Traditional SWR - flawed at the core

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

willthrill81 wrote: Sat Jan 22, 2022 12:08 pm
nigel_ht wrote: Sat Jan 22, 2022 12:06 pm The only thing SWR tells you is that historically you won’t deplete your portfolio if you spend X% Or LESS.
No, it doesn't tell you this, and no metric can. It tells you that your portfolio would not have been prematurely depleted in worst historic period of the PAST.

Without a crystal ball, no one knows what the forward SWR of any strategy will be.
I said the same thing you did in a slightly different way. What is different from saying “historically you won’t deplete your portfolio” and “would not have been prematurely depleted in worst historic period of the PAST”?

I even used wording closer to yours later in the same post:

“It’s useful to know that your intended withdrawal rate is higher than what would have survived the historical worst case scenario.”
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Traditional SWR - flawed at the core

Post by willthrill81 »

nigel_ht wrote: Sat Jan 22, 2022 12:12 pm
willthrill81 wrote: Sat Jan 22, 2022 12:08 pm
nigel_ht wrote: Sat Jan 22, 2022 12:06 pm The only thing SWR tells you is that historically you won’t deplete your portfolio if you spend X% Or LESS.
No, it doesn't tell you this, and no metric can. It tells you that your portfolio would not have been prematurely depleted in worst historic period of the PAST.

Without a crystal ball, no one knows what the forward SWR of any strategy will be.
I said the same thing you did in a slightly different way. What is different from saying “historically you won’t deplete your portfolio” and “would not have been prematurely depleted in worst historic period of the PAST”?
I glossed over your use of the word 'historically'. My apologies. I was focusing on your use of 'won't' (i.e., 'will not'), which has to do with the future.
The Sensible Steward
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

nigel_ht wrote: Sat Jan 22, 2022 12:10 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:52 am
randomguy wrote: Sat Jan 22, 2022 10:36 am
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am

You make a fair point about the horizon not being constant between the two investors in the OP. The problem would be easy to show if we had different SWR for different horizons. But if we have only a 30 year SWR, it's not possible to keep things equal between the two people retiring at different dates. But we can keep things almost equal by reducing the time between the two retirement dates. Consider the following modification of the scenarios in OP:

Investor 3: Retired 1 week ago at 60 with $2M 60/40 portfolio and 4% withdrawal. Markets crashed last week and his portfolio is down to $1.5M. Assume 0% inflation for simplicity and he is happily withdrawing $80,000 per year ($2M x 4%) and considers that appropriate.

Investor 4: Retires today at 60 with $1.5M 60/40 portfolio and 4% withdrawal. He is happily withdrawing $60,000 per year ($1.5M x 4%) and considers that appropriate.

Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
Investor 1 is taking on a 5% chance of failure. Investor 4 is taking on a 0% chance. Less risk, less money..
Both followed the same methodology to arrive at their withdrawal amount. Why is the methodology giving different risk/rewards to different people when they didn't ask for it?
Both have 5% failure since that’s what 4% over 30 years represents.
Both have almost the same portfolio and horizon, but very different withdrawal amounts. So they now have different failure rates.

They may have had the same failure rates when they each started. But that's ancient history. Why does it matter? The investor withdrawing $80K now has a higher failure rate (>5%) than the investor withdrawing $60K (5%).
Total Portfolio Allocation and Withdrawal (TPAW)
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

willthrill81 wrote: Sat Jan 22, 2022 12:17 pm
nigel_ht wrote: Sat Jan 22, 2022 12:12 pm
willthrill81 wrote: Sat Jan 22, 2022 12:08 pm
nigel_ht wrote: Sat Jan 22, 2022 12:06 pm The only thing SWR tells you is that historically you won’t deplete your portfolio if you spend X% Or LESS.
No, it doesn't tell you this, and no metric can. It tells you that your portfolio would not have been prematurely depleted in worst historic period of the PAST.

Without a crystal ball, no one knows what the forward SWR of any strategy will be.
I said the same thing you did in a slightly different way. What is different from saying “historically you won’t deplete your portfolio” and “would not have been prematurely depleted in worst historic period of the PAST”?
I glossed over your use of the word 'historically'. My apologies. I was focusing on your use of 'won't' (i.e., 'will not'), which has to do with the future.
Lol…as one of the more vocal SWR defenders around here I would never hear the end of overselling the capabilities of SWR…
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Traditional SWR - flawed at the core

Post by randomguy »

nigel_ht wrote: Sat Jan 22, 2022 12:10 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:52 am
randomguy wrote: Sat Jan 22, 2022 10:36 am
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am

You make a fair point about the horizon not being constant between the two investors in the OP. The problem would be easy to show if we had different SWR for different horizons. But if we have only a 30 year SWR, it's not possible to keep things equal between the two people retiring at different dates. But we can keep things almost equal by reducing the time between the two retirement dates. Consider the following modification of the scenarios in OP:

Investor 3: Retired 1 week ago at 60 with $2M 60/40 portfolio and 4% withdrawal. Markets crashed last week and his portfolio is down to $1.5M. Assume 0% inflation for simplicity and he is happily withdrawing $80,000 per year ($2M x 4%) and considers that appropriate.

Investor 4: Retires today at 60 with $1.5M 60/40 portfolio and 4% withdrawal. He is happily withdrawing $60,000 per year ($1.5M x 4%) and considers that appropriate.

Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
Investor 1 is taking on a 5% chance of failure. Investor 4 is taking on a 0% chance. Less risk, less money..
Both followed the same methodology to arrive at their withdrawal amount. Why is the methodology giving different risk/rewards to different people when they didn't ask for it?
Both have 5% failure since that’s what 4% over 30 years represents.
No they are using the same methodology but different data sets. Imagine there are 10 people in a room and 1 is 6' tall and the rest are 5'10. What are your odds of picking out a 6'er? 1/10. Now remove the 6' from the room. Is your odds of picking a 6'er 10% (original dataset) or 0% (new data set). Obviously 0%. Investor 4 has the added info that the market just dropped ~50%. That means the 4% rule calculated without that constraint doesn't apply. You need to calculate the SWR using only the data that accounts for the market drop the week before. Again this is all backed up with studies that show that the SWR during low valuation periods isn't 4%. It is higher.

This also works the other way. When week 1 starts is Investor 3's failure rate 5%? Heck no. The failure rate after a 50% market drop is probably up around 25%+. The risk of failure was 5% when everything started but with the new info it is higher since we have gone from a world of a lot of possibilities (market up 20%, market flat) to the worst outcome . Investor 4 is faced with the choice to match investor 1's new risk of failure (take out 80k and have a 25% risk of failure) or adopt a lower one (40k with 0% risk of failure).
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

Ben Mathew wrote: Sat Jan 22, 2022 12:22 pm
nigel_ht wrote: Sat Jan 22, 2022 12:10 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:52 am
randomguy wrote: Sat Jan 22, 2022 10:36 am
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am

You make a fair point about the horizon not being constant between the two investors in the OP. The problem would be easy to show if we had different SWR for different horizons. But if we have only a 30 year SWR, it's not possible to keep things equal between the two people retiring at different dates. But we can keep things almost equal by reducing the time between the two retirement dates. Consider the following modification of the scenarios in OP:

Investor 3: Retired 1 week ago at 60 with $2M 60/40 portfolio and 4% withdrawal. Markets crashed last week and his portfolio is down to $1.5M. Assume 0% inflation for simplicity and he is happily withdrawing $80,000 per year ($2M x 4%) and considers that appropriate.

Investor 4: Retires today at 60 with $1.5M 60/40 portfolio and 4% withdrawal. He is happily withdrawing $60,000 per year ($1.5M x 4%) and considers that appropriate.

Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
Investor 1 is taking on a 5% chance of failure. Investor 4 is taking on a 0% chance. Less risk, less money..
Both followed the same methodology to arrive at their withdrawal amount. Why is the methodology giving different risk/rewards to different people when they didn't ask for it?
Both have 5% failure since that’s what 4% over 30 years represents.
Both have almost the same portfolio and horizon, but very different withdrawal amounts. So they now have different failure rates.

They may have had the same failure rates when they each started. But that's ancient history. Why does it matter? The investor withdrawing $80K now has a higher failure rate (>5%) than the investor withdrawing $60K (5%).
Nope. Historically both have the same failure rate.

Now that doesn’t tell you that investor 1 might have run out of money at year 25 and investor 2 ran out of money at year 29.5.

The first is obviously worse than the second one.

But still it’s an apples to oranges comparison. If both have a $60K yearly spend they have the same outcomes.

And again, if they both have $80K a year spends the second person can’t retire.

SWR is a ceiling and not a floor.

Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Traditional SWR - flawed at the core

Post by Marseille07 »

nigel_ht wrote: Sat Jan 22, 2022 12:30 pm Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Imo Ben's example is contrived because the granularity of SWR study is per year. IOW it doesn't account for a scenario Ben describes, where Investor 3 retires with 2M and a week later their portfolio is 1.5M.

in order to compare apples to apples, Investor 4 needs to retire a year later.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

randomguy wrote: Sat Jan 22, 2022 12:23 pm
nigel_ht wrote: Sat Jan 22, 2022 12:10 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:52 am
randomguy wrote: Sat Jan 22, 2022 10:36 am
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am

You make a fair point about the horizon not being constant between the two investors in the OP. The problem would be easy to show if we had different SWR for different horizons. But if we have only a 30 year SWR, it's not possible to keep things equal between the two people retiring at different dates. But we can keep things almost equal by reducing the time between the two retirement dates. Consider the following modification of the scenarios in OP:

Investor 3: Retired 1 week ago at 60 with $2M 60/40 portfolio and 4% withdrawal. Markets crashed last week and his portfolio is down to $1.5M. Assume 0% inflation for simplicity and he is happily withdrawing $80,000 per year ($2M x 4%) and considers that appropriate.

Investor 4: Retires today at 60 with $1.5M 60/40 portfolio and 4% withdrawal. He is happily withdrawing $60,000 per year ($1.5M x 4%) and considers that appropriate.

Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
Investor 1 is taking on a 5% chance of failure. Investor 4 is taking on a 0% chance. Less risk, less money..
Both followed the same methodology to arrive at their withdrawal amount. Why is the methodology giving different risk/rewards to different people when they didn't ask for it?
Both have 5% failure since that’s what 4% over 30 years represents.
No they are using the same methodology but different data sets. Imagine there are 10 people in a room and 1 is 6' tall and the rest are 5'10. What are your odds of picking out a 6'er? 1/10. Now remove the 6' from the room. Is your odds of picking a 6'er 10% (original dataset) or 0% (new data set). Obviously 0%. Investor 4 has the added info that the market just dropped ~50%. That means the 4% rule calculated without that constraint doesn't apply. You need to calculate the SWR using only the data that accounts for the market drop the week before. Again this is all backed up with studies that show that the SWR during low valuation periods isn't 4%. It is higher.

This also works the other way. When week 1 starts is Investor 3's failure rate 5%? Heck no. The failure rate after a 50% market drop is probably up around 25%+. The risk of failure was 5% when everything started but with the new info it is higher since we have gone from a world of a lot of possibilities (market up 20%, market flat) to the worst outcome . Investor 4 is faced with the choice to match investor 1's new risk of failure (take out 80k and have a 25% risk of failure) or adopt a lower one (40k with 0% risk of failure).
It helps if everyone understands that failure rate is running out of money anytime before then end of the computed period (30 years for many).

That means failure on day 1 is equivalent to failure on day 10,949.

Just looking at the overall failure percentage doesn’t give you the full picture but given the future won’t be exactly like the past I shoot for 0% failure and assume that my estimates of future expenses are far more likely to occur than being unlucky enough to experience a worse than worst case scenario.
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Traditional SWR - flawed at the core

Post by randomguy »

Marseille07 wrote: Sat Jan 22, 2022 12:32 pm
nigel_ht wrote: Sat Jan 22, 2022 12:30 pm Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Imo Ben's example is contrived because the granularity of SWR study is per year. IOW it doesn't account for a scenario Ben describes, where Investor 3 retires with 2M and a week later their portfolio is 1.5M.

in order to compare apples to apples, Investor 4 needs to retire a year later.
The time thing is a distraction. The SWR for 25 year periods just isn't that different from 30. It isn't remotely enough to account for the difference.

It is an information problem. After the markets drop 50% (either in 1 week or 5 years), your odds of failure for the remaining period are not 5%. They are 25%+ since a lot of the good outcome paths are no longer available. The second person needs to increase their SWR well above 4% to even the risk out.

Even if you believe that things are totally random in the market (they aren't) imagine that to have portfolio failure you need to roll double 6's on a 2 dice . what are the odds? 3%. Now imagine you roll a 6 (i.e. the market dropped 40%). Your partner then joins you and you will share the next dice roll and then they get another roll. What are the odds of failure? Your odds are 1/6 (you have already lost the first dice roll). Your partner though is 1/36.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

randomguy wrote: Sat Jan 22, 2022 1:11 pm
Marseille07 wrote: Sat Jan 22, 2022 12:32 pm
nigel_ht wrote: Sat Jan 22, 2022 12:30 pm Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Imo Ben's example is contrived because the granularity of SWR study is per year. IOW it doesn't account for a scenario Ben describes, where Investor 3 retires with 2M and a week later their portfolio is 1.5M.

in order to compare apples to apples, Investor 4 needs to retire a year later.
The time thing is a distraction. The SWR for 25 year periods just isn't that different from 30. It isn't remotely enough to account for the difference.
The difference is 5 years of spending at $80K…$400,000.

It’s a huge difference.

For a 25 year period at $80K a year you need $2M.

ONE year makes a difference in outcome because it’s a double bonus…one less year of spending and one more year of saving and growth.
It is an information problem. After the markets drop 50% (either in 1 week or 5 years), your odds of failure for the remaining period are not 5%.
Yep they are…because we aren’t rolling dice (aka Monte Carlo) but looking at historical results.

1929 was followed by 1932. That’s why it’s one of the two worst case scenarios. The events are historically linked and occur together 100% of the time.

If you use some sort of Monte Carlo as the basis of computing SWR you’d end up with a different number.
Even if you believe that things are totally random in the market (they aren't) imagine that to have portfolio failure you need to roll double 6's on a 2 dice . what are the odds? 3%. Now imagine you roll a 6 (i.e. the market dropped 40%). Your partner then joins you and you will share the next dice roll and then they get another roll. What are the odds of failure? Your odds are 1/6 (you have already lost the first dice roll). Your partner though is 1/36.
Your partner already rolled a 6 unless they weren’t in the market so the odds are the same.

If you can generate $1.5M in a week you don’t need to worry about SWR…
Chico78
Posts: 26
Joined: Thu Mar 14, 2019 12:53 pm

Re: Traditional SWR - flawed at the core

Post by Chico78 »

I’m the (only?) guy that actually has used the 4% “rule” in retirement……but it’s been a good four years and I do an annual reasonableness check. I compare it to VPW, ABW, RMD, CAPE and a reset to 4% of my current portfolio value. So far so good.
We always seem to talk about retirement as if it's a rifle shot(a dumb ballistic trajectory); compute all the variables, pull the trigger and it’s a hit or a miss. I think of it more like an aircraft flight (monitorable and adjustable); come up with a good flight plan, do a Howgoezit/ reasonableness check and adjust it when and if you need to. YMMV.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Traditional SWR - flawed at the core

Post by Marseille07 »

Chico78 wrote: Sat Jan 22, 2022 3:19 pm I’m the (only?) guy that actually has used the 4% “rule” in retirement……but it’s been a good four years and I do an annual reasonableness check. I compare it to VPW, ABW, RMD, CAPE and a reset to 4% of my current portfolio value. So far so good.
We always seem to talk about retirement as if it's a rifle shot(a dumb ballistic trajectory); compute all the variables, pull the trigger and it’s a hit or a miss. I think of it more like an aircraft flight (monitorable and adjustable); come up with a good flight plan, do a Howgoezit/ reasonableness check and adjust it when and if you need to. YMMV.
Well, if you reset to 4% of your current portfolio value then you're talking about constant-% WR rather than SWR.

SWR is unique because it gives you a spending ceiling that you more or less know in advance. Percentage-based methods don't give you that.
Chico78
Posts: 26
Joined: Thu Mar 14, 2019 12:53 pm

Re: Traditional SWR - flawed at the core

Post by Chico78 »

Nope , I compare it to a 4% of existing. I can of course adjust at any time.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Traditional SWR - flawed at the core

Post by Marseille07 »

Chico78 wrote: Sat Jan 22, 2022 3:27 pm Nope , I compare it to a 4% of existing. I can of course adjust at any time.
I don't follow. Withdrawing 4% of your remaining portfolio balance *is* constant-%. SWR doesn't work that way.
Chico78
Posts: 26
Joined: Thu Mar 14, 2019 12:53 pm

Re: Traditional SWR - flawed at the core

Post by Chico78 »

Sorry, I used 4% plus inflation, the “rule” , and compare it to the other methods including the 4% reset as a reasonableness check.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

Chico78 wrote: Sat Jan 22, 2022 3:27 pm Nope , I compare it to a 4% of existing. I can of course adjust at any time.
The difference is using SWR vs what you are doing is that under SWR for every million at the start you can pull up to around $40K a year.

