Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Maybe Wade Pfau was over-optimistic …

Post by randomguy »

afan wrote: Sat Oct 01, 2022 12:29 pm What are the correct countries, periods and events to include in an effort to guess the future? If you say "only post WWII US," then you get an optimistic answer.
Once you admit that other countries and other periods have had far lower returns, then you need to consider whether the optimistic outlook is justified. You need to consider how bad the low end could be.

Whether one adjusts spending and retirement date depends on attitudes towards risk. 1.9% is no less or more correct than 4%. Both are derived from datasets of SOME real world outcomes. Pick the one you think is most likely to reflect future events. Then switch off your crystal ball.
Yes and historically with those far lower returns, the SWR is still up around 3.5% in the couple of studies that have tried it. Why is this number lower in this study?

1) They aren't using historical returns. They are using a Monte carlo simulator that used historical returns. You can read their methodology and decide how good it is

2) They used longer time lines. The US 40 year SWR is more like 3.5% instead of 4%

3) they used different data. I will let you decide how you feel about 1890s data

4) their international returns excludes US. That makes no sense. The US is part of the developed world. I didn't see a global cap weighted SWR in the study.

You will have to decide how you feel about all these issues.
billaster
Posts: 2881
Joined: Wed Apr 27, 2022 2:21 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by billaster »

randomguy wrote: Sat Oct 01, 2022 6:36 pm
afan wrote: Sat Oct 01, 2022 12:29 pm What are the correct countries, periods and events to include in an effort to guess the future? If you say "only post WWII US," then you get an optimistic answer.
Once you admit that other countries and other periods have had far lower returns, then you need to consider whether the optimistic outlook is justified. You need to consider how bad the low end could be.

Whether one adjusts spending and retirement date depends on attitudes towards risk. 1.9% is no less or more correct than 4%. Both are derived from datasets of SOME real world outcomes. Pick the one you think is most likely to reflect future events. Then switch off your crystal ball.
Yes and historically with those far lower returns, the SWR is still up around 3.5% in the couple of studies that have tried it. Why is this number lower in this study?

1) They aren't using historical returns. They are using a Monte carlo simulator that used historical returns. You can read their methodology and decide how good it is

2) They used longer time lines. The US 40 year SWR is more like 3.5% instead of 4%

3) they used different data. I will let you decide how you feel about 1890s data

4) their international returns excludes US. That makes no sense. The US is part of the developed world. I didn't see a global cap weighted SWR in the study.

You will have to decide how you feel about all these issues.
5) They include newborns who won't even begin withdrawing for another 65 years.
User avatar
SimpleGift
Posts: 4477
Joined: Tue Feb 08, 2011 2:45 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by SimpleGift »

afan wrote: Sat Oct 01, 2022 12:29 pm Once you admit that other countries and other periods have had far lower returns, then you need to consider whether the optimistic outlook is justified. You need to consider how bad the low end could be.
Yes, but "how low the bad end was" during the economic disasters for developed countries in the 20th century (table below) seems far removed from "how low the bad end could be" for developed countries in the 21st century:
  • Image
    Source: Barro, Rare Disasters & Asset Markets in the 20th Century.
In the 21st century, should we be planning for geopolitical events in which major developed economies suffer 25% or greater concurrent losses in real GDP-per-capita? With two world wars and a global depression, this happened three times in the last century. The scale and scope of this economic instability in the developed world seems highly unlikely to recur in the century ahead, to my mind.
User avatar
squirrel1963
Posts: 1253
Joined: Wed Jun 21, 2017 10:12 am
Location: Portland OR area

Re: Maybe Wade Pfau was over-optimistic …

Post by squirrel1963 »

TN_Boy wrote: Sat Oct 01, 2022 8:37 am
McQ wrote: Fri Sep 30, 2022 8:55 pm A new paper looks at sustainable withdrawal rates using contemporary mortality and a much more comprehensive set of international stock and bond returns: “The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets,” https://papers.ssrn.com/sol3/papers.cfm ... id=4227132

Mark Hulbert glosses it here if you would prefer the Cliff Notes version: https://www.marketwatch.com/story/forge ... =home-page.

It’s way lower than anything seen in Bengen or the Trinity study*

*And Morningstar too, [Disrespectful comment removed by admin LadyGeek].
It's difficult to take this real seriously, although I do think that worrying about retirements longer than 30 years makes sense. In the original studies, even for retirements that "succeeded" sometimes the retiree was left with little money at the 30 year mark.

Here is a headliner ...
To achieve a 1% ruin probability, for example, retirees must adopt a withdrawal rate of just 0.80% (i.e., $8,000 of withdrawals per year for $1 million in savings).
Hmm. I'm having trouble with that.

My biggest problem was understanding exactly how they handled the pooled ex-US developed sample. If I missed the full description, just tell me what page it is on.

1) Did they just average the monthly returns of the developed countries (ranging from large economies like England, Germany, Japan, etc to tiny countries like Luxemburg) or did they weight by ... GDP for example? GDP would obviously vary over the years.

2) Did they correct for WWs?

3) I don't personally find data ranging back to the late 1800s interesting for either US or international securities.

I think that how a researcher handles non-US data is pretty important. For example, the stock market performance of a really small country is not interesting. It's just not. I think you have to look at the larger economies with well functioning stock markets and correct for issues like being bombed to rubble in WW II.

There have been other studies looking at how SWR for non-US investors would have worked and I think they were less ... pessimistic.
I don't know about the dataset used by the authors of this paper, but in regard to WW2, if I recall correctly (I believe from Bernstein "rational expectations):

All Anglo-Saxon countries had good positive real returns in the past 100 years, better than non-anglosaxon countries.
Germany had a positive real returns in the past 100 years despite WW2.
Italy had negative real returns in the past 100 years.
LMP | Liability Matching Portfolio | safe portfolio: TIPS ladder + I-bonds + Treasuries | risky portfolio: US stocks / US REIT / International stocks
User avatar
squirrel1963
Posts: 1253
Joined: Wed Jun 21, 2017 10:12 am
Location: Portland OR area

Re: Maybe Wade Pfau was over-optimistic …

Post by squirrel1963 »

SimpleGift wrote: Sat Oct 01, 2022 10:36 am
MH2 wrote: Sat Oct 01, 2022 7:42 am
SimpleGift wrote: Fri Sep 30, 2022 11:25 pm The returns data from this study comes from 38 developed countries over the 1890 to 2019 period. As shown in the study's summary table below, the ex-U.S. developed markets SWRs have been a fraction of the U.S. market's SWRs. No doubt that the U.S. has been favored historically, with no world wars on its home soil, and with little of the geopolitical turmoil experienced by many of the ex-U.S. developed countries since 1890.
The frustrating thing is that all of these issues can be accounted for with a couple of controls. This would be a good opportunity to show, e.g., the effects that something like hyper-inflation or a global war could have on withdraw rates.
Right. While showing the effects of hyper-inflation or world war on historical withdrawal rates might be a noteworthy addition to the academic body of financial knowledge, it still strikes me that these studies have close to zero relevance for prospective retirees who are planning their 21st century retirement portfolios.

In short: A world of difference between "financial archaeology" and what retirees need to plan for their future today.
It's tough to know how relevant the past 100 years financial history will be to the next 100 years, but a study of the past is the best thing we can do, no? If that's not the case then the Trinity study cannot be trusted either, which means any SWR determination would be just guesswork.
I personally find it useful to look at historical datasets to have some informed opinion on what future SWR will be.
LMP | Liability Matching Portfolio | safe portfolio: TIPS ladder + I-bonds + Treasuries | risky portfolio: US stocks / US REIT / International stocks
JackoC
Posts: 4710
Joined: Sun Aug 12, 2018 11:14 am

Re: Maybe Wade Pfau was over-optimistic …

Post by JackoC »

Escapevelocity wrote: Fri Sep 30, 2022 10:19 pm
CletusCaddy wrote: Fri Sep 30, 2022 10:05 pm Any analysis that produces less than a 3.33% SWR is literally non-sensical.

That’s because the SWR period is defined as 30 years, and with TIPS you could literally immunize your entire sequence of withdrawals from inflation with zero real return for 30 years.

A 100% TIPS portfolio defines the floor of SWR, in other words.

