Variable Percentage Withdrawal (VPW)

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sat Aug 03, 2013 3:58 pm

longinvest wrote:
siamond wrote:Well, here you are, you gave me the answer. Year after year, just add the inflation-equivalent to the 'heirs money", and any return beyond that should be added/removed from the capital that is used for the VPW formula. The 'heirs money' will be protected, and any extra return (or possible loss) will be used for my own benefit...

If I was to adopt such a strategy, I would be careful not to invest the heirs money into the markets. I would keep this money safe in CDs or Money Market Funds (MMF).

I would invest the rest and apply VPW on it. If, any year, interests don't cover inflation on the heirs money, I would take off the necessary amount from the annual VPW payment and add it into CDs/MMF.

The reason I wouldn't invest the heirs money is that, otherwise, I would be back into implementing a constant-dollar-withdrawal-like strategy (with all its sequence of return problems) for trying to keep the heirs money up to a constant-dollar balance.

Alternately, if you don't mind your heirs waiting, just put this money into TIPS (but not into TIPS funds). This way, your heirs will assuredly get their money if they wait until maturity. No risk.

On a personal note, I am a believer in giving the money while I am alive. As a primitive risk management strategy, I try not to create situations where I am more valuable to others dead than alive. :oops:


I would tend to agree on the personal note, but maybe not 100% of it (personal view). Anyhoo, this only re-inforces the point of having a portion of (inflation-adjusted) capital which is 'reasonably' protected. Whether you give after 10 years, 30 years or after your death, same thing. Also please remember that I was asking for such 'protected capital' concept, not only for heirs, but also to protect scenarios past life expectancy (which MAY change quite a lot 20 years from now, who knows), and maybe a safety cushion for some form of LTC needs too. Without such a way of protecting a slice of one's portfolio, I fear that VPW is very unpractical.

Now (and I realize it depends on personal goals), I certainly didn't mean to leave the 'protected capital' on a low-return/low-risk investment. This would seem quite a waste and a disservice to the heirs, notably if you enjoy a long retirement (or time period before you give most of it). It should stay on the regular AA, like the rest, and then you adjust stocks/bonds portfolio as you see fit. And when you decide to give, you reverse-balance on the whole thing.

So here is what I was thinking. Basic idea is to start from a portfolio of say $1M, and have a fixed amount you want to protect, say $400k. Then let the $400k evolve with the inflation (hence stay protected), and use anything beyond it with your nifty VPW algorithm. Anything beyond it means:
1. the way the $1M minus $400k part of the portfolio evolves based on market returns and the annual withdrawals
2. the returns of the $400k beyond the inflation (in most years, this is a positive return, but it can turn into a negative return)

Since your algorithm neatly adjusts itself to whatever current value of the portfolio to consider, it can be applied to the sum of #1 and #2. One could also factor in occasional lump sum cash inflows, by the way, which is really cool.

I actually did a model of such 'Protected VPW' idea this morning, this is real easy (after a good night of sleep!), and it seems to work fine. The hypersensitivity of the overall model to the market gyrations continues to make me wary, but that's another discussion... Give it a try! :wink:

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sat Aug 03, 2013 4:43 pm

siamond wrote: Also please remember that I was asking for such 'protected capital' concept, not only for heirs, but also to protect scenarios past life expectancy (which MAY change quite a lot 20 years from now, who knows), and maybe a safety cushion for some form of LTC needs too. Without such a way of protecting a slice of one's portfolio, I fear that VPW is very unpractical.


Of course VPW is very unpractical for what you are trying to use it for!.....Take your money you want to 'Protect' and put it in Tips and a separate account. Then run VPW with the remaining part of your portfolio.

What I fear, is that the Original Poster will incorporate ideas into his excellent tool to render it Unpractical for what it was designed. I believe the OP stated this tool was designed with the intent to try to Use up All Capital at the 'End of Plan'. Leaving a Pile at the End negates his Original Design.

umfundi
Posts: 3361
Joined: Tue Jun 07, 2011 5:26 pm

Re: Variable Percentage Withdrawal

Post by umfundi » Sat Aug 03, 2013 5:14 pm

Yes,

The intent is to have an excellent wine cellar, and to have but one bottle left at the end. :happy

Keith
Déjà Vu is not a prediction

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sat Aug 03, 2013 6:22 pm

Cut-Throat wrote:
siamond wrote:Of course VPW is very unpractical for what you are trying to use it for!.....Take your money you want to 'Protect' and put it in Tips and a separate account. Then run VPW with the remaining part of your portfolio.

What I fear, is that the Original Poster will incorporate ideas into his excellent tool to render it Unpractical for what it was designed. I believe the OP stated this tool was designed with the intent to try to Use up All Capital at the 'End of Plan'. Leaving a Pile at the End negates his Original Design.


Did you even read my post? I doubt it... I wanted to AVOID putting a specialized chunk of portfolio in TIPS with low returns... Choosing our AA and choosing your SWR method seem rather orthogonal choices to me. Both should apply to your entire portfolio.

And just zeroing the value to protect would of course restore the original model.

I am not saying this is the way to go. I am just exploring ideas to make an extension of the concept more applicable to real life. At least for my own personal goals...