Even if 1929 happens again* (89%) and your portfolio value craters.

Using constant percentage you wouldn’t be able to.

—-

* it would have to happen again with deflation and pretty much everything else the same. It won’t so the next Big One will be different so SWR is only HISTORICALLY safe. Still pretty safe tho’. You are far more likely to seriously misjudge your expenses 20 years from now than to encounter Depression 2.0.
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Traditional SWR - flawed at the core

Post by randomguy »

nigel_ht wrote: Sat Jan 22, 2022 2:45 pm
randomguy wrote: Sat Jan 22, 2022 1:11 pm
Marseille07 wrote: Sat Jan 22, 2022 12:32 pm
nigel_ht wrote: Sat Jan 22, 2022 12:30 pm Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Imo Ben's example is contrived because the granularity of SWR study is per year. IOW it doesn't account for a scenario Ben describes, where Investor 3 retires with 2M and a week later their portfolio is 1.5M.

in order to compare apples to apples, Investor 4 needs to retire a year later.
The time thing is a distraction. The SWR for 25 year periods just isn't that different from 30. It isn't remotely enough to account for the difference.
The difference is 5 years of spending at $80K…$400,000.

It’s a huge difference.

For a 25 year period at $80K a year you need $2M.

ONE year makes a difference in outcome because it’s a double bonus…one less year of spending and one more year of saving and growth.
It is an information problem. After the markets drop 50% (either in 1 week or 5 years), your odds of failure for the remaining period are not 5%.
Yep they are…because we aren’t rolling dice (aka Monte Carlo) but looking at historical results.

1929 was followed by 1932. That’s why it’s one of the two worst case scenarios. The events are historically linked and occur together 100% of the time.

If you use some sort of Monte Carlo as the basis of computing SWR you’d end up with a different number.
Even if you believe that things are totally random in the market (they aren't) imagine that to have portfolio failure you need to roll double 6's on a 2 dice . what are the odds? 3%. Now imagine you roll a 6 (i.e. the market dropped 40%). Your partner then joins you and you will share the next dice roll and then they get another roll. What are the odds of failure? Your odds are 1/6 (you have already lost the first dice roll). Your partner though is 1/36.
Your partner already rolled a 6 unless they weren’t in the market so the odds are the same.

If you can generate $1.5M in a week you don’t need to worry about SWR…
-The SWR for 30 years is 4% (95% success). The SWR for 25 years is 4.3%. I consider that a minimal difference. The question is why our guy can only take out 60k instead of 80k. Saving they can take out 65k explains a tiny portion of the difference.

- Yep and that is why your are in trouble. This isn't exactly right but it is close enough to demonstrate the point. There are 100 chances when you retire. 5 of them fail. 75 times the market is up in year 1, 25% the market is down. If the market is up you never fail. If your market is down after year 1, what are your failure chances over the next 29 years? You have gone from 5% (5/100) to 25% (5/25) once you learn that year 1 is a down year.

-Nope. By the rules of the game to go broke if I need to roll one more 6. My partner needs to roll 2 6's to go broke. One is clearly much more likely than the other. If my partner took out 80k instead of 60k, they would have been playing from the start. When they chose to take out 60k they are starting a new game.

Again this is all grossly simplified to make it clearer. You can't get too crazy with historical data. We just don't have enough of it. Just because if the market was up in year 1 and never failed might not transition well in the future. The difference between being up +3% or down 3% is in the noise. But there are general trends of what is a bad come (down or flat markets for 5 years) and what are good outcomes (i.e. you make the normal 7%+). You can see how the lack of data shows up when we figure out portfolio failure rates. 1929 works if you start in Jan. That big bull market the first 9 months of the year gives you a big cushion. Start every month and now October 1929 is a failure case. Details of methodology matter (are you looking at years, months, or days, taking money out Jan 1, Dec 31, every month, and exact holdings) when looking at these edge cases.
JackoC
Posts: 4714
Joined: Sun Aug 12, 2018 11:14 am

Re: Traditional SWR - flawed at the core

Post by JackoC »

Ben Mathew wrote: Sat Jan 22, 2022 11:42 am
JackoC wrote: Sat Jan 22, 2022 10:03 am
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am
JSNO wrote: Sat Jan 22, 2022 12:18 am
Ben Mathew wrote: Fri Jan 21, 2022 6:34 pm Yes, that's how the methodology works. And that methodology results in the following contradiction: two people in identical situations are being prescribed different things based on a history that is irrelevant.
If as of today, both people are planning to live for another 30 years, then there is a contradiction because both people still need to fund 30 years of retirement.

However, both people planning to live another 30 years from today violates the assumptions underlying the 4% SWR as laid out in the Trinity study.

The contradiction is not inherent in the SWR.
You make a fair point about the horizon not being constant between the two investors in the OP. The problem would be easy to show if we had different SWR for different horizons. But if we have only a 30 year SWR, it's not possible to keep things equal between the two people retiring at different dates. But we can keep things almost equal by reducing the time between the two retirement dates. Consider the following modification of the scenarios in OP:
The flaw then would seem to be the application of the same 'received wisdom' SWR for different lengths of retirement. *That's* a contradiction.
You seem to be evaluating SWR only as a tool to tell you when to retire. You are not considering SWR as a retirement spending strategy. The contradiction OP was trying to get at is from using SWR as an actual retirement spending strategy, sticking to a withdrawal calculated at the start of retirement.

But even as a tool for evaluating when to retire, the SWR methodology (though not the 4% estimate) is pretty bad. The 5% failure rate leads to very conservative withdrawals, and would make people work far longer than they need to.

You might say, well we can easily fix that by allowing a 10% failure rate, or 20%. It can be made less conservative. True, but the problem is that it's hard to evaluate what percentage to use for a reasonable estimate. That's because SWR is assuming you won't adjust and so it's not creating the right picture to base our decisions on. If you will adjust, then you're not going to run out of money.
JackoC wrote: Sat Jan 22, 2022 10:03 am But my simple (simplistic?) point, if *at* retirement you say 'Trinity Study, 30yrs, 4%, I have it, I'm retiring' or 'ABW, I'll assume 1.2% real after tax return, 30 yrs, 4%, I have it, I'm retiring', what's the 'contradiction'? And if the ABW person assumes significantly more than 1.2%, as is practically likely, they'll decide to retire sooner. :happy
If ABW says someone can take out $40,000 if they retired this year, and $40,000 is about sufficient, that person shouldn't retire yet! If g=0%, then there's roughly a 50% chance that their spending will fall below $40,000 next year. They will need a cushion. Entering lower than expected returns as you did above is a simple way to accomplish this. Or one can take a more explicitly probabilistic approach like the ABW simulator...
JackoC wrote: Sat Jan 22, 2022 10:03 am [...] using the distribution of past returns as if it's 'the' stationary distribution of returns irrespective of starting valuations. That's a clearly optimistic assumption IMO at historically high starting asset valuations like now. Those valuations don't mean 'crash coming, but you can avoid it by listening to my tactical market calls!', it means unknown future realized return, I have no tactical market call to give, but future returns are reasonably assumed centered around a lower centroid, ie the expected return, than average past return. Therefore '% success' numbers from past studies are too optimistic unless you posit the expected variance is also lower than average past variance, a heroic assumption I'm not willing to make though I can't prove it's untrue.
Fully agree.
I'm glad you agree with last point but it seems perhaps at odds with the earlier statement that a 5% failure rate would lead to excessively conservative withdrawal rate. Or I could be misunderstanding that first point. If you use a realistic distribution of future returns, centroid IMO 2-3% pa below that of the long term average past return then aim for 5% 'failure rate' I agree that's too low, because IMO it's out of whack with all the other bad things that can happen (dying, dementia so that you don't even know your living standard, etc.) that are cumulatively much more likely. But that's entirely in the realm of opinion including general outlook beyond finance. OTOH if you use past returns data (like FIREcalc) to estimate 5% failure rate the real failure rate is probably much higher than that on the future return distribution. I think 4% SWR is aggressive not conservative for a healthy person in the 'professional/managerial class', predominant here, planning to retire before the traditional 65 as most in that class can afford to consider voluntarily doing (plenty of people in lower social classes retire before 65 but often due to health, unemployability or sometimes generous traditional pensions). 4% very likely to fail? I don't think so; but no way would I have quit when I had only that much money.