As a matter of fact, right now TIPS across the entire yield curve are yielding 2% real. So the SWR floor is actually more like 5% right now.
I’ve been waiting for someone to make this argument. Of course, we only arrived at this reality with TIPS very recently as they were yielding zero or less real until just a short time ago. The real yield of circa 2% with TIPS changes everything
Well TIPS are only yielding 2% if you round to the nearest %, and not sure why we'd do that if we can see the exact number, it's not an estimate of some unknown future value. But just about to start an expected 30 yr retirement, using the rule of thumb maturity half of horizon, 15 yr TIPS rate is 1.68% vs. -0.76% at the beginning of the year, quite a difference. As Keynes said “When the facts change, I change my mind. What do you do, sir?” One thing that would be nonsensical (though not saying either poster in this series said this) would be to counter statements about the expected return on TIPS made months ago by citing the yields now. Those yields have changed, and if one could have predicted that they should have been making money trading the bond market not debating SWR's.

But it's still not fair to criticize Pfau's result even if you're correcting for a simply different expected return right now than fairly recently. Another thing is 'SWR is for 30 yrs'. There's no point in semantic debate what's 'standard'. In practical reality there are threads all the time where people talk about retiring in early 60's, 50's, even sometimes earlier than that and still quote 4% as 'standard' SWR. Also Pfau's work IIRC focuses on possibility of lifespan/retirement age combination that make the span end up more than 30 yrs. A third thing is how taxes are factored in. The valid way IMO is a subtraction from portfolio return, in which case *after tax* return of TIPS in a taxable account is still only barely positive for high bracket taxpayers, assuming inflation at the TIPS BE, less if inflation ends up higher (which possibility is the reason to buy TIPS). If most of one's withdrawals are coming out of trad IRA/401k the 'portfolio' is really only the nominal IRA balance*(1-tax rate) but the growth rate of that discounted balance is the pre tax yield. However it's not rare for people to have most assets in taxable. And when people say OTOH 'I'm going to live on 4% of my portfolio' but are going to pay income taxes *from* that 4% that's not really 4% as I see it. A big chunk of the 4% in that case is not doing anything to support your lifestyle except keeping you out of jail for tax evasion :happy

Depending on definitions and situation, a 'SWR' of even less than 3% could be the answer even now, though that wasn't a common answer even when yields were at their lowest, if the only goal is not running out of money oneself (as opposed to ensuring a bequest). But saying 3.33% is the rational minimum is not only making the stated assumption of 30 yr retirement, which while reasonable doesn't always apply, it depends on assumption about how taxes are factored in which isn't stated.
Last edited by JackoC on Sat Oct 01, 2022 9:33 pm, edited 1 time in total.
User avatar
William Million
Posts: 1130
Joined: Wed May 05, 2010 4:41 am
Location: A Deep Mountain

Re: Maybe Wade Pfau was over-optimistic …

Post by William Million »

In the many years since I've been on this site, many have advocaded a sub-4% SWR, often in very strident terms. In all cases, these Bogleheads have been wrong.

Got nothing against reducing to 3% or 3.5% as a measure of caution. 1.9% is too silly to even debate.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Maybe Wade Pfau was over-optimistic …

Post by randomguy »

William Million wrote: Sat Oct 01, 2022 9:32 pm In the many years since I've been on this site, many have advocaded a sub-4% SWR, often in very strident terms. In all cases, these Bogleheads have been wrong.

Got nothing against reducing to 3% or 3.5% as a measure of caution. 1.9% is too silly to even debate.
A 1.9% SWR means that there is a 5% chance of it showing up. The other 95% you are taking out more money. You should only expect a 5% case to happen every 20 years. Probably more like 50 years given clustering. Bogleheads hasn't been around remotely long enough for use to see one. Odds are we are getting 20-80% probability returns and not those bottom 5% ones.

It does seem unlikely though that we are going to get like 30%+ worse returns than the historical numbers for a global portfolio....
User avatar
oldcomputerguy
Moderator
Posts: 17878
Joined: Sun Nov 22, 2015 5:50 am
Location: Tennessee

Re: Maybe Wade Pfau was over-optimistic …

Post by oldcomputerguy »

Some trolling comments were removed. As a reminder, see: General Etiquette
At all times we must conduct ourselves in a respectful manner to other posters. Attacks on individuals, insults, name calling, trolling, baiting or other attempts to sow dissension are not acceptable.
There is only one success - to be able to spend your life in your own way. (Christopher Morley)
tj
Posts: 9317
Joined: Wed Dec 23, 2009 11:10 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by tj »

SimpleGift wrote: Sat Oct 01, 2022 7:42 pm
afan wrote: Sat Oct 01, 2022 12:29 pm Once you admit that other countries and other periods have had far lower returns, then you need to consider whether the optimistic outlook is justified. You need to consider how bad the low end could be.
Yes, but "how low the bad end was" during the economic disasters for developed countries in the 20th century (table below) seems far removed from "how low the bad end could be" for developed countries in the 21st century:
  • Image
    Source: Barro, Rare Disasters & Asset Markets in the 20th Century.
In the 21st century, should we be planning for geopolitical events in which major developed economies suffer 25% or greater concurrent losses in real GDP-per-capita? With two world wars and a global depression, this happened three times in the last century. The scale and scope of this economic instability in the developed world seems highly unlikely to recur in the century ahead, to my mind.
I mean, if a global pandemic didn't cause a historically bad scenario, what will?
User avatar
SimpleGift
Posts: 4477
Joined: Tue Feb 08, 2011 2:45 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by SimpleGift »

tj wrote: Sun Oct 02, 2022 12:24 am I mean, if a global pandemic didn't cause a historically bad scenario, what will?
Yes, from the data I've seen, the declines in real GDP-per-capita from the Covid pandemic were in the range of -2% to -10% for the developed countries of the world. This isn't even in the same ballpark as the -20% to -60% declines in real GDP-per-capita for many developed countries during the economic disasters of the last century.
Wrench
Posts: 1046
Joined: Sun Apr 28, 2019 10:21 am

New study on Safe Withdrawal Rate (SWR)

Post by Wrench »

[Thread merged into here --admin LadyGeek]

This study:
https://papers.ssrn.com/sol3/papers.cfm ... id=4227132
argues SWRs are much lower than 4%. Here is the abstract:
We use a comprehensive new dataset of asset-class returns in 38 developed countries to exam-
ine a popular class of retirement spending rules that prescribe annual withdrawals as a constant
percentage of the retirement account balance. A 65-year-old couple willing to bear a 5% chance
of financial ruin can withdraw just 2.26% per year, a rate materially lower than convention-
al advice (e.g., the 4% rule). Our estimates of failure rates under conventional withdrawal
policies have important implications for individuals (e.g., savings rates, retirement timing, and
retirement consumption), public policy (e.g., participation rates in means-tested programs), and
society (e.g., elderly poverty rates).
My summary: the original Bengan study used a narrow time-limited data set from U.S. only, which because the U.S. in that time period had exception returns compared to other developed countries, significantly overestimates the SWR for a standard 60/40 stock/bond portfolio. Using a longer dataset across more developed countries coupled with longer life expectancies for current and future retirees, leads to much lower SWRs.
Wanderingwheelz
Posts: 3125
Joined: Mon Mar 04, 2019 8:52 am

Re: New study on Safe Withdrawal Rate (SWR)

Post by Wanderingwheelz »

I think this thread is already covering the subject you raised here.

viewtopic.php?p=6897306#p6897306
Being wrong compounds forever.
User avatar
tennisplyr
Posts: 3700
Joined: Tue Jan 28, 2014 12:53 pm
Location: Sarasota, FL

Re: New study on Safe Withdrawal Rate (SWR)

Post by tennisplyr »

Retired 11years and have been withdrawing what I needed over that time. Currently my portfolio is >30% higher than it was when I retired. Granted the bull market has been in play much of that time. For me, net net, use common sense, watch your expenses and live your life…it goes fast!
“Those who move forward with a happy spirit will find that things always work out.” -Retired 13 years 😀
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by LadyGeek »

I merged Wrench's thread into the ongoing discussion. I also clarified the thread title.

(Thanks to the member who reported the post and explained what's wrong.)
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.
Leesbro63
Posts: 10581
Joined: Mon Nov 08, 2010 3:36 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by Leesbro63 »

Sounds like Wade Pfau is basically arguing to limit spending to the 2-3% interest & dividends that are thrown off of an indexed 60/40 to 40/60 portfolio
User avatar
Garco
Posts: 1078
Joined: Wed Jan 23, 2013 1:04 am
Location: U.S.A.

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by Garco »

I am in my late 70's. I have a retirement fund accumulation of >$3 million. I am taking RMD's from a tax-deferred investment account.

While I'm taking what's required from the tax-deferred account, I'm not not spending it all. I also receive monthly Social Security checks. And I have property, some of which will be turned into cash within the next year.

