #10 on: March 17, 2015, 03:01:24 PM »
I noticed that's how he did it, but the implications didn't click.  Thanks for the clarification.

Well that's a bummer.  I wonder what it each of those strategies (and the corresponding chart) would look like with historical data.
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #11 on: March 17, 2015, 03:37:37 PM »
Cfiresim lets you model at least some of the variable spending strategies Pfau discussed (and others which he didn't), so it shouldn't take too much effort to produce similar tables with partially complete figures using historical results if someone less lazy than me is willing to volunteer to do so.  But the main problem is that cfiresim doesn't give all the data needed to compile the same charts.

I just left a comment on Pfau's blog about this to get his thoughts (but I suspect he will say that he doesn't believe we can count on performance in the coming decades to repeat the results of the past century).

bearkat

  • Stubble
  • **
  • Posts: 120
  • Age: 34
  • Location: California
Re: Pfau on Variable Withdrawal Strategies
« Reply #12 on: March 17, 2015, 05:55:28 PM »
Bengen's Floor and Ceiling, at 3.33% WR, actually look the most appealing to me.   But yikes if you're in the bottom 10% and your real wealth is cut from 1MM to 5k after 30 years.  GG.  Those of us with significantly longer timelines than 30 years (perhaps even double that) may need a lower WR than we've been thinking.  The "be flexible" we mention may not be quite enough, it may be "start with a lower WR and be flexible"

It's important to note that the methodology Pfau used for the paper (Monte Carlo simulations with the capital market expectations described in the appendix to the article) produced significantly more pessimistic results than actual historical performance would predict.

If you run simulations using the same parameters in cfiresim based on historical data, the results are substantially better.  For the Bengen Floor and Ceiling spending plan referencedin your quote above (using the same parameters as Pfau's paper -- 3.33% initial WR, 50/50 allocation, 0.50% expense ratio, spending band of 15% below real value of first year's withdrawal and 20% above real value of first year's withdrawal), the median ending value after 30 years using historical data is $78,091 (as opposed to $69,270 using Pfau's methodology).  Cfiresim doesn't tell you the 10th percentile figure, but the absolute worst case ending value (i.e., bottom <1st percentile) is $7,823, which is more than fifteen times higher than Pfau's $500 figure for the bottom 10th percentile.

For the straight constant inflation-adjusted spending plan (the normal non-variable SWR spending approach we usually talk about), the history-based figures are even more proportionally optimistic -- median ending value of $116,850 (compared to Pfau's $78,070) and worst-case (bottom <1st percentile) of $22,438 (compared to $510 for Pfau's bottom 10th percentile).

So I read the paper, and WOW do things look bleak. 2.88% WR and still a 10% chance of basically having no money after 30 years.

Some questions:

How much would we expect a more aggressive AA to change the results? I basically never see a 50/50 AA on these boards.

Is there a MMM forum consensus on Monte Carlo vs cfiresim? From brooklynguy's analysis above above, I'm seeing a major deviation between the two.

Any thoughts on how early a retiree would be able to recognize if he's in the bottom 10% of possibilities? 2 years? 5 years? 20 years? Never because no one has a properly working crystal ball.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #13 on: March 17, 2015, 08:42:13 PM »
So I read the paper, and WOW do things look bleak. 2.88% WR and still a 10% chance of basically having no money after 30 years.

Some questions:

I'm going to give my answers to your questions in a different order than you asked them.

Quote
Is there a MMM forum consensus on Monte Carlo vs cfiresim? From brooklynguy's analysis above above, I'm seeing a major deviation between the two.

I prefer cfiresim using actual historical data (note that cfiresim also has a Monte Carlo simulator).  The problem with most Monte Carlo simulators (like the one Pfau used) is that they are only as good as the assumptions they make about the future (the "garbage in, garbage out" problem).  Vanguard's Monte Carlo simulator doesn't really have this problem because it uses actual historical results, but just randomizes the sequence of returns.  But I still think cfiresim's straight actual historical results are the best we have to go on -- it is impossible to predict the future, and the best we can do is look at the past (looking at randomly-generated Monte Carlo results can be an informative supplement, but if I were forced to chose one tool I would go with actual historical data).

Using cfiresim's standard history-based calculator (with identical parameters to what Pfau used) tells us that in actual historical periods, there was a 0% chance of running out of money and the hapless retiree who was unfortunate enough to retire at the absolute worst time in history was still left with over 20% of his starting portfolio value (in real terms) at the end of the 30 year period (a far cry from the 10% chance of basically running out of money generated by Pfau's methodology).