I do agree it would be nuts to pick an SWR either from a study or the result of a pre-retirement ABW calculation and never reconsider it during retirement regardless of realized returns or spending pattern changes. That seems obvious, but as has been said umpteen times on this thread and so many others, who really does that? The key variable seems to me is how much money you have to have before voluntarily retiring, which is where I see no particular advantage to SWR derived from a (n IMO outdated) study vs. ABW using too high a return assumption. I do also agree the ABW method is more flexible pre-retirement to put in a return that's appropriately low. But getting everybody to agree what that is? :happy Also my general opinion is that the more black-boxy sophisticated the tool, the more likely people are to overlook basic flaw in the method as with FIREcalc. I bet there's some tool that lets you simulate with all your own inputs, but I'm not sure it's better than just assuming no real after tax return so initial SWR low 3's from early 60's or so. But if that results in 'wasting too much of your life' working, then assume something higher. In which case even the outdated studies or FIREcalc are *some* indication of what's not way too high. And I'm not saying 4% starting point is way too high, if willing to fairly likely have to make downward adjustments, or count on natural decay of real spending. It's not what I did or would do, risk preference is personal. Plus here we add in spending preference relative to how much one managed to accumulate as of the time ending regular day job became seriously attractive.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

JackoC wrote: Sat Jan 22, 2022 4:48 pm And I'm not saying 4% starting point is way too high, if willing to fairly likely have to make downward adjustments, or count on natural decay of real spending.
Basic SWR ignores SS.

That’s not as big a factor for early FIRE but for the mid 50’s barely early retirement it would.

So there’s quite a bit of margin built into “4% covers expenses” as a retirement threshold.
JackoC
Posts: 4714
Joined: Sun Aug 12, 2018 11:14 am

Re: Traditional SWR - flawed at the core

Post by JackoC »

nigel_ht wrote: Sat Jan 22, 2022 5:12 pm
JackoC wrote: Sat Jan 22, 2022 4:48 pm And I'm not saying 4% starting point is way too high, if willing to fairly likely have to make downward adjustments, or count on natural decay of real spending.
Basic SWR ignores SS.

That’s not as big a factor for early FIRE but for the mid 50’s barely early retirement it would.

So there’s quite a bit of margin built into “4% covers expenses” as a retirement threshold.
I mentioned that last post. There are many people retiring at/near 'normal' age where SS (and/or a pension) covers most of their needs, but they aren't including it in the PV that determines SWR. This makes the risk of choosing a wrong SWR on their 'other money' a smaller to in the limit trivial issue depending how adequate SS by itself is. But for some people living on SS alone would be a big reduction in living standard (living just on SS is not 'beneath us' but it would be a big come down and that matters to us). And like you said people looking to retire quite early have to last a long time without SS. But lots of discussions here involve people talking past each other due to combination of different personal circumstances and lack of standard definitions of 'what counts'.
Last edited by JackoC on Sat Jan 22, 2022 5:29 pm, edited 1 time in total.
BogleFan510
Posts: 1039
Joined: Tue Aug 04, 2020 2:13 pm

Re: Traditional SWR - flawed at the core

Post by BogleFan510 »

This post represents the danger of only reading a headline 'e.g. 4% is a "safe rate"' instead of actually taking the time to read and understand the paper that led to the conclusions. The paper is and was never meant to dictate any retirement withdraw rate absolutes.

To call it 'flawed at the core' as a headline is a clickbait style I find offputting and not helpful to investors planning for retirement. Explaining sequence of returns risk and the assumptions regarding retirement duration, as a concept, are much more helpful to people than scare tactic headlines.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

JackoC wrote: Sat Jan 22, 2022 5:23 pm
nigel_ht wrote: Sat Jan 22, 2022 5:12 pm
JackoC wrote: Sat Jan 22, 2022 4:48 pm And I'm not saying 4% starting point is way too high, if willing to fairly likely have to make downward adjustments, or count on natural decay of real spending.
Basic SWR ignores SS.

That’s not as big a factor for early FIRE but for the mid 50’s barely early retirement it would.

So there’s quite a bit of margin built into “4% covers expenses” as a retirement threshold.
I mentioned that last post. There are many people retiring at/near 'normal' age where SS (and/or a pension) covers most of their needs, but they aren't including it in the PV that determines SWR. This makes the risk of choosing a wrong SWR on their 'other money' a smaller to in the limit trivial issue depending how adequate SS by itself is. But for some people living on SS alone would be a big reduction in living standard (living just on SS is not 'beneath us' but it would be a big come down and that matters to us). And like you said people looking to retire quite early have to last a long time without SS. But lots of discussions here involve people talking past each other due to combination of different personal circumstances and lack of standard definitions of 'what counts'.
But there is no “wrong SWR”…the SWR is based on duration of the retirement not your needs.

Whether your needs are met with a withdrawal rate at or below SWR informs you whether you have enough to survive the historical worst cases.

Using the computed SWR for your retirement duration is mostly safe when trying to determine if it is safe to FIRE. Generally that becomes 33X expenses vs 25X expenses.

SS can be used as a hedge for either worse than worst case or, more likely, you screwed up your estimate on expenses OR you can include that in the calculation of how much withdrawals you need to meet income to tell you if it is safe to retire with a low risk of depletion based on historical backrests.

If SS meets the majority or all of your needs then the only withdrawal strategy you need is “spend what you want but don’t go crazy”.

Forget SWR, ABW, VPW etc.
tibbitts
Posts: 23716
Joined: Tue Feb 27, 2007 5:50 pm

Re: Traditional SWR - flawed at the core

Post by tibbitts »

nigel_ht wrote: Sat Jan 22, 2022 5:12 pm Basic SWR ignores SS.
Basic SWR does, but so does super-ultra-advanced SWR. It also ignores other pensions, employer supplements for health insurance, and countless other factors that might increase or decrease the need for withdrawals from a portfolio. SS just doesn't belong is a SWR discussion.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

tibbitts wrote: Sat Jan 22, 2022 7:41 pm
nigel_ht wrote: Sat Jan 22, 2022 5:12 pm Basic SWR ignores SS.
Basic SWR does, but so does super-ultra-advanced SWR. It also ignores other pensions, employer supplements for health insurance, and countless other factors that might increase or decrease the need for withdrawals from a portfolio. SS just doesn't belong is a SWR discussion.
If you are doing apples to apples comparisons with VPW and other methods you need to consider it if they do or it’s an uneven playing field.

When you factor all the assumptions/requirements in then the difference in annual budget isn’t quite as high as some folks believe.

Plus if you use tools like cFIREsim switching withdrawal methods is a menu selection. Considering SWR with the same SS as something else becomes simpler.

So why wouldn’t “Advanced SWR” consider bridge to SS or SPIAs or pensions?
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Traditional SWR - flawed at the core

Post by randomguy »

nigel_ht wrote: Sat Jan 22, 2022 8:07 pm
tibbitts wrote: Sat Jan 22, 2022 7:41 pm
nigel_ht wrote: Sat Jan 22, 2022 5:12 pm Basic SWR ignores SS.
Basic SWR does, but so does super-ultra-advanced SWR. It also ignores other pensions, employer supplements for health insurance, and countless other factors that might increase or decrease the need for withdrawals from a portfolio. SS just doesn't belong is a SWR discussion.
If you are doing apples to apples comparisons with VPW and other methods you need to consider it if they do or it’s an uneven playing field.

When you factor all the assumptions/requirements in then the difference in annual budget isn’t quite as high as some folks believe.

Plus if you use tools like cFIREsim switching withdrawal methods is a menu selection. Considering SWR with the same SS as something else becomes simpler.

So why wouldn’t “Advanced SWR” consider bridge to SS or SPIAs or pensions?
VPW doesn't change with SS. The VPW folks just like to include SS to make the volatility seem less...
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Traditional SWR - flawed at the core

Post by vineviz »

Ben Mathew wrote: Sat Jan 22, 2022 1:26 am Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
SWR was never a methodology. It is simply a guideline and a model. Like any model, it will break if you push the assumptions hard enough.

So I would say that YOUR version of SWR is "anchoring on starting portfolio value and not adjusting", but that need not be universally true.

For one thing, many practitioners will use the average portfolio value over some period of time to avoid precisely the conundrum you illustrated of having the initial withdrawal amount be tied to the portfolio balance on a particular day. It's common to use the average value over the course of a year, for example.

A second difficulty with your example is that your two investors are using the same withdrawal rate despite having different expected returns. The long-run expected return of stocks is not the same before a 40+% crash as it is after that crash, and the initial withdrawal rate should take that into account.

SWR is a simple heuristic device that allows us to readily make sense of a reality that is infinitely complicated. All models have flaws, but many of those models are useful BECAUSE of those flaws instead of despite them.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

JackoC wrote: Sat Jan 22, 2022 4:48 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:42 am But even as a tool for evaluating when to retire, the SWR methodology (though not the 4% estimate) is pretty bad. The 5% failure rate leads to very conservative withdrawals, and would make people work far longer than they need to.