So what is "SAFE"? Even if I live to 90 I'm not going to run out of cash.
Wrench
Posts: 1046
Joined: Sun Apr 28, 2019 10:21 am

Re: Maybe Wade Pfau was over-optimistic …

Post by Wrench »

Lawrence of Suburbia wrote: Sat Oct 01, 2022 2:24 pm
CletusCaddy wrote: Sat Oct 01, 2022 1:21 pm
Lee_WSP wrote: Sat Oct 01, 2022 1:20 pm
Wouldn’t an annuity also provide a floor?
Not against inflation
In my view, the function of an annuity (SPIA) today is to provide a sort of minimum income you can't outlive. It's longevity insurance, as opposed to an 'investment', really. The way it is used in a lot of retirement planning is to pay for your most basic expenses (food, housing, utilities perhaps) for a good number of years, while freeing up the retiree to buy more risky assets (more stocks, fewer bonds?) to hopefully stay ahead of inflation.

The big caveat here obviously is, it seems to me it probably only works out in the sort of low-inflation environment we've been blessed with in recent decades.

I'd love it if Wade (or anyone else hip to SPIAs in a retirement plan) would chip in here and correct my assumptions about this.
That's the way I think of my SPIA. Social Security, my SPIA (which happens to be inflation adjusted - got it before they disappeared) will cover ~100% of my expected expenses. I never considered using SWR as a strategy. For me, retirement income >= expected retirement expenses (with a cushion) for me and my spouse's expected lifetimes. I developed a plan to achieve that. What's left after that is for one time expenses, long term care and a legacy if there is anything left. The SPIA does allow me to invest more aggressively with the balance of my funds. There has always been too much uncertainty for me with SWR so I chose a different route. YMMV.

If one does not like to lose control of the initial premium of an SPIA, a TIPS ladder can achieve ~4% payout today for 30 years (as a previous poster indicated). One loses lifetime protection, but gains higher control. Alternatively, a 30 year STRIPS ladder which increases the number in each rung periodically can also achieve >4% payout if one prefers a fixed annual increase rather than CPI adjustment. Finally, it is quite possible to inflation adjust a nominal income (e.g., from an SPIA) using various techniques. There are several threads in BHs on how to do that. So, whether a 60/40 portfolio has a SWR of 4% or 1.9% could very well be a moot point for many BHs.

Wrench
iim7V7IM7
Posts: 466
Joined: Sun Nov 21, 2021 12:26 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by iim7V7IM7 »

I recommend reading Wade’s 2017 book “How Much Can I Spend in Retirement?” for his comprehensive review of various fixed and variable withdrawal strategies. He provides a very nice overview of their pros and cons and how they perform. He covers

- Bengen’s constant inflation adjusted strategy,
- fixed percentage withdrawals,
- endowment formulas,
- Bengen’s hard dollar floor and ceiling approach,
- Vanguard’s floor and ceiling approach,
- Kitces ratcheting rule,
- Guyton and Klinger’s decision rules,
- Zolt’s glide path rule and
- RMDs (actuarial methods such as Blanchett’s)

He also re-introduces the concept of a Pay Rule from his 2015 paper to evaluate each against vs a simple probability of failure (e.g., retiree accepts X% probability that their wealth falls below a threshold amount of $Y (in inflation adjusted terms) by year Z of retirement (probability, amount and year). This is a better way in my opinion for a retiree to assess performance parameters of each approach.

The “dash board” on his website is over interpreted in my opinion. 1) he has not updated it since April 2021 (things have changed quite a bit in 1-1/2 years), 2) the 2.39% and 3.50% withdrawal rates listed have very different acceptance criteria from Bill Bengen’s.

- The 2.39% for a 50/50 asset allocation assumes a 90% probability that real wealth does not fall below 15% of its initial level by year 30 of retirement
- The 3.50% for a 75/25 asset allocation assumes 80% probability that real wealth does not fall below 10% of its initial level by year 25 of retirement

These really cannot be compared to Bengen’s 1994 paper because they add different things to the estimate. Bengen just wanted to have enough money at the beginning of year 30 to support the inflation adjusted withdrawal.

I feel that Wade gets way too much guff here as being an annuity salesman, being too conservative or in some cases using unrealistic assumptions such as longevity or annual investment expenses. I believe his research contributions to retirement planning thinking are valuable to consider. They frankly have helped me think through things.
heyyou
Posts: 4452
Joined: Tue Feb 20, 2007 3:58 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by heyyou »

The entire premise of SWR is for the retiree to never modify his/her initial, inflation adjusted spending level for the entire 30 year period, regardless of remaining portfolio value during the whole period. SWR was the first, researched-based retirement spending tool, superior to all the unresearched suggestions that preceded it, but it is an imperfect spending plan due to its major feature of only and always increasing spending due to inflation. In some historical sequences, the SWR retiree has more than the initial portfolio amount after 30 years of spending, and the opposite case of portfolio failure if high inflation occurs for too many years, starting early in retirement. Yes, SWR is a good planning tool, but no longer one of the better spending plans.

The simple, easy sidestep is to use a newer, different spending plan, preferable one that also makes small adaptations annually (in either direction), in response to remaining portfolio values. That is comparable to re-retiring each year with your recent portfolio value and your expected longevity. To me, that is continuing to live within your means during retirement.

For a decade starting in the early 1990s?, I kept the copy of Mutual Funds magazine that had interviewed all of the then famous fund advocates about retirement spending. None of them had considered that question, and their answers would now be considered as useless. Salute Wm Bengen for his pioneering research, then move on to spending plans with better features than his, i.e. adapting your spending annually to each recent year-end portfolio value. David Zolt suggested criteria for whether to take each of SWR's annual increases, and Sun and Webb use longevity table info on each recent annual portfolio value.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by randomguy »

[quote=Leesbro63 post_id=6897905 time=1664717801 user_id=22011]
Sounds like Wade Pfau is basically arguing to limit spending to the 2-3% interest & dividends that are thrown off of an indexed 60/40 to 40/60 portfolio
[/quote]

Not really. Divs are 2-3% of current portfolio value not 2-4% real of initial portfolio value like a SWR. It is just coincidence the numbers are similiar right now.
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic …

Post by TN_Boy »

tj wrote: Sun Oct 02, 2022 12:24 am
SimpleGift wrote: Sat Oct 01, 2022 7:42 pm
afan wrote: Sat Oct 01, 2022 12:29 pm Once you admit that other countries and other periods have had far lower returns, then you need to consider whether the optimistic outlook is justified. You need to consider how bad the low end could be.
Yes, but "how low the bad end was" during the economic disasters for developed countries in the 20th century (table below) seems far removed from "how low the bad end could be" for developed countries in the 21st century:
  • Image
    Source: Barro, Rare Disasters & Asset Markets in the 20th Century.
In the 21st century, should we be planning for geopolitical events in which major developed economies suffer 25% or greater concurrent losses in real GDP-per-capita? With two world wars and a global depression, this happened three times in the last century. The scale and scope of this economic instability in the developed world seems highly unlikely to recur in the century ahead, to my mind.
I mean, if a global pandemic didn't cause a historically bad scenario, what will?
The war in Ukraine going nuclear, pulling NATO into the conflict, and (in addition to wider-spread fighting in Europe) things like major pipelines all over the world being attacked, leading to worldwide persistent severe energy shortages that cannot be addressed in the short term. Lots of people get cold and hungry the next 24 to 48 months, lots of industries are shutdown due to lack of power, or very expensive power (this is already a bit of a problem in parts of Europe).

You asked. I don't think it will happen :-) though. But I think the chances are non-zero of that debacle happening. Not that I have an investment plan for that contingency, since I can't figure out what said plan for that scenario would be.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by randomguy »

heyyou wrote: Sun Oct 02, 2022 10:57 am The entire premise of SWR is for the retiree to never modify his/her initial, inflation adjusted spending level for the entire 30 year period, regardless of remaining portfolio value during the whole period. SWR was the first, researched-based retirement spending tool, superior to all the unresearched suggestions that preceded it, but it is an imperfect spending plan due to its major feature of only and always increasing spending due to inflation. In some historical sequences, the SWR retiree has more than the initial portfolio amount after 30 years of spending, and the opposite case of portfolio failure if high inflation occurs for too many years, starting early in retirement. Yes, SWR is a good planning tool, but no longer one of the better spending plans.

The simple, easy sidestep is to use a newer, different spending plan, preferable one that also makes small adaptations annually (in either direction), in response to remaining portfolio values. That is comparable to re-retiring each year with your recent portfolio value and your expected longevity. To me, that is continuing to live within your means during retirement.

Sure and what happens with any of those type of schemes when you have a SOR that yield a 1.9% fixed SWR? Obviously you would need to put your scheme into their simulators to get the real numbers but a decade or so of spending 1.5%(i.e. a worse outcome) is a pretty reasonable guess given the performance of pretty much all the variable schemes using historical data.