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sat Aug 03, 2013 6:53 pm

siamond wrote:Did you even read my post? I doubt it... I wanted to AVOID putting a solid chunk of portfolio in TIPS with low returns... Choosing our AA and choosing your SWR method seem rather orthogonal choices to me. Both should apply to your entire portfolio.

And just zeroing the value to protect would of course restore the original model.

I am not saying this is the way to go. I am just exploring ideas to make an extension of the concept more applicable to real life. At least for my own personal goals...


Yes, I read your Post. Did you Read the OPs intent when designing this tool? He did not design it for YOUR personal goals, but designed it for someone that wished to use up all of their Capital at "end of Plan'. The more requirements that you throw at any planning tool, the smaller the Withdrawal rate will be. His initial design was to maximize the withdrawal rate early in retirement, without putting the 'end of Plan' in Jeopardy.

As we used to say in the Software Business "There is No Right Way to do the Wrong Thing'.

I suggest you look at FireCalc, which has a portfolio Floor option already.

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sat Aug 03, 2013 7:56 pm

Which is exactly why I spoke of exploring an "extension of the concept".

If you're not interested by going one inch beyond the original intent, be it. Sounds that we're talking past each other anyway. Hopefully, other folks will be a bit more curious to such avenues.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 04, 2013 11:39 am

Hi, I've uploaded a new version with Prof. Robert Shiller's data (it goes back to 1871).

One thing I am unsure about is my calculation for Bonds total returns. Here's what I did:
  • I assumed that RLONG(N) was the interest rate on a newly issued 10-year bond in year N.
  • An investment in a $100 bond, at the start of year N will generate a RLONG(N) interest payment at the start of year N+1.
  • The difficulty is to value this bond at the start of year N+1. To approximate it, I use the present value of a bond with coupon payment of RLONG(N), future value of $100 in 9 years, with a yield to maturity of RLONG(N+1). This is like assuming that the yield curve is flat (a 9-years bond has the same interest rate as a 10-years bond).
  • The total return, for year N, is RLONG(N)+Value-100

Please look at it and tell me if it makes sense.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 5:53 am

I've added an optional spending cap to further smooth out consumption.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 7:40 am

In his spreadsheet, Adamodar does a calculation similar to mine to transform interest rates into bond returns with one difference; he computes the present value of the bond (at year N+1) using a 10-years period, instead of a 9-years period.

Using Adamodar's calculation would increase the gains when interest rates go down, and decrease the gains (or increase the losses) when interest rates go up.

When I compare Simba's inflation numbers with Shiller's, they seem pretty similar (Simba in blue, Shiller in red):
Image

When I compare Simba's stock numbers with Shiller's, they seem pretty similar (Simba in blue, Shiller in red):
Image

But, when I compare Simba's bond numners with Shiller's, Shiller's numbers have much more volatility (Simba in blue, Shiller in red):
Image

Using Adamodar's calculation, I get a wider dispersion of Shiller's returns, as expected:
Image

This leads me to lose confidence in approximated bond returns based on interest rates. Please do not assume that the approximated bond returns are accurate, based on Shiller's data.

In the spreadsheet, should I add a warning, or should I limit using Shiller's data to US stocks (no bonds data)? What do you think?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 9:05 am

Mixing a portfolio of 50% 10-year Bonds, 50% 1-year notes brings Shiller bond numbers (10-year government bonds) much closer to Simbas (TBM). THis is probably because it reduces the duration closer to TBM's 5.5.

Here's the visual comparison (Simba blue, Shiller red):
Image

I'll use that in the spreadsheet.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 9:40 am

I've just uploaded a new version that uses estimated Intermediate-Bond returns (similar to TBM) based on Shiller 1-year and 10-year interest rates. This gives a much smoother ride. I guess that the Bogleheads recommendation to use TBM or an Intermediate-Bond fund has some pretty solid ground behind it.

I don't know if there's much interest for all this as nobody is commenting.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Mon Aug 05, 2013 10:33 am

longinvest wrote:I've just uploaded a new version that uses estimated Intermediate-Bond returns (similar to TBM) based on Shiller 1-year and 10-year interest rates. This gives a much smoother ride. I guess that the Bogleheads recommendation to use TBM or an Intermediate-Bond fund has some pretty solid ground behind it.

I don't know if there's much interest for all this as nobody is commenting.


I think a few folks have commented that I respect. It is the best Withdrawal Calculator that I have seen thus far.

User avatar
boknows
Posts: 122
Joined: Mon Jan 28, 2013 3:11 pm
Location: Fairfax, VA
Contact:

Re: Variable Percentage Withdrawal

Post by boknows » Mon Aug 05, 2013 10:42 am

longinvest wrote:I've just uploaded a new version that uses estimated Intermediate-Bond returns (similar to TBM) based on Shiller 1-year and 10-year interest rates. This gives a much smoother ride. I guess that the Bogleheads recommendation to use TBM or an Intermediate-Bond fund has some pretty solid ground behind it.

I don't know if there's much interest for all this as nobody is commenting.


I've been interested in this, as I've been developing a calculator for online use that uses historical data (sort of like firecalc) and have been curious to add different models of withdrawal methods into it. I appreciate all