[...]
JackoC wrote: Sat Jan 22, 2022 10:03 am [...] using the distribution of past returns as if it's 'the' stationary distribution of returns irrespective of starting valuations. That's a clearly optimistic assumption IMO at historically high starting asset valuations like now. Those valuations don't mean 'crash coming, but you can avoid it by listening to my tactical market calls!', it means unknown future realized return, I have no tactical market call to give, but future returns are reasonably assumed centered around a lower centroid, ie the expected return, than average past return. Therefore '% success' numbers from past studies are too optimistic unless you posit the expected variance is also lower than average past variance, a heroic assumption I'm not willing to make though I can't prove it's untrue.
Fully agree.
I'm glad you agree with last point but it seems perhaps at odds with the earlier statement that a 5% failure rate would lead to excessively conservative withdrawal rate.
5% failure rate would lead to excessively conservative withdrawal rates if applied to a forward looking return distribution based on current valuations. The only reason SWR 5% failure rate is currently producing a fairly reasonable 4% withdrawal rate is that the historical returns being used is much higher. This just happens to be a lucky period where SWR produces a reasonable estimate because its inherent conservativeness is being canceled by an optimistic return distribution. If it had been a time where historical returns were a good estimate of the future distribution, the SWR @ 5% failure rate would produce an extremely conservative withdrawal.
JackoC wrote: Sat Jan 22, 2022 4:48 pm I do agree it would be nuts to pick an SWR either from a study or the result of a pre-retirement ABW calculation and never reconsider it during retirement regardless of realized returns or spending pattern changes. That seems obvious, but as has been said umpteen times on this thread and so many others, who really does that?
(Almost) nobody does it. But then how useful is a calculation that assumes people do? Is that calculation providing a clear picture of retirement?
JackoC wrote: Sat Jan 22, 2022 4:48 pm The key variable seems to me is how much money you have to have before voluntarily retiring, which is where I see no particular advantage to SWR derived from a (n IMO outdated) study vs. ABW using too high a return assumption. I do also agree the ABW method is more flexible pre-retirement to put in a return that's appropriately low. But getting everybody to agree what that is? :happy Also my general opinion is that the more black-boxy sophisticated the tool, the more likely people are to overlook basic flaw in the method as with FIREcalc.
All the more reason that black-boxy tools should force people to enter their assumptions and not hard code too much into the black box. People won't agree on what's a high, medium and low expected return to use. But at least they'll know what they are assuming.
JackoC wrote: Sat Jan 22, 2022 4:48 pm I bet there's some tool that lets you simulate with all your own inputs, but I'm not sure it's better than just assuming no real after tax return so initial SWR low 3's from early 60's or so.
I find it hard to interpret SWR results meaningfully because of the fixed withdrawal assumption. A 5% failure rate is too conservative. Is 10% appropriate? Is 20%? It's hard to understand what these numbers actually mean for a retirement. Far more useful I think to actually model the strategy being pursued and show the range of outcomes. The information I need is the probability distribution of retirement spending by age. This is easily produced through simulations of any strategy, whether ABW or fixed. Why settle for hard to interpret numbers when you can simply look at the relevant distribution? Why make the problem harder by answering a different hypothetical and having people guess from that?

The low tech way that makes sense to me is entering a low expected return in ABW. A fuller probabilistic picture is easily available by ABW simulations, so for me that's an easy choice.
Total Portfolio Allocation and Withdrawal (TPAW)
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

randomguy wrote: Sat Jan 22, 2022 8:13 pm
nigel_ht wrote: Sat Jan 22, 2022 8:07 pm
tibbitts wrote: Sat Jan 22, 2022 7:41 pm
nigel_ht wrote: Sat Jan 22, 2022 5:12 pm Basic SWR ignores SS.
Basic SWR does, but so does super-ultra-advanced SWR. It also ignores other pensions, employer supplements for health insurance, and countless other factors that might increase or decrease the need for withdrawals from a portfolio. SS just doesn't belong is a SWR discussion.
If you are doing apples to apples comparisons with VPW and other methods you need to consider it if they do or it’s an uneven playing field.

When you factor all the assumptions/requirements in then the difference in annual budget isn’t quite as high as some folks believe.

Plus if you use tools like cFIREsim switching withdrawal methods is a menu selection. Considering SWR with the same SS as something else becomes simpler.

So why wouldn’t “Advanced SWR” consider bridge to SS or SPIAs or pensions?
VPW doesn't change with SS. The VPW folks just like to include SS to make the volatility seem less...
Ah gotcha.

But there is a disinclination to promise any minimal withdrawal amount and a required flexibility test.

I wonder how many would meet that requirement without either SS/pension or so much money that any withdrawal method is relatively unimportant unless you really do want to spend everything down.
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

vineviz wrote: Sat Jan 22, 2022 8:17 pm
Ben Mathew wrote: Sat Jan 22, 2022 1:26 am Both investors have almost the same horizons (1 week apart) and almost the same and portfolios (1 week's withdrawal apart), but the withdrawals are wildly different ($80K vs $60K). It's because SWR is anchoring on starting portfolio value and not adjusting.
SWR was never a methodology. It is simply a guideline and a model. Like any model, it will break if you push the assumptions hard enough.

So I would say that YOUR version of SWR is "anchoring on starting portfolio value and not adjusting", but that need not be universally true.

For one thing, many practitioners will use the average portfolio value over some period of time to avoid precisely the conundrum you illustrated of having the initial withdrawal amount be tied to the portfolio balance on a particular day. It's common to use the average value over the course of a year, for example.

A second difficulty with your example is that your two investors are using the same withdrawal rate despite having different expected returns. The long-run expected return of stocks is not the same before a 40+% crash as it is after that crash, and the initial withdrawal rate should take that into account.

SWR is a simple heuristic device that allows us to readily make sense of a reality that is infinitely complicated. All models have flaws, but many of those models are useful BECAUSE of those flaws instead of despite them.
Good models are simple. But not too simple.

SWR is too simple.

The SWR model has caused people to arrive at faulty conclusions, one of which is that an upward sloping glide path in retirement reduces risk. This result is based on SWR studies and it's driven by SWR's fixed withdrawal assumption. If we assume variable withdrawals, then the fixed asset allocation is optimal (as derived in Merton's 1969 paper). An upward sloping glidepath increases risk in late retirement. You can't see this in an SWR model. It's too simple. You need a variable spending model to see this. So this is a case where the overly simplistic modeling in SWR is tripping people up and leading to bad decision making.
Total Portfolio Allocation and Withdrawal (TPAW)
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

Ben Mathew wrote: Sat Jan 22, 2022 8:19 pm
I find it hard to interpret SWR results meaningfully because of the fixed withdrawal assumption. A 5% failure rate is too conservative. Is 10% appropriate? Is 20%? It's hard to understand what these numbers actually mean for a retirement. Far more useful I think to actually model the strategy being pursued and show the range of outcomes.
SWR does just that…shows the range of outcomes based on duration and withdrawal rate.

20% failure means that the withdrawal rate ceiling selected failed in 20% of historical cases…you should probably pick something lower unless you are in poor health…
The information I need is the probability distribution of retirement spending by age. This is easily produced through simulations of any strategy, whether ABW or fixed. Why settle for hard to interpret numbers when you can simply look at the relevant distribution? Why make the problem harder by answering a different hypothetical and having people guess from that?
There are actual metrics regarding spending by age.

Here is the Kitces article about that.

https://www.kitces.com/blog/age-banding ... -category/
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Traditional SWR - flawed at the core

Post by vineviz »

Ben Mathew wrote: Sat Jan 22, 2022 8:41 pm The SWR model has caused people to arrive at faulty conclusions, one of which is that an upward sloping glide path in retirement reduces risk. This result is based on SWR studies and it's driven by SWR's fixed withdrawal assumption.
The rising equity glide path advocacy is based on a naive use of the historical record. It had nothing to do with SWR: any withdrawal method would have done better with a rising equity glide path because of the particular sequence of returns from the mid-1960s to the mid-1990s.
Ben Mathew wrote: Sat Jan 22, 2022 8:41 pm If we assume variable withdrawals, then the fixed asset allocation is optimal (as derived in Merton's 1969 paper). An upward sloping glidepath increases risk in late retirement. You can't see this in an SWR model. It's too simple. You need a variable spending model to see this. So this is a case where the overly simplistic modeling in SWR is tripping people up and leading to bad decision making.
Again, not the case. Fixed asset allocation is optimal with constant real withdrawals as well.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

nigel_ht wrote: Sat Jan 22, 2022 8:47 pm
Ben Mathew wrote: Sat Jan 22, 2022 8:19 pm
I find it hard to interpret SWR results meaningfully because of the fixed withdrawal assumption. A 5% failure rate is too conservative. Is 10% appropriate? Is 20%? It's hard to understand what these numbers actually mean for a retirement. Far more useful I think to actually model the strategy being pursued and show the range of outcomes.
SWR does just that…shows the range of outcomes based on duration and withdrawal rate.