Obviously if you really think we are in a 2% world, you don't invest 60/40 or the like. You load up on tips and the like.
furwut
Posts: 2122
Joined: Tue Jun 05, 2012 8:54 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by furwut »

Phyneas wrote: Sat Oct 01, 2022 4:32 am As something of an aside, is it a normal result of SWR studies that increasing your equity allocation beyond 60% is not expected to increase the SWR?
I think so. A bad series of returns in the early years of retirement (Sequence of Risk) can decimate a heavily laden equities portfolio leaving too little remaining to support future spending even with good returns.
MnD
Posts: 5184
Joined: Mon Jan 14, 2008 11:41 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MnD »

zaboomafoozarg wrote: Sat Oct 01, 2022 2:49 pm I'm saving for a 2% withdrawal rate because of research like this.

The future doesn't look good compared to the past, as the past year has made abundantly clear.
How many more years of working will this entail?
What experiences, both while working and after you retire are you going to have to forgo to achieve this?
Likewise for goods like nice reliable newer (and safer) vehicles?
70/30 AA for life, Global market cap equity. Rebalance if fixed income <25% or >35%. Weighted ER< .10%. 5% of annual portfolio balance SWR, Proportional (to AA) withdrawals.
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic …

Post by TN_Boy »

McQ wrote: Sat Oct 01, 2022 6:04 pm
TN_Boy wrote: Sat Oct 01, 2022 8:37 am ...
My biggest problem was understanding exactly how they handled the pooled ex-US developed sample. If I missed the full description, just tell me what page it is on.

1) Did they just average the monthly returns of the developed countries (ranging from large economies like England, Germany, Japan, etc to tiny countries like Luxemburg) or did they weight by ... GDP for example? GDP would obviously vary over the years.

2) Did they correct for WWs?

3) I don't personally find data ranging back to the late 1800s interesting for either US or international securities.

I think that how a researcher handles non-US data is pretty important. For example, the stock market performance of a really small country is not interesting. It's just not. I think you have to look at the larger economies with well functioning stock markets and correct for issues like being bombed to rubble in WW II.

There have been other studies looking at how SWR for non-US investors would have worked and I think they were less ... pessimistic.
Hello TN_Boy, here is a stab at answering the question you asked. I'm going off my understanding of their section 3. I am only moderately confident that I understand the method and hope others will correct as needed.

Your #1:
-returns are never averaged, except across assets within country/month. International returns, country by country, are capitalization-weighted; but international returns are not included in their base case. Rather, there are two pools, developed and US. Within each pool there is, say, N strings of stock returns, one string for each country. The string could be as long as (2019 - 1890) * 12. The actual historical sequence is preserved within string--1929 returns precede 1930 precede 1931 etc. The four asset returns for a month within country are locked together and preserved.

-at this point they could have stopped and analyzed all possible rolls of length M1, M2, Mn (=mortality). That would have been a strictly historical analysis, redone country by country and mortality length by mortality length, and they could still have got odds of failure (from a frequentist perspective).

-instead, from the complete set of four-fold strings they pick a start month and a country;
-then they draw a block length (mean length 120 months = 10 years). If that exceeds the country string length available (e.g., a 2015 start date in the US, with only four years left), they draw another block length and apply it to some other part of the sample and another start date, splicing that string onto the first.
-they do this until whatever mortality length being tested is filled (string spliced enough to be long enough).
-and then they make withdrawals against the spliced string and find if/when the funds run out before mortality.
-and, very important: they do that millions of times, which is where the odds of failure come from.

An unusual sort of Monte Carlo simulation, in which many but not all 10-year historical sequences are preserved, but where most mortality lengths contain a mash up of strings of different provenance. Possibly intended to give the analysis enough mathematical heft to have a shot at a peer-reviewed academic finance journal.

your 2 and 3:
yes, all years are included (Germany in 1923, Japan in 1945, etc.). Missing months (NYSE closed most of the last half of 2014) received interpolated returns.

re Old data. That is a hardy perennial in these SWR discussions: whether including the distant past makes the analysis worse or better. Your dismissal of 1890 etc. would carry more weight if you would put down a marker: "nothing before this date, and everything after, because X fundamentally changed." A standard example: don't mix bond returns from before and after the gold standard lapsed.

But could I not also say: "nothing before 20xx, because there weren't any internet or social media stocks then, with their new and awesome business models, making those returns irrelevant when looking forward from today." Slippery slope, once you start discarding history.

Myself, I prefer to include every bit of quality data available.
McQ,

Thanks for getting back to me. In reverse order, okay, I don't like 1890, what start date do I like :-). Fair question. I don't know. Frankly the post WW II period seems a lot more useful than earlier times, due to massive changes in the world, but obviously that leaves us with relatively few data points, and well, 2020 is rather different than 1946 too. I will say that when you start with mid 20th century, you have a situation where the US is the dominant economic and military power, and the US dollar is the world's reserve currency. I believe those factors do affect US stock returns.

Including WWs ... yeah, I just think (for example) the German stock market experience in the 20s through the 40s doesn't tell me much about what SWR is achievable. I don't know the best way to work around that sort of thing. More generally, what returns from what countries are somehow "not valid?"

As to the 1st question, the method still seems odd to me. I'm looking at page 8. Let me ponder that a bit more. If the block length is 10 years, then of course they are randomly picking non-adjacent blocks for a country to get the full 30+ year simulation for a given country?

Going to Figure 1, page 22. The graph has a single line for developed sample and a single line for the US sample. So the developed sample is just the totaled up results of all the non-US country samples. Where each simulation result had the same "weight." I.e. if 50% of the simulations using, for example, Slovakia failed, that counts the same as if 50% of the samples using England failed? And the overall failure at each payout rate is the total failures across simulations from all countries. And of course that figure is where the incredibly low value for a 1% failure rate comes from. Hmm.
MnD
Posts: 5184
Joined: Mon Jan 14, 2008 11:41 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MnD »

furwut wrote: Sun Oct 02, 2022 12:49 pm
Phyneas wrote: Sat Oct 01, 2022 4:32 am As something of an aside, is it a normal result of SWR studies that increasing your equity allocation beyond 60% is not expected to increase the SWR?
I think so. A bad series of returns in the early years of retirement (Sequence of Risk) can decimate a heavily laden equities portfolio leaving too little remaining to support future spending even with good returns.
For the 1966 retiree, tilting increasingly to nominal bonds resulted in portfolio failure sooner than all-equity.

100% equity portfolio failure in year 24
70:30 year 24
50:50 year 23
100% bonds year 21

https://www.cfiresim.com/050a67e3-9283- ... 162c26817f
70/30 AA for life, Global market cap equity. Rebalance if fixed income <25% or >35%. Weighted ER< .10%. 5% of annual portfolio balance SWR, Proportional (to AA) withdrawals.
User avatar
PicassoSparks
Posts: 417
Joined: Tue Apr 28, 2020 5:41 am

Re: Maybe Wade Pfau was over-optimistic …

Post by PicassoSparks »

SimpleGift wrote: Sat Oct 01, 2022 2:22 pm Since most of us did not experience firsthand the war-related economic dislocations of the 20th century, it is easy to underestimate their impact. But personally — and perhaps myopically — I have a hard time seeing this degree of economic disruption in the developed world, from any source, repeating in the 21st century.
As we are posting, there is a ground war happening in the developed world. For Ukrainian & Russian investors, that degree of disruption is currently in progress.
User avatar
SimpleGift
Posts: 4477
Joined: Tue Feb 08, 2011 2:45 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by SimpleGift »

PicassoSparks wrote: Sun Oct 02, 2022 1:54 pm As we are posting, there is a ground war happening in the developed world. For Ukrainian & Russian investors, that degree of disruption is currently in progress.
The question is not whether we'll see regional conflicts in the 21st century. It's that, in our current nuclear age of mutually-assured destruction — where the impacts of global war are incalculable and can't be priced by the markets — should we instead anticipate the scale and scope of global, war-related disasters that the world experienced in the 20th century (where GDP-per-capita was down 25%-50% in many developed countries)?

I don't see this as likely — and thus have a hard time accepting that asset returns from the war-ravaged, developed countries of the 20th century can be of much help to us in retirement planning for the 21st century.
MarkRoulo
Posts: 1192
Joined: Mon Jun 22, 2015 10:25 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MarkRoulo »

TN_Boy wrote: Sun Oct 02, 2022 1:09 pm ... snip ...