Quote
How much would we expect a more aggressive AA to change the results? I basically never see a 50/50 AA on these boards.

Pretty drastically.  For the above 2.88% WR case, using actual historical data, going from 50/50 to 90/10 stocks/bonds almost doubles the portfolio ending values (and retains a 100% success rate).   

Quote
Any thoughts on how early a retiree would be able to recognize if he's in the bottom 10% of possibilities? 2 years? 5 years? 20 years? Never because no one has a properly working crystal ball.

Here's one recent thread where we discussed the ability to gauge whether your retirement plan is headed towards potential failure (including the strategy of periodically re-executing cfiresim throughout retirement using up-to-date numbers to see if you are still comfortable with your then-current historical success rate):  http://forum.mrmoneymustache.com/welcome-to-the-forum/when-fire-plans-go-wrong-your-experiences/

sol

  • Walrus Stache
  • *******
  • Posts: 8433
  • Age: 47
  • Location: Pacific Northwest
Re: Pfau on Variable Withdrawal Strategies
« Reply #14 on: March 17, 2015, 08:58:38 PM »
Pfau's research has always been overly pessimistic.  I haven't yet read this new publication, but all of his recent work has assumed a 1% MER.  If you're using low-fee Vanguard funds, you can add almost that entire 1% back on to his recommendations.

edit:  looks like he's assuming a 0.5% MER this time, so maybe he's heard from enough complainers.  Adding that 0.5% back on to his recommended 2.85% yields 3.35%, which is actually HIGHER than I would expect this type of Monte Carlo simulation to predict.

As previously discussed here and elsewhere, there are good reasons why Monte Carlo simulations like this (and Vanguard's) are a poor choice.  Market returns have always been correlated, both between years and between asset classes, in a way that this analyses neglects.  Interest rates tend to be higher in years when stocks have recently done well, for example, because that's how we manage monetary and fiscal policy in America.  Putting your faith in this kind of simulation is the same as assuming the future US economy will be run by monkeys.  Yes, things look gloomy when you assume any random horrible thing can happen at any time.

There are a ton of different ways to calculate future expected returns from current reality.  This is literally the worst one I can think of.
« Last Edit: March 17, 2015, 09:13:13 PM by sol »

sol

  • Walrus Stache
  • *******
  • Posts: 8433
  • Age: 47
  • Location: Pacific Northwest
Re: Pfau on Variable Withdrawal Strategies
« Reply #15 on: March 17, 2015, 09:30:03 PM »
After reading this entire paper, I've decided there's nothing new there that we haven't already discussed at length here. 

If anyone disagrees, I'd love to hear about it.  Maybe I missed something?
« Last Edit: March 17, 2015, 09:33:46 PM by sol »

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #16 on: March 17, 2015, 09:51:05 PM »
After reading this entire paper, I've decided there's nothing new there that we haven't already discussed at length here. 

If anyone disagrees, I'd love to hear about it.  Maybe I missed something?

IDK, most of our discussion is "4% is fine if you're flexible" - this is an example of flexible (as it uses the various adjustable spending strategies) and 4% doesn't work with any of them.

It does have those limitations noted above though (monte carlo over actual historical data and a 50/50 split instead of more aggressive one) -- like I mentioned, I would like to see the data for a chart with historical data and a 80/20 or 90/10 split.

So no, it may not offer a TON more than we've discussed, because of the limitations leaving us to shrug on what it looks like in those scenarios much more common to most of us, but it does put a small damper, IMO, on the "4% is fine just be flexible" idea.
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.

sol

  • Walrus Stache
  • *******
  • Posts: 8433
  • Age: 47
  • Location: Pacific Northwest
Re: Pfau on Variable Withdrawal Strategies
« Reply #17 on: March 17, 2015, 10:13:36 PM »
IDK, most of our discussion is "4% is fine if you're flexible" - this is an example of flexible (as it uses the various adjustable spending strategies) and 4% doesn't work with any of them.

It looks to me like 4% still works with four the strategies listed in Table 2, even after his 0.5% fee that previous SWR research doesn't include.

Quote
like I mentioned, I would like to see the data for a chart with historical data and a 80/20 or 90/10 split.

The MC approach greatly exaggerates the downside impact of higher stock allocations.  The random sequencing of returns means that you get a much larger spread of simulated outcomes than you get from the historical record, but the upside tail is still just "success" regardless of final balance.  The downside tail of the distribution, by contrast, looks much worse as a percentage of all possible outcomes.  In any model where you're trying to identify the likelihood of extreme tail events, adding artificial variability will make them look more likely.