20% failure means that the withdrawal rate ceiling selected failed in 20% of historical cases…you should probably pick something lower unless you are in poor health…
But the outcomes are not success and failure of the portfolio. If you adjust spending (like almost everyone is saying they will), then the outcome is more or less spending in retirement. It would be something like at

- Age 73, income will be $40,000 +- $3, 000 with 95% probability
- Age 74, income will be $40,000 +- $4, 000 with 95% probability
...

etc. for all retirement ages.

Plus the distribution for legacy if desired.

That's what's shown by the ABW simulator upthread.

A graph or table like this can be produced for any strategy, including a fixed strategy.

This spending distribution provides a realistic picture of retirement outcomes. The probability of success and failure is a red herring. Because there is no success and failure in retirement. Only more and less spending (including legacy if that matters).
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Traditional SWR - flawed at the core

Post by vineviz »

Ben Mathew wrote: Sat Jan 22, 2022 9:02 pm
This spending distribution provides a realistic picture of retirement outcomes.
It provides a picture of ONE ASPECT of retirement outcomes.

ABW shows you what income looks like if you hold the terminal value fixed. A SWR model shows you what the terminal value looks like if you hold the income fixed. It's just two different views of the same object.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
tibbitts
Posts: 23716
Joined: Tue Feb 27, 2007 5:50 pm

Re: Traditional SWR - flawed at the core

Post by tibbitts »

nigel_ht wrote: Sat Jan 22, 2022 8:07 pm So why wouldn’t “Advanced SWR” consider bridge to SS or SPIAs or pensions?
SWR has to do with calculating/estimating how much you can withdrawal from a portfolio for a specified length of time, based on historical return data. It has nothing to do with alternative sources of income, or how much you owe for taxes, or how much you need to maintain your lifestyle, or how much you need to barely survive, or anything else. Mixing in these other factors makes this more into a discussion of "safe spending rate."
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

vineviz wrote: Sat Jan 22, 2022 8:58 pm
Ben Mathew wrote: Sat Jan 22, 2022 8:41 pm The SWR model has caused people to arrive at faulty conclusions, one of which is that an upward sloping glide path in retirement reduces risk. This result is based on SWR studies and it's driven by SWR's fixed withdrawal assumption.
The rising equity glide path advocacy is based on a naive use of the historical record. It had nothing to do with SWR: any withdrawal method would have done better with a rising equity glide path because of the particular sequence of returns from the mid-1960s to the mid-1990s.
Ben Mathew wrote: Sat Jan 22, 2022 8:41 pm If we assume variable withdrawals, then the fixed asset allocation is optimal (as derived in Merton's 1969 paper). An upward sloping glidepath increases risk in late retirement. You can't see this in an SWR model. It's too simple. You need a variable spending model to see this. So this is a case where the overly simplistic modeling in SWR is tripping people up and leading to bad decision making.
Again, not the case. Fixed asset allocation is optimal with constant real withdrawals as well.
Not true. This is mathematical, not an artifact of historical data.

Risk is minimized when it's spread evenly across time: Instead of taking a big bet in one year and a small bet in another year, you take medium sized bets both years.

In variable withdrawals, the portfolio risk is being spread across all ages. If the portfolio does 10% better than expected, all remaining ages get 10% more than expected. If the portfolio does 10% worse than expected, all ages get 10% less. So if you hold the same AA on the entire portfolio every year, consumption in any given future year is equally sensitive to stocks. So the stock risk being applied to each age is constant.

By contrast, with fixed withdrawals, all the risk is being applied to the the last age currently funded. Spending in the near term does not change. The entire gain/loss of the portfolio is concentrated on the years at the end--not distributed across all ages as in variable withdrawals. So a fixed allocation would not spread stock risk evenly. At the start of retirement, when the portfolio is large, the gain/loss on the last funded age is large. As the portfolio dwindles over the course of retirement, the gain/loss gets smaller. To keep risk constant on the last years, one would need to hold a constant (PV adjusted) dollar stock allocation. Not a constant percentage stock allocation. A constant dollar stock allocation is of course a rising equity glide path.
Total Portfolio Allocation and Withdrawal (TPAW)
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

Ben Mathew wrote: Sat Jan 22, 2022 8:41 pm
Good models are simple. But not too simple.

SWR is too simple.
Repeated assertions don’t make something true…
The SWR model has caused people to arrive at faulty conclusions, one of which is that an upward sloping glide path in retirement reduces risk. This result is based on SWR studies and it's driven by SWR's fixed withdrawal assumption.

If we assume variable withdrawals, then the fixed asset allocation is optimal (as derived in Merton's 1969 paper). An upward sloping glidepath increases risk in late retirement. You can't see this in an SWR model. It's too simple. You need a variable spending model to see this. So this is a case where the overly simplistic modeling in SWR is tripping people up and leading to bad decision making.
Many variable spending models don’t assume/support a minimum required spending level while proponents assume that SWR requires maximum spending (ie you must take out the computed amount every year).

If you make that assumption then the implication is that the expected expenses match that of the SWR withdrawal amount. Dropping below that means you aren’t meeting expenses. This is a failure case.

At this point the withdrawals stop being variable…or at least they can only vary upwards and not down…and you can’t vary upwards very much in the worst case scenario.

So the upward glide path analysis isn’t incorrect because they make that same assumption: that the SWR amount is the spending floor. For a longer retirement higher equity results in a higher chance of success because your portfolio amount is marginal.

If you make this same assumption for VPW or ABW…that there is a spending floor…then the majority of differences disappear.

When the floor = SWR then there is no room for variability.

If the floor is above SWR then expenses won’t be met in the historical worst cases. All methods fail.

If the floor is below SWR then you can use SWR, VPW, ABW equally successfully using either a fixed AA or rising glide path…whatever floats your boat.
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

vineviz wrote: Sat Jan 22, 2022 9:11 pm
Ben Mathew wrote: Sat Jan 22, 2022 9:02 pm
This spending distribution provides a realistic picture of retirement outcomes.
It provides a picture of ONE ASPECT of retirement outcomes.

ABW shows you what income looks like if you hold the terminal value fixed. A SWR model shows you what the terminal value looks like if you hold the income fixed. It's just two different views of the same object.
Whatever anyone plans to do in retirement, I suggest that they actually model it and see the probability distribution of the outcomes. It's not hard to do. Don't rely on answers to hypothetical questions that aren't giving sufficient or relevant information.

If someone plans to keep their spending fixed, model that. If someone actually plans to vary their spending, model that. If they plan to do a bit of both, model that. If they plan to take double in leap years, model that. Whatever they plan to do, model it and see what retirement will look like. And don't just look at the probability of success and failure--however success and failure is defined--because that hides the magnitudes of the success and failure. Instead, see the range of possible outcomes for retirement spending distribution at each age (and legacy if that matters). Everything else is half the story--sometimes relevant, sometimes irrelevant, and sometimes misleading.
Total Portfolio Allocation and Withdrawal (TPAW)
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

randomguy wrote: Sat Jan 22, 2022 3:44 pm
nigel_ht wrote: Sat Jan 22, 2022 2:45 pm
randomguy wrote: Sat Jan 22, 2022 1:11 pm
Marseille07 wrote: Sat Jan 22, 2022 12:32 pm
nigel_ht wrote: Sat Jan 22, 2022 12:30 pm Not sure how many times that needs to be repeated in a forum where folks regularly post they are spending at 2% of initial portfolio rate and still want more safety…
Imo Ben's example is contrived because the granularity of SWR study is per year. IOW it doesn't account for a scenario Ben describes, where Investor 3 retires with 2M and a week later their portfolio is 1.5M.

in order to compare apples to apples, Investor 4 needs to retire a year later.
The time thing is a distraction. The SWR for 25 year periods just isn't that different from 30. It isn't remotely enough to account for the difference.
The difference is 5 years of spending at $80K…$400,000.

It’s a huge difference.

For a 25 year period at $80K a year you need $2M.

ONE year makes a difference in outcome because it’s a double bonus…one less year of spending and one more year of saving and growth.
It is an information problem. After the markets drop 50% (either in 1 week or 5 years), your odds of failure for the remaining period are not 5%.
Yep they are…because we aren’t rolling dice (aka Monte Carlo) but looking at historical results.

1929 was followed by 1932. That’s why it’s one of the two worst case scenarios. The events are historically linked and occur together 100% of the time.