McQ,

Thanks for getting back to me. In reverse order, okay, I don't like 1890, what start date do I like :-). Fair question. I don't know. Frankly the post WW II period seems a lot more useful than earlier times, due to massive changes in the world, but obviously that leaves us with relatively few data points, and well, 2020 is rather different than 1946 too. I will say that when you start with mid 20th century, you have a situation where the US is the dominant economic and military power, and the US dollar is the world's reserve currency. I believe those factors do affect US stock returns.

Including WWs ... yeah, I just think (for example) the German stock market experience in the 20s through the 40s doesn't tell me much about what SWR is achievable. I don't know the best way to work around that sort of thing. More generally, what returns from what countries are somehow "not valid?"

As to the 1st question, the method still seems odd to me. I'm looking at page 8. Let me ponder that a bit more. If the block length is 10 years, then of course they are randomly picking non-adjacent blocks for a country to get the full 30+ year simulation for a given country?

Going to Figure 1, page 22. The graph has a single line for developed sample and a single line for the US sample. So the developed sample is just the totaled up results of all the non-US country samples. Where each simulation result had the same "weight." I.e. if 50% of the simulations using, for example, Slovakia failed, that counts the same as if 50% of the samples using England failed? And the overall failure at each payout rate is the total failures across simulations from all countries. And of course that figure is where the incredibly low value for a 1% failure rate comes from. Hmm.
My *guess* is that they randomly pick a start year and then take the next ten years. Three of these picks and you'd have thirty years. Two specific questions that I'd love to know the answer to, though are:
  • Do they pick with or without replacement? e.g., can you randomly pick 1930 twice when assembling a single 30-year (or whatever) sequence?
  • If they don't pick with replacement, can they randomly select overlapping regions (e.g. 1930 and 1935) or do they insist that any given yearly return be counted only once in a given generated sequence?
I'm sure these aren't a secret, I just don't know the answer.
Poe22
Posts: 249
Joined: Thu Sep 15, 2022 1:21 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by Poe22 »

petulant wrote: Sat Oct 01, 2022 1:20 pm You might as well run a Monte Carlo situation based on 65 million years of data inputs and conclude that the SWR is 0% because in the worst 5% of cases a meteor will strike the Yucatan peninsula and cause a mass extinction.
I think you nailed it. We mustn't forget what Monte Carlo simulations are: Simulations of the absolute worst to the absolute best outcomes. The absolute worst outcome will always be 0% SWR: the total destruction of the economy/market.

Planning with the worst SWR is like living as if you'd die within the next minute. Who in his right mind does that?
User avatar
Charles Joseph
Posts: 2366
Joined: Tue Apr 05, 2022 10:49 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by Charles Joseph »

McQ wrote: Fri Sep 30, 2022 8:55 pm A new paper looks at sustainable withdrawal rates using contemporary mortality and a much more comprehensive set of international stock and bond returns: “The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets,” https://papers.ssrn.com/sol3/papers.cfm ... id=4227132

Mark Hulbert glosses it here if you would prefer the Cliff Notes version: https://www.marketwatch.com/story/forge ... =home-page.

It’s way lower than anything seen in Bengen or the Trinity study*

*And Morningstar too, [Disrespectful comment removed by admin LadyGeek].
Just take what you think you need out of your portfolio the first year (after making reasonable estimates of course), see how that stacks up, then make appropriate adjustments as needed, tightening your belt or loosening it as tougher times dictate or better times allow.

A lot of this endless, incessant safe withdrawal stuff is really a waste of ink. Ask Taylor about the Larimore Retirement Withdrawal Plan. Just pure old common sense.
"The big money is not in the buying and selling, but in the waiting." - Charles Munger
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Maybe Wade Pfau was over-optimistic …

Post by vineviz »

Poe22 wrote: Sun Oct 02, 2022 3:26 pm
petulant wrote: Sat Oct 01, 2022 1:20 pm You might as well run a Monte Carlo situation based on 65 million years of data inputs and conclude that the SWR is 0% because in the worst 5% of cases a meteor will strike the Yucatan peninsula and cause a mass extinction.
I think you nailed it. We mustn't forget what Monte Carlo simulations are: Simulations of the absolute worst to the absolute best outcomes. The absolute worst outcome will always be 0% SWR: the total destruction of the economy/market.

Planning with the worst SWR is like living as if you'd die within the next minute. Who in his right mind does that?
Yeah, that's not how Monte Carlo simulations work. Certainly not when the parameters are coherently specified.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic …

Post by TN_Boy »

MarkRoulo wrote: Sun Oct 02, 2022 3:18 pm
TN_Boy wrote: Sun Oct 02, 2022 1:09 pm ... snip ...

McQ,

Thanks for getting back to me. In reverse order, okay, I don't like 1890, what start date do I like :-). Fair question. I don't know. Frankly the post WW II period seems a lot more useful than earlier times, due to massive changes in the world, but obviously that leaves us with relatively few data points, and well, 2020 is rather different than 1946 too. I will say that when you start with mid 20th century, you have a situation where the US is the dominant economic and military power, and the US dollar is the world's reserve currency. I believe those factors do affect US stock returns.

Including WWs ... yeah, I just think (for example) the German stock market experience in the 20s through the 40s doesn't tell me much about what SWR is achievable. I don't know the best way to work around that sort of thing. More generally, what returns from what countries are somehow "not valid?"

As to the 1st question, the method still seems odd to me. I'm looking at page 8. Let me ponder that a bit more. If the block length is 10 years, then of course they are randomly picking non-adjacent blocks for a country to get the full 30+ year simulation for a given country?

Going to Figure 1, page 22. The graph has a single line for developed sample and a single line for the US sample. So the developed sample is just the totaled up results of all the non-US country samples. Where each simulation result had the same "weight." I.e. if 50% of the simulations using, for example, Slovakia failed, that counts the same as if 50% of the samples using England failed? And the overall failure at each payout rate is the total failures across simulations from all countries. And of course that figure is where the incredibly low value for a 1% failure rate comes from. Hmm.
My *guess* is that they randomly pick a start year and then take the next ten years. Three of these picks and you'd have thirty years. Two specific questions that I'd love to know the answer to, though are:
  • Do they pick with or without replacement? e.g., can you randomly pick 1930 twice when assembling a single 30-year (or whatever) sequence?
  • If they don't pick with replacement, can they randomly select overlapping regions (e.g. 1930 and 1935) or do they insist that any given yearly return be counted only once in a given generated sequence?
I'm sure these aren't a secret, I just don't know the answer.
Good questions. I also think they pick a 10 year block, then randomly another 10 year block, but don't know how they handle the situations you mention.

Plus by picking non-sequential 10 year blocks (e.g. 1940 to 1950 then 1960 to 1970) you have the situation where mean reversion - if one believes in that - is sometimes ignored, as you might pick two really bad sequences in a row. But Monte Carlo simulations are complicated, and sometimes controversial.
MarkRoulo
Posts: 1192
Joined: Mon Jun 22, 2015 10:25 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MarkRoulo »

McQ wrote: Sat Oct 01, 2022 5:08 pm
petulant wrote: Sat Oct 01, 2022 12:17 pm ... I am disappointed that user McQ is now apparently trying to match marketwatch headlines for clickbait ...
I note for the record that you are the third Boglehead to accuse me of clickbait (I keep a running tally, I doubt you will be the last).

Two inferences occur:
1. Where there’s smoke, there must be fire;
Or
2. Time to dust off my copy of Plato’s Republic and find the passage where he bans poets because of their fatal predilection for clickbait and other violations of logic and probity.

More than a whiff of that Plato here at Bogleheads, in my experience thus far.

Conversely, your point about whether SWR analyses should include wars does raise a key question. I don't see that the answer is a foregone conclusion, and hope we can debate the question here in this thread.
I don't think we will reach a conclusion.

I will ILLUSTRATE (not prove) why with a few examples that begin as questions.

Question #1: What was the homicide rate in New York City from 2000 - 2010?

Homicide rate is commonly reported as homicides per 100,000 and the US as a whole was a bit over 9.0 in 2000. This dropped to about 4.5 in 2014 and has since risen to 6.5. For reference, Baltimore runs over 50. And upper middle class suburbs often see a rate of around 1.0.

So ... New York City saw around 6,000 homicides over those ten year (so 600 per year) and had a population of around 8 million. This works out to about 7.5 per 100,000. Quite good for a large city compared to the US as a whole.

But there is a catch and it is a big one. The 6,000 count does NOT include the 3,000 people killed on 9/11. I would have a difficult time arguing that flying airplanes into a building full of people didn't count as homicide, but the official stats for NYC *exclude* those deaths for 2001 (the official number of homicides as 649).

If one is trying to ascertain the chances of getting killed in NYC over the next 30-50 years should those deaths be excluded from the backward looking sample because this is unlikely to happen again? Or because it is rare? Or should they be included because every so often this sort of thing DOES happen even if we probably won't get an exact repeat?