Quote
So no, it may not offer a TON more than we've discussed, because of the limitations leaving us to shrug on what it looks like in those scenarios much more common to most of us, but it does put a small damper, IMO, on the "4% is fine just be flexible" idea.

The key message for these variable spending strategies has always been that you can spend more early on if you're willing to spend less later in life.  Which is the natural pattern of real retirees, btw, regardless of available assets.  bo_knows and I spent a whole thread detailing variables on these floor/ceiling plans for cFIREsim that can reproduce Pfau's strategies with historic returns, and we generally saw SWRs pushing 5% for cases we considred reasonable.

As your retirement unfolds and your time horizon shrinks, you effectively get to eat away half of your risk profile.  Planning a 10 year retirement at age 80 is a very different thing than planning a 50 year retirement at age 40, but Pfau's approach makes the risks look larger for the 80 year old than the 40 year old.

I love Pfau for doing the spreadsheet gruntwork that most people won't bother with, but I don't understand why he's such a pessimist.  Sometimes I think he's trying to mislead people when he phrases things in such a way that he can publish the lowest justifiable SWR numbers he can find.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #18 on: March 18, 2015, 07:37:59 AM »
After reading this entire paper, I've decided there's nothing new there that we haven't already discussed at length here. 

I think there are two (related) new things here, but neither of them are particularly noteworthy.

One is that this is the first rigorous apples-to-apples performance comparison of the competing variable spending strategies that I've seen.  Unfortunately, as discussed above, the pessimistic methodology Pfau chose to employ limits the utility of the data for decision-making about whether you've accumulated enough to commence FIRE (but it is still useful to compare the performance of the various spending strategies).  As discussed above, if someone were to reproduce this work using actual historical data (and throw in more aggressive allocation alternatives), I would be very interested to see the results.

The other is the concept of the "XYZ measure" (as an alternative to the traditional failure rate of total portfolio depletion) that he used to conduct the apples-to-apples comparison. 


It looks to me like 4% still works with four the strategies listed in Table 2, even after his 0.5% fee that previous SWR research doesn't include.


The four strategies listed in Table 2 with a 4% or greater initial withdrawal rate don't "work" in the sense we usually mean (i.e., ~95% chance of success for retirements having a strong likelihood of extending beyond 30 years).  In each of those cases, the portfolio is almost depleted to zero at the 30-year mark at least 10% of the time.  But again, those figures suffer from the use of Pfau's overly pessimistic methodology.

Quote
Planning a 10 year retirement at age 80 is a very different thing than planning a 50 year retirement at age 40, but Pfau's approach makes the risks look larger for the 80 year old than the 40 year old.

I don't follow this.  Pfau's MC methodology is more pessimistic for the reasons you gave and the reasons discussed earlier in this thread, but I don't see how his approach fails to account for the shorter time horizon of an older retiree. 

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 43
Re: Pfau on Variable Withdrawal Strategies
« Reply #19 on: March 18, 2015, 09:22:38 AM »
FYI, here's Pfau's response in the comments on his blog:

brooklynguy:

"Great attempt to provide a framework for making apples-to-apples comparisons across the different variable withdrawal strategies and for introducing an alternative to the traditional total-depletion “failure rate” metric!

I am curious what the results would have looked like if you had used actual historical data rather than Monte Carlo simulations with capital market performance assumptions. Based on some spot-checking of the figures in the tables against the results of history-based simulations on cfiresim.com using identical parameters, it appears that the results would have been substantially more optimistic."


Pfau:

"Thank you. I haven’t done this specific analysis with historical data, but I’m sure you’re right that results would have come out more optimistic. I think the main problem now is that there are not enough instances of such low interest rates in the historical data to really have a good idea about the implications of starting retirement in a low interest rate environment."


arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 28444
  • Age: -997
  • Location: Seattle, WA
Re: Pfau on Variable Withdrawal Strategies
« Reply #20 on: March 18, 2015, 09:33:31 AM »
Yeah, but it'd be nice if he still provided it, but with that caveat.   Of course it's his choice as to what framework to use, but I think many of us disagree with that decision.
I am a former teacher who accumulated a bunch of real estate, retired at 29, spent some time traveling the world full time and am now settled with three kids.
If you want to know more about me, this Business Insider profile tells the story pretty well.
I (rarely) blog at AdventuringAlong.com. Check out the Now page to see what I'm up to currently.