If you use some sort of Monte Carlo as the basis of computing SWR you’d end up with a different number.
Even if you believe that things are totally random in the market (they aren't) imagine that to have portfolio failure you need to roll double 6's on a 2 dice . what are the odds? 3%. Now imagine you roll a 6 (i.e. the market dropped 40%). Your partner then joins you and you will share the next dice roll and then they get another roll. What are the odds of failure? Your odds are 1/6 (you have already lost the first dice roll). Your partner though is 1/36.
Your partner already rolled a 6 unless they weren’t in the market so the odds are the same.

If you can generate $1.5M in a week you don’t need to worry about SWR…
-The SWR for 30 years is 4% (95% success). The SWR for 25 years is 4.3%. I consider that a minimal difference. The question is why our guy can only take out 60k instead of 80k. Saving they can take out 65k explains a tiny portion of the difference.
Given that 5 years expenses is $400K then sometime over the course of 5 years the market has dropped $100K for the first investor.

Investor 1 hasn’t experienced the worse case scenario of 1929. It might be 1966 but it’s hard to tell with the information we have.

That means that the 1929 and 1966 outcomes are still as likely for the second investor as the first…so 4.3% with a potential 5% failure rate.
- Yep and that is why your are in trouble. This isn't exactly right but it is close enough to demonstrate the point. There are 100 chances when you retire. 5 of them fail. 75 times the market is up in year 1, 25% the market is down. If the market is up you never fail. If your market is down after year 1, what are your failure chances over the next 29 years? You have gone from 5% (5/100) to 25% (5/25) once you learn that year 1 is a down year.

-Nope. By the rules of the game to go broke if I need to roll one more 6. My partner needs to roll 2 6's to go broke. One is clearly much more likely than the other. If my partner took out 80k instead of 60k, they would have been playing from the start. When they chose to take out 60k they are starting a new game.
No, the sequence of rolls remains the same. It’s not a new game but your partner joining an existing one already started.

The sequence is x, x, x, x, 6 where x is any number not 6.

Investor 2 joins the game at this point for the next 25 rolls.

The next roll will either be a 6 or not. The odds of failure are the same for both.

That’s the way the game works as there is only one stock market.

You can argue that the game resets when two sixes are rolled because valuations get reset to something lower. That doesn’t apply to 1966 though.
Again this is all grossly simplified to make it clearer. You can't get too crazy with historical data. We just don't have enough of it. Just because if the market was up in year 1 and never failed might not transition well in the future.
Yes. And a 80% drop may be less survivable than the 1929 sequence because of inflation or some other factor.

“Worse then historical” is fairly imprecise…but it’s what we have. You can use mathematical models but how often do they actually validate them against real world sequences?
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

Ben Mathew wrote: Sat Jan 22, 2022 9:02 pm
nigel_ht wrote: Sat Jan 22, 2022 8:47 pm
Ben Mathew wrote: Sat Jan 22, 2022 8:19 pm
I find it hard to interpret SWR results meaningfully because of the fixed withdrawal assumption. A 5% failure rate is too conservative. Is 10% appropriate? Is 20%? It's hard to understand what these numbers actually mean for a retirement. Far more useful I think to actually model the strategy being pursued and show the range of outcomes.
SWR does just that…shows the range of outcomes based on duration and withdrawal rate.

20% failure means that the withdrawal rate ceiling selected failed in 20% of historical cases…you should probably pick something lower unless you are in poor health…
But the outcomes are not success and failure of the portfolio. If you adjust spending (like almost everyone is saying they will), then the outcome is more or less spending in retirement. It would be something like at

- Age 73, income will be $40,000 +- $3, 000 with 95% probability
- Age 74, income will be $40,000 +- $4, 000 with 95% probability
...

etc. for all retirement ages.
There is a point where spending can’t be reduced without significant consequences.

Again if this is equal to the SWR amount then the variable methods have no variability.

If it is above the SWR amount all methods fail.
Plus the distribution for legacy if desired.

That's what's shown by the ABW simulator upthread.

A graph or table like this can be produced for any strategy, including a fixed strategy.

This spending distribution provides a realistic picture of retirement outcomes. The probability of success and failure is a red herring. Because there is no success and failure in retirement. Only more and less spending (including legacy if that matters).
Well that’s clearly wrong.

If there is no success or failure in retirement then any withdrawal percentage is 100% successful as zero spending is still “success”.

There is always a floor. If there is no floor then you don’t need a withdrawal strategy at all.

Just do whatever you want because your needs will always be met.

Also, the probability that I will correctly estimate my age 75 spending to +/- $4K with 95% confidence is very slim.

I doubt you can do much better.
User avatar
Topic Author
9-5 Suited
Posts: 1307
Joined: Thu Jun 23, 2016 12:14 pm

Re: Traditional SWR - flawed at the core

Post by 9-5 Suited »

BogleFan510 wrote: Sat Jan 22, 2022 5:27 pm This post represents the danger of only reading a headline 'e.g. 4% is a "safe rate"' instead of actually taking the time to read and understand the paper that led to the conclusions. The paper is and was never meant to dictate any retirement withdraw rate absolutes.

To call it 'flawed at the core' as a headline is a clickbait style I find offputting and not helpful to investors planning for retirement. Explaining sequence of returns risk and the assumptions regarding retirement duration, as a concept, are much more helpful to people than scare tactic headlines.
I didn’t give more than a few seconds thought to the title, and if I had it to do over would have chosen something not only less clickbaity but also something that better represented the issue. I have read the study and hours of discussion and subsequent writings on the study, but funny enough the original post need not reference the trinity study at all. It was intended only to be a critique of any practical withdrawal method that uses the starting portfolio value as the anchor for all subsequent withdrawal decisions and 4% was only used as an example but any rate could have been chosen. Other posters subsequently introduced the trinity study specifically.

Many people do offhandedly imply that this is a real withdrawal strategy, while many others claim it isn’t and wasn’t ever intended to be that. Fine enough on the second point, which I hope is actually true.

And since this got to Page 3 from the last time I checked it, I guess click bait works ;)
JackoC
Posts: 4714
Joined: Sun Aug 12, 2018 11:14 am

Re: Traditional SWR - flawed at the core

Post by JackoC »

Ben Mathew wrote: Sat Jan 22, 2022 8:19 pm
JackoC wrote: Sat Jan 22, 2022 4:48 pm
Ben Mathew wrote: Sat Jan 22, 2022 11:42 am But even as a tool for evaluating when to retire, the SWR methodology (though not the 4% estimate) is pretty bad. The 5% failure rate leads to very conservative withdrawals, and would make people work far longer than they need to.

[...]
JackoC wrote: Sat Jan 22, 2022 10:03 am [...] using the distribution of past returns as if it's 'the' stationary distribution of returns irrespective of starting valuations. That's a clearly optimistic assumption IMO at historically high starting asset valuations like now.
Fully agree.
I'm glad you agree with last point but it seems perhaps at odds with the earlier statement that a 5% failure rate would lead to excessively conservative withdrawal rate.
1. 5% failure rate would lead to excessively conservative withdrawal rates if applied to a forward looking return distribution based on current valuations. ... This just happens to be a lucky period where SWR produces a reasonable estimate because its inherent conservativeness is being canceled by an optimistic return distribution.
JackoC wrote: Sat Jan 22, 2022 4:48 pm I do agree it would be nuts to pick an SWR either from a study or the result of a pre-retirement ABW calculation and never reconsider it during retirement regardless of realized returns or spending pattern changes. That seems obvious, but as has been said umpteen times on this thread and so many others, who really does that?
2. (Almost) nobody does it. But then how useful is a calculation that assumes people do? Is that calculation providing a clear picture of retirement?
JackoC wrote: Sat Jan 22, 2022 4:48 pm The key variable seems to me is how much money you have to have before voluntarily retiring, which is where I see no particular advantage to SWR derived from a (n IMO outdated) study vs. ABW using too high a return assumption.
3. All the more reason that black-boxy tools should force people to enter their assumptions and not hard code too much into the black box. People won't agree on what's a high, medium and low expected return to use. But at least they'll know what they are assuming.
JackoC wrote: Sat Jan 22, 2022 4:48 pm I bet there's some tool that lets you simulate with all your own inputs, but I'm not sure it's better than just assuming no real after tax return so initial SWR low 3's from early 60's or so.
4. I find it hard to interpret SWR results meaningfully because of the fixed withdrawal assumption. A 5% failure rate is too conservative. Is 10% appropriate? Is 20%? It's hard to understand what these numbers actually mean for a retirement. Far more useful I think to actually model the strategy being pursued and show the range of outcomes. The information I need is the probability distribution of retirement spending by age. This is easily produced through simulations of any strategy, whether ABW or fixed. Why settle for hard to interpret numbers when you can simply look at the relevant distribution? Why make the problem harder by answering a different hypothetical and having people guess from that?