Rare events are tough to model, partially because they are rare so we don't have a good idea of the true underlying probability.

I'm not going to offer an answer, but I think this question illustrates well the stats problem to solve.

Question #2 (This is a variation on #1): What is the average homicide rate in Europe compared to the US?

If one looks up the basic statistics one finds something like, "European countries tend to have homicide rates 1/3 - 1/4 that of the US."

Which is a pretty reasonable answer and addresses the question most people care about if they are going to Europe on holiday or trying to make some sort of point about gun control.

But it excludes the 12,000,000 people murdered in the Holocaust (1933 - 1945).

Another way to view the homicide rate is that in any given year the European rate has been quite a bit lower than that of the US, but there have been a few years that were INSANELY HIGH compared to the US.

Note that we can play this game even today because when Russian missiles are aimed at Ukrainian hospitals that almost certainly doesn't show up in the basic homicide statistics for Ukraine though the folks are just as dead as if they had been murdered.

So ... retain the data with World Wars and Great Depressions (and Argentina and Japan post-WW2) or remove it because these events are: (a) rare and, (b) unlikely to be repeated?

I don't think "we" are going to converge on an answer.
MarkRoulo
Posts: 1192
Joined: Mon Jun 22, 2015 10:25 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MarkRoulo »

TN_Boy wrote: Sun Oct 02, 2022 3:46 pm
MarkRoulo wrote: Sun Oct 02, 2022 3:18 pm
TN_Boy wrote: Sun Oct 02, 2022 1:09 pm ... snip ...

McQ,

Thanks for getting back to me. In reverse order, okay, I don't like 1890, what start date do I like :-). Fair question. I don't know. Frankly the post WW II period seems a lot more useful than earlier times, due to massive changes in the world, but obviously that leaves us with relatively few data points, and well, 2020 is rather different than 1946 too. I will say that when you start with mid 20th century, you have a situation where the US is the dominant economic and military power, and the US dollar is the world's reserve currency. I believe those factors do affect US stock returns.

Including WWs ... yeah, I just think (for example) the German stock market experience in the 20s through the 40s doesn't tell me much about what SWR is achievable. I don't know the best way to work around that sort of thing. More generally, what returns from what countries are somehow "not valid?"

As to the 1st question, the method still seems odd to me. I'm looking at page 8. Let me ponder that a bit more. If the block length is 10 years, then of course they are randomly picking non-adjacent blocks for a country to get the full 30+ year simulation for a given country?

Going to Figure 1, page 22. The graph has a single line for developed sample and a single line for the US sample. So the developed sample is just the totaled up results of all the non-US country samples. Where each simulation result had the same "weight." I.e. if 50% of the simulations using, for example, Slovakia failed, that counts the same as if 50% of the samples using England failed? And the overall failure at each payout rate is the total failures across simulations from all countries. And of course that figure is where the incredibly low value for a 1% failure rate comes from. Hmm.
My *guess* is that they randomly pick a start year and then take the next ten years. Three of these picks and you'd have thirty years. Two specific questions that I'd love to know the answer to, though are:
  • Do they pick with or without replacement? e.g., can you randomly pick 1930 twice when assembling a single 30-year (or whatever) sequence?
  • If they don't pick with replacement, can they randomly select overlapping regions (e.g. 1930 and 1935) or do they insist that any given yearly return be counted only once in a given generated sequence?
I'm sure these aren't a secret, I just don't know the answer.
Good questions. I also think they pick a 10 year block, then randomly another 10 year block, but don't know how they handle the situations you mention.

Plus by picking non-sequential 10 year blocks (e.g. 1940 to 1950 then 1960 to 1970) you have the situation where mean reversion - if one believes in that - is sometimes ignored, as you might pick two really bad sequences in a row. But Monte Carlo simulations are complicated, and sometimes controversial.
The point to the 10-year periods is to pick up some/most of the effects of mean reversion WITHOUT having to explicitly set parameters in their simulation for mean reversion. I think this is a pretty good (though not perfect ... but that isn't a choice here ...) way to get some mean reversion (which selecting single years randomly does not do) while avoiding the problem of explicitly coding up a reversion term and getting it wrong.
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by TN_Boy »

Charles Joseph wrote: Sun Oct 02, 2022 3:33 pm
McQ wrote: Fri Sep 30, 2022 8:55 pm A new paper looks at sustainable withdrawal rates using contemporary mortality and a much more comprehensive set of international stock and bond returns: “The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets,” https://papers.ssrn.com/sol3/papers.cfm ... id=4227132

Mark Hulbert glosses it here if you would prefer the Cliff Notes version: https://www.marketwatch.com/story/forge ... =home-page.

It’s way lower than anything seen in Bengen or the Trinity study*

*And Morningstar too, [Disrespectful comment removed by admin LadyGeek].
Just take what you think you need out of your portfolio the first year (after making reasonable estimates of course), see how that stacks up, then make appropriate adjustments as needed, tightening your belt or loosening it as tougher times dictate or better times allow.

A lot of this endless, incessant safe withdrawal stuff is really a waste of ink. Ask Taylor about the Larimore Retirement Withdrawal Plan. Just pure old common sense.
Well, I always have to point this out, Taylor retired in 1982 with a 1M or so portfolio (maybe a pension too, I forget). Look up

1) what 1M in 82 is worth now and
2) stock and bond returns beginning in 1982.

It was basically the best time in US history for someone to retire partially living off an investment portfolio. Common sense is harder when returns are not so kind, which is why people wrangle about these things. Of course Taylor did not KNOW that 1982 and forward would be so great, nor did we have the advantage of good SWR studies. But the fact remains, one's portfolio starting in 1982 was basically doing really good things year after year.
MarkRoulo
Posts: 1192
Joined: Mon Jun 22, 2015 10:25 am

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by MarkRoulo »

Charles Joseph wrote: Sun Oct 02, 2022 3:33 pm
McQ wrote: Fri Sep 30, 2022 8:55 pm A new paper looks at sustainable withdrawal rates using contemporary mortality and a much more comprehensive set of international stock and bond returns: “The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets,” https://papers.ssrn.com/sol3/papers.cfm ... id=4227132

Mark Hulbert glosses it here if you would prefer the Cliff Notes version: https://www.marketwatch.com/story/forge ... =home-page.

It’s way lower than anything seen in Bengen or the Trinity study*

*And Morningstar too, [Disrespectful comment removed by admin LadyGeek].
Just take what you think you need out of your portfolio the first year (after making reasonable estimates of course), see how that stacks up, then make appropriate adjustments as needed, tightening your belt or loosening it as tougher times dictate or better times allow.

A lot of this endless, incessant safe withdrawal stuff is really a waste of ink. Ask Taylor about the Larimore Retirement Withdrawal Plan. Just pure old common sense.
I don't expect many people to follow the 4% rule as a withdrawal strategy.

But one still has to decide when to retire even if one intends to follow a more flexible spending rule.

4% works out to 25x projected/expected/desired spending.

If 25x isn't enough then is 50x? And if one should just take what you need out of your portfolio and then make appropriate adjustments what is the reasonable estimate to start with? 25x? Or more or less?

In many respects this is a debate about how much money one "needs" to retire it is just expressed as a fixed/inflation-adjusted spending ratio rather than as a multiple of reasonably estimated expenses.
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic …

Post by TN_Boy »

MarkRoulo wrote: Sun Oct 02, 2022 3:49 pm
TN_Boy wrote: Sun Oct 02, 2022 3:46 pm
MarkRoulo wrote: Sun Oct 02, 2022 3:18 pm
TN_Boy wrote: Sun Oct 02, 2022 1:09 pm ... snip ...

McQ,

Thanks for getting back to me. In reverse order, okay, I don't like 1890, what start date do I like :-). Fair question. I don't know. Frankly the post WW II period seems a lot more useful than earlier times, due to massive changes in the world, but obviously that leaves us with relatively few data points, and well, 2020 is rather different than 1946 too. I will say that when you start with mid 20th century, you have a situation where the US is the dominant economic and military power, and the US dollar is the world's reserve currency. I believe those factors do affect US stock returns.

Including WWs ... yeah, I just think (for example) the German stock market experience in the 20s through the 40s doesn't tell me much about what SWR is achievable. I don't know the best way to work around that sort of thing. More generally, what returns from what countries are somehow "not valid?"

As to the 1st question, the method still seems odd to me. I'm looking at page 8. Let me ponder that a bit more. If the block length is 10 years, then of course they are randomly picking non-adjacent blocks for a country to get the full 30+ year simulation for a given country?

Going to Figure 1, page 22. The graph has a single line for developed sample and a single line for the US sample. So the developed sample is just the totaled up results of all the non-US country samples. Where each simulation result had the same "weight." I.e. if 50% of the simulations using, for example, Slovakia failed, that counts the same as if 50% of the samples using England failed? And the overall failure at each payout rate is the total failures across simulations from all countries. And of course that figure is where the incredibly low value for a 1% failure rate comes from. Hmm.
My *guess* is that they randomly pick a start year and then take the next ten years. Three of these picks and you'd have thirty years. Two specific questions that I'd love to know the answer to, though are:
  • Do they pick with or without replacement? e.g., can you randomly pick 1930 twice when assembling a single 30-year (or whatever) sequence?
  • If they don't pick with replacement, can they randomly select overlapping regions (e.g. 1930 and 1935) or do they insist that any given yearly return be counted only once in a given generated sequence?
I'm sure these aren't a secret, I just don't know the answer.
Good questions. I also think they pick a 10 year block, then randomly another 10 year block, but don't know how they handle the situations you mention.

Plus by picking non-sequential 10 year blocks (e.g. 1940 to 1950 then 1960 to 1970) you have the situation where mean reversion - if one believes in that - is sometimes ignored, as you might pick two really bad sequences in a row. But Monte Carlo simulations are complicated, and sometimes controversial.
The point to the 10-year periods is to pick up some/most of the effects of mean reversion WITHOUT having to explicitly set parameters in their simulation for mean reversion. I think this is a pretty good (though not perfect ... but that isn't a choice here ...) way to get some mean reversion (which selecting single years randomly does not do) while avoiding the problem of explicitly coding up a reversion term and getting it wrong.
Well 10 year strings might be better than 1 year random picks, but you can select a 10 year string ending with a bad sequence following by picking a 10 year string starting with a bad sequence and wind up with a single really really bad sequence.

You are probably right this is still better though. One way to test that would be to see if simulation results vary much if you pick different length strings, one year, three years, five years, 10 years, etc, including a 30 year string. That would tell you how sensitive the results are to string length. Which would kinda be an interesting study all by itself.
Zeno
Posts: 1042
Joined: Wed Sep 12, 2018 10:44 am

Re: Maybe Wade Pfau was over-optimistic …

Post by Zeno »

deleted
Last edited by Zeno on Wed Dec 28, 2022 5:46 pm, edited 1 time in total.
MarkRoulo
Posts: 1192
Joined: Mon Jun 22, 2015 10:25 am

Re: Maybe Wade Pfau was over-optimistic …

Post by MarkRoulo »

TN_Boy wrote: Sun Oct 02, 2022 3:57 pm
MarkRoulo wrote: Sun Oct 02, 2022 3:49 pm ... snip ...

The point to the 10-year periods is to pick up some/most of the effects of mean reversion WITHOUT having to explicitly set parameters in their simulation for mean reversion. I think this is a pretty good (though not perfect ... but that isn't a choice here ...) way to get some mean reversion (which selecting single years randomly does not do) while avoiding the problem of explicitly coding up a reversion term and getting it wrong.
Well 10 year strings might be better than 1 year random picks, but you can select a 10 year string ending with a bad sequence following by picking a 10 year string starting with a bad sequence and wind up with a single really really bad sequence.

You are probably right this is still better though. One way to test that would be to see if simulation results vary much if you pick different length strings, one year, three years, five years, 10 years, etc, including a 30 year string. That would tell you how sensitive the results are to string length. Which would kinda be an interesting study all by itself.
I think the authors would argue that there might be a chance of two bad sequences happening back-to-back (though they don't know what that probability is) and they want to capture this. If one was sure that this could never happen, then one could embed that reversion to the mean logic/probability in a Monte Carlo simulation. The catch with excluding it is that if the chance IS REAL and non-zero then excluding it gives us a false sense of security.

Note also that this entire discussion is really just another variation of, "We don't have enough sample data to be very confident in our conclusions."

The Trinity study worked with data from 1926. At the time (1995?) this was fewer than three independent 30-year returns. There just isn't that much you can conclude from three independ data points. Extending the data to the rest of the world brings its own problems:
  • The markets in the rest of the world is correlated with the US and with each other so you don't get fully independent new data, and
  • You have to decide what to do about WW1, WW2, Argentina, ...
The error bars (if one does the stats carefully) are just larger than most people want them to be and this leads to 95% confidence withdrawal rates that are depressingly low. Because we just don't know!
User avatar
9-5 Suited
Posts: 1306
Joined: Thu Jun 23, 2016 12:14 pm

Re: Maybe Wade Pfau was over-optimistic …

Post by 9-5 Suited »

William Million wrote: Sat Oct 01, 2022 9:32 pm In the many years since I've been on this site, many have advocaded a sub-4% SWR, often in very strident terms. In all cases, these Bogleheads have been wrong.

Got nothing against reducing to 3% or 3.5% as a measure of caution. 1.9% is too silly to even debate.
Imagine if you are a typical strip mall brokerage house client and pay a 1% AUM fee, have 0.75% average mutual fund fees, and experience 0.25% in trading cost and tax drag. Under the 1.9% scenario, you can take out $0 at the beginning to fund your retirement expenses since money is fungible and all costs are created equal.
User avatar
SimpleGift
Posts: 4477
Joined: Tue Feb 08, 2011 2:45 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by SimpleGift »

MarkRoulo wrote: Sun Oct 02, 2022 4:09 pm Note also that this entire discussion is really just another variation of, "We don't have enough sample data to be very confident in our conclusions."
Sure, we never have enough data. But it seems the essential question we are trying to answer in retirement planning for U.S. investors is: "What are the likely expected returns to capitalist enterprise in the 21st century?"

And we do have over 150 years of stock market return data from developed, capitalist countries around the world whose legal systems respect the rights of shareholders, whose stocks markets were not closed and confiscated by revolution, whose economies were not devastated by world wars, etc. Think the U.S., the U.K., Australia, Switzerland, Canada, Sweden, and the like.

If we are trying to project expected returns into the 21st century, why not use the historical returns of these relatively stable, capitalist democracies — and then just call it a day, since it is the best that we can now forecast for U.S. investors?
User avatar
Lawrence of Suburbia
Posts: 691
Joined: Mon Aug 08, 2022 12:04 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by Lawrence of Suburbia »

Charles Joseph wrote: Sun Oct 02, 2022 3:33 pm Just take what you think you need out of your portfolio the first year (after making reasonable estimates of course), see how that stacks up, then make appropriate adjustments as needed, tightening your belt or loosening it as tougher times dictate or better times allow.

[...] Ask Taylor about the Larimore Retirement Withdrawal Plan. Just pure old common sense.
While I do find all the discussions about SWR fascinating, I find myself sort of deer-in-the-headlights about it all. I mean, what if we get some sort of hot WWIII situation? North America has no longer got the natural defense of geographic isolation; even puny North Korea can now plaster the western U.S. with nukes. We've seen with the Nord pipeline attacks how easy it is to disrupt the world's energy supplies. Etc., etc., etc. ... Adjusting one's SWR won't make a blind bit of difference in a real 21st century world conflict or other unknown, planet-wide catastrophe.

So, I'm trying to operate as if a more positive outcome is the likeliest outcome; assuming that the U.S. and most of the civilized world muddle through, science advances, markets continue to operate, and people of good will somehow find a way to get along. I give a positive outcome about a 66% chance.

My withdrawal rate currently ~3% ... just 'cos that's what I'm currently needing to live. And going forward, my plan is to take money out of my stash whenever I need to cover my living expenses, eat out somewhere once or twice a month, every 2-3 years take a trip somewhere, maintain my health as best as possible, and ... smell the roses, gratefully, with my face towards the Sun. And not to worry anymore about portfolio percentages.

(Note: I'm saying this more for myself than anyone else; maybe I should incorporate it into some sort of investment plan statement.)
74% VTHRX/8% DODWX/12% TIAA Traditional/6% SWVXX
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Maybe Wade Pfau was over-optimistic …

Post by randomguy »

MarkRoulo wrote: Sun Oct 02, 2022 4:09 pm
I think the authors would argue that there might be a chance of two bad sequences happening back-to-back (though they don't know what that probability is) and they want to capture this. If one was sure that this could never happen, then one could embed that reversion to the mean logic/probability in a Monte Carlo simulation. The catch with excluding it is that if the chance IS REAL and non-zero then excluding it gives us a false sense of security.
On the other hand by not having mean reversion, they give a false sense of pessimism. Let's say you get a 0% decade 10% of the time. What are the odds of having a dead decade after another one. Monte carlo says it is still 10%. Historically though reversion to the mean might say it is more like 1%. So you way over state the risk. And maybe it is possible to get 1923-1932 followed by 2000-9. Maybe it is possible to have 7 down years like that in a row. But not even the Nikki 225 dropped like that.