The low tech way that makes sense to me is entering a low expected return in ABW. A fuller probabilistic picture is easily available by ABW simulations, so for me that's an easy choice.
1. My point was just that saying a target of 5% failure is too conservative is an opinion, distinguishing that from saying expected return is now significantly lower than past average historical realized which is a fact (we might politely call opinion on a polite forum but is really pretty obvious). However if one shares our joint opinion that 5% failure is too low a target and recognizes the fact that past return distribution is centered too high for % failure numbers it generates to be accurate, then yeah there's some fortunate offset in those two things.

2. Well again referring to point 3, prior to retirement the target is just an amount of money and the tactical way you'd manage it after retirement doesn't yet enter it. But I mainly grant this point, especially since practically 'SWR' is usually based on outdated studies.

3. If you use a black box tool, it should be like that I agree. I'm not sure everybody benefits from using black box tools at all though.

4. Here the assumption though is that, having rejected the past distribution of returns as being able to generate meaningful 'probability of failure', we could construct a distribution that would be able to. I'm not sure. I'd have reasonable confidence in one metric, the mean. I would make the default assumption that expected variance is the long term past variance but with less confidence. I'd have less confidence still in specifying skew and fat tailed-ness and so forth. I've looked at FIREcalc results with the simple adjustment of a dummy 'expense' number of 2-3% to reflect the lower mean of the forward looking distribution and I don't think that's *as* worthless as the raw results, but I wouldn't make my plan based on the details of any simulation. I've just made (what I believe are) quite conservative simplistic assumptions. But looking at simulations with an attempt at correct forward looking inputs is OK if others find value in it.
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Traditional SWR - flawed at the core

Post by randomguy »

nigel_ht wrote: Sat Jan 22, 2022 10:33 pm
No, the sequence of rolls remains the same. It’s not a new game but your partner joining an existing one already started.

The sequence is x, x, x, x, 6 where x is any number not 6.

Investor 2 joins the game at this point for the next 25 rolls.

The next roll will either be a 6 or not. The odds of failure are the same for both.

That’s the way the game works as there is only one stock market.

You can argue that the game resets when two sixes are rolled because valuations get reset to something lower. That doesn’t apply to 1966 though.
We should play this game for a $10/ game for 1 million games. I will let you be investor 1 since the odds are the same:) The odds of the next roll being a 6 is the same but the effect of rolling a 6 is different. If the next dice roll is a a 6, investor 1 loses (2 6s in a row). Investor 2 on the other hand gets another roll since they have only rolled 1 6.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Traditional SWR - flawed at the core

Post by Marseille07 »

9-5 Suited wrote: Sat Jan 22, 2022 10:49 pm Many people do offhandedly imply that this is a real withdrawal strategy, while many others claim it isn’t and wasn’t ever intended to be that. Fine enough on the second point, which I hope is actually true.
It can be a fine real withdrawal strategy, so long as you understand what it means.

In fact ERN did a follow-up study on this where the success criteria was more strict than simply having $0+ in your account after 30&60 years. Using something like that is probably more reasonable.
User avatar
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: Traditional SWR - flawed at the core

Post by Ben Mathew »

JackoC wrote: Sun Jan 23, 2022 11:38 am prior to retirement the target is just an amount of money and the tactical way you'd manage it after retirement doesn't yet enter it.
The target needs to be converted to retirement spending in order to know if it's sufficient, and that requires some assumptions about the retirement strategy. How much does $1 million at age 65 mean for retirement spending if markets go as well as expected? What will spending be if markets do better or worse than expected? Is there enough of a cushion? The answers to that will be different for different strategies. It's useful to think about it before retirement to inform decisions like how much to save and when to retire.
JackoC wrote: Sun Jan 23, 2022 11:38 am 4. Here the assumption though is that, having rejected the past distribution of returns as being able to generate meaningful 'probability of failure', we could construct a distribution that would be able to. I'm not sure. I'd have reasonable confidence in one metric, the mean. I would make the default assumption that expected variance is the long term past variance but with less confidence. I'd have less confidence still in specifying skew and fat tailed-ness and so forth. I've looked at FIREcalc results with the simple adjustment of a dummy 'expense' number of 2-3% to reflect the lower mean of the forward looking distribution and I don't think that's *as* worthless as the raw results, but I wouldn't make my plan based on the details of any simulation. I've just made (what I believe are) quite conservative simplistic assumptions. But looking at simulations with an attempt at correct forward looking inputs is OK if others find value in it.
The simulators in ABW and TPAW simply lets the user to shift the historical log return distribution so that the mean return matches the user's expected return. This keeps the variance and skews and fat tails and all other aspects of the historical distribution. That seemed "good enough" to me.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
vanbogle59
Posts: 1314
Joined: Wed Mar 10, 2021 7:30 pm

Re: Traditional SWR - flawed at the core

Post by vanbogle59 »

9-5 Suited wrote: Sat Jan 22, 2022 10:49 pm the original post ...was intended only to be a critique of any practical withdrawal method that uses the starting portfolio value as the anchor for all subsequent withdrawal decisions
9-5 Suited wrote: Fri Jan 21, 2022 9:47 am There are two well-known behavioral errors...
...how do you rationalize this when variable withdrawal strategies exist (VPW, ABW, fixed %) that would eliminate this obviously contradictory example above?
Do you think the R in ABW is immune to behavioral error?
"The ABW calculation requires the user to input the expected rate of return of the portfolio."
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Traditional SWR - flawed at the core

Post by vineviz »

vanbogle59 wrote: Sun Jan 23, 2022 12:58 pm Do you think the R in ABW is immune to behavioral error?
"The ABW calculation requires the user to input the expected rate of return of the portfolio."
To be fair, the SWR estimation requires an assumed rate of real portfolio growth as well. It's often not explicit but it's always there.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Traditional SWR - flawed at the core

Post by willthrill81 »

vineviz wrote: Sun Jan 23, 2022 1:01 pm
vanbogle59 wrote: Sun Jan 23, 2022 12:58 pm Do you think the R in ABW is immune to behavioral error?
"The ABW calculation requires the user to input the expected rate of return of the portfolio."
To be fair, the SWR estimation requires an assumed rate of real portfolio growth as well. It's often not explicit but it's always there.
Indeed. It's literally impossible to determine mathematically how much can be withdrawn without any estimate of forward returns. Even 0% real is a forward return estimate.
The Sensible Steward
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

randomguy wrote: Sun Jan 23, 2022 12:10 pm
nigel_ht wrote: Sat Jan 22, 2022 10:33 pm
No, the sequence of rolls remains the same. It’s not a new game but your partner joining an existing one already started.

The sequence is x, x, x, x, 6 where x is any number not 6.

Investor 2 joins the game at this point for the next 25 rolls.

The next roll will either be a 6 or not. The odds of failure are the same for both.

That’s the way the game works as there is only one stock market.

You can argue that the game resets when two sixes are rolled because valuations get reset to something lower. That doesn’t apply to 1966 though.
We should play this game for a $10/ game for 1 million games. I will let you be investor 1 since the odds are the same:) The odds of the next roll being a 6 is the same but the effect of rolling a 6 is different. If the next dice roll is a a 6, investor 1 loses (2 6s in a row). Investor 2 on the other hand gets another roll since they have only rolled 1 6.
If you are modeling the market as a game then anytime two 6’s show up we both lose money.

It’s not independent…
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Traditional SWR - flawed at the core

Post by Marseille07 »

willthrill81 wrote: Sun Jan 23, 2022 1:04 pm
vineviz wrote: Sun Jan 23, 2022 1:01 pm
vanbogle59 wrote: Sun Jan 23, 2022 12:58 pm Do you think the R in ABW is immune to behavioral error?
"The ABW calculation requires the user to input the expected rate of return of the portfolio."
To be fair, the SWR estimation requires an assumed rate of real portfolio growth as well. It's often not explicit but it's always there.
Indeed. It's literally impossible to determine mathematically how much can be withdrawn without any estimate of forward returns. Even 0% real is a forward return estimate.
Sure, but isn't the OP talking about ABW & SWR as withdrawal methodologies?

Estimating SWR is a different ballgame than choosing a cell on the SWR AA & WR table and start using it. The latter does not require any estimation, as the table is crafted using historical data.
nigel_ht
Posts: 4742
Joined: Tue Jan 01, 2019 9:14 am

Re: Traditional SWR - flawed at the core

Post by nigel_ht »

vineviz wrote: Sun Jan 23, 2022 1:01 pm
vanbogle59 wrote: Sun Jan 23, 2022 12:58 pm Do you think the R in ABW is immune to behavioral error?
"The ABW calculation requires the user to input the expected rate of return of the portfolio."
To be fair, the SWR estimation requires an assumed rate of real portfolio growth as well. It's often not explicit but it's always there.
Well yes, but it’s a historically crappy one…
Post Reply