And you can do the same logic with interest/bond rates. Imagine you pull 1972-1981. But after that you pull say 1950-1959. You get good stock returns but you don't get any of those great bonds returns from falling rates. You paid the price on the way up but get none of the benefit on the way down.

It should be trivial for the authors to calculate the historical SWR rates and compare them to what their simulation is popping out. You can then decide if we just got lucky and had a better than average outcome or if their model is pessimistic...
User avatar
Charles Joseph
Posts: 2366
Joined: Tue Apr 05, 2022 10:49 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by Charles Joseph »

Lawrence of Suburbia wrote: Sun Oct 02, 2022 6:15 pm
Charles Joseph wrote: Sun Oct 02, 2022 3:33 pm Just take what you think you need out of your portfolio the first year (after making reasonable estimates of course), see how that stacks up, then make appropriate adjustments as needed, tightening your belt or loosening it as tougher times dictate or better times allow.

[...] Ask Taylor about the Larimore Retirement Withdrawal Plan. Just pure old common sense.
While I do find all the discussions about SWR fascinating, I find myself sort of deer-in-the-headlights about it all. I mean, what if we get some sort of hot WWIII situation? North America has no longer got the natural defense of geographic isolation; even puny North Korea can now plaster the western U.S. with nukes. We've seen with the Nord pipeline attacks how easy it is to disrupt the world's energy supplies. Etc., etc., etc. ... Adjusting one's SWR won't make a blind bit of difference in a real 21st century world conflict or other unknown, planet-wide catastrophe.

So, I'm trying to operate as if a more positive outcome is the likeliest outcome; assuming that the U.S. and most of the civilized world muddle through, science advances, markets continue to operate, and people of good will somehow find a way to get along. I give a positive outcome about a 66% chance.

My withdrawal rate currently ~3% ... just 'cos that's what I'm currently needing to live. And going forward, my plan is to take money out of my stash whenever I need to cover my living expenses, eat out somewhere once or twice a month, every 2-3 years take a trip somewhere, maintain my health as best as possible, and ... smell the roses, gratefully, with my face towards the Sun. And not to worry anymore about portfolio percentages.

(Note: I'm saying this more for myself than anyone else; maybe I should incorporate it into some sort of investment plan statement.)
Colonel Lawrence - Excellent post. Thank you. And I left out one part in my comments about safe withdrawal rate discussions being a waste of time: I obsess about my possible safe withdrawal rates. I guess at the end of the day I'll just see how I do.

You do bring up a fascinating point about certain risks. I've come to peace with some of those in this way: I view certain risks as no risks at all. In other words, I ignore them. Things like a nuclear attack or, say, the default of the US government on its debt obligations, these things I can't really plan for or mitigate, so I don't consider them as risks for planning purposes. Interest rate risk, credit risk, etc. - those things I can work on. The others I really can't control. So I just stick my head in the sand (from a finance perspective).
"The big money is not in the buying and selling, but in the waiting." - Charles Munger
TN_Boy
Posts: 4076
Joined: Sat Jan 17, 2009 11:51 am

Re: Maybe Wade Pfau was over-optimistic …

Post by TN_Boy »

MarkRoulo wrote: Sun Oct 02, 2022 4:09 pm
TN_Boy wrote: Sun Oct 02, 2022 3:57 pm
MarkRoulo wrote: Sun Oct 02, 2022 3:49 pm ... snip ...

The point to the 10-year periods is to pick up some/most of the effects of mean reversion WITHOUT having to explicitly set parameters in their simulation for mean reversion. I think this is a pretty good (though not perfect ... but that isn't a choice here ...) way to get some mean reversion (which selecting single years randomly does not do) while avoiding the problem of explicitly coding up a reversion term and getting it wrong.
Well 10 year strings might be better than 1 year random picks, but you can select a 10 year string ending with a bad sequence following by picking a 10 year string starting with a bad sequence and wind up with a single really really bad sequence.

You are probably right this is still better though. One way to test that would be to see if simulation results vary much if you pick different length strings, one year, three years, five years, 10 years, etc, including a 30 year string. That would tell you how sensitive the results are to string length. Which would kinda be an interesting study all by itself.
I think the authors would argue that there might be a chance of two bad sequences happening back-to-back (though they don't know what that probability is) and they want to capture this. If one was sure that this could never happen, then one could embed that reversion to the mean logic/probability in a Monte Carlo simulation. The catch with excluding it is that if the chance IS REAL and non-zero then excluding it gives us a false sense of security.

Note also that this entire discussion is really just another variation of, "We don't have enough sample data to be very confident in our conclusions."

The Trinity study worked with data from 1926. At the time (1995?) this was fewer than three independent 30-year returns. There just isn't that much you can conclude from three independ data points. Extending the data to the rest of the world brings its own problems:
  • The markets in the rest of the world is correlated with the US and with each other so you don't get fully independent new data, and
  • You have to decide what to do about WW1, WW2, Argentina, ...
The error bars (if one does the stats carefully) are just larger than most people want them to be and this leads to 95% confidence withdrawal rates that are depressingly low. Because we just don't know!
It's not just the stats. Are market returns from say 1900 (any country) relevant to a discussion of what might happen next year? That's not clear to me. We don't just lack data points; we lack an understanding of which data points we have are relevant to forward looking projections. A simulation with irrelevant data is not useful.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Maybe Wade Pfau was over-optimistic …

Post by vineviz »

randomguy wrote: Sun Oct 02, 2022 7:01 pm
MarkRoulo wrote: Sun Oct 02, 2022 4:09 pm
I think the authors would argue that there might be a chance of two bad sequences happening back-to-back (though they don't know what that probability is) and they want to capture this. If one was sure that this could never happen, then one could embed that reversion to the mean logic/probability in a Monte Carlo simulation. The catch with excluding it is that if the chance IS REAL and non-zero then excluding it gives us a false sense of security.
On the other hand by not having mean reversion, they give a false sense of pessimism.
Except that the methodology they used DOES include whatever mean reversion the markets exhibit: that's the point of the 10-year block bootstrapping.

From the paper:
Block lengths are drawn from a geometric distribution with an average block length of 120 months. These long blocks allow the simulation to reflect short-term characteristics such as persistent return volatility and long-term characteristics like mean reversion in stock returns.
In any event, in practical terms it's a relatively unimportant factor. At a 5th or 10th percentile level, the presence or absence of realistic levels of mean reversion is nearly invisible in the outputs.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
9-5 Suited
Posts: 1306
Joined: Thu Jun 23, 2016 12:14 pm

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by 9-5 Suited »

Alrighty, so I fully acknowledge that this may be a 'hot take' of sorts. But, as a practical matter, I have simply chosen to ignore these continued attempts at reframing the "safe withdrawal rate". In a way similar to the constant search for more factors which explain returns, they are beginning to feel like academic pursuits that exist primarily to bring attention and career success to the authors.

They are so ridiculously one-directional, biased heavily in favor of ever-lower withdrawal rates. Some may say it's unwise, but I just don't care to adjust my plans because the losers of World Wars had terrible outcomes or because of what happened to 1897 retirees.

Further, as I get closer and closer to my own early retirement - now a month away - I already realize how irrelevant these "SWR" figures are beyond being rough guidelines. We will never withdraw the exact same amount, inflation-adjusted, for 50 years. My wife still has work plans. I'm already being approached for part-time consulting opportunities and figuring out a new, passion-based career option I hadn't previously thought a lot about. We have a gigantic amount of flexible expenses in our budget, and by virtue of me not working will be able to save even more than expected on some other budget items.

I'm just absolutely done with this pessimistic stuff. I love Morgan Housel's comments about how pessimism looks so "wise" and optimism looks so foolish, but it is the optimists who win in almost every corner of life including investing.

I guess take a 1.9% withdrawal rate if you wish - I promise I will not be, and I feel not a single ounce of regret about it.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Maybe Wade Pfau was over-optimistic … [New study on Safe Withdrawal Rate]

Post by vineviz »

9-5 Suited wrote: Sun Oct 02, 2022 8:34 pm I guess take a 1.9% withdrawal rate if you wish - I promise I will not be, and I feel not a single ounce of regret about it.
I don't think retirement researcher has "make "9-5 Suited" feel regret" as a goal.

The goal is pretty much simply to help investors make sensible plans for their retirement, by providing evidence about what is likely to work and what is unlikely to work.

It's not manufactured pessimism, but rather an acknowledgement of the reality of the situation we face.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Post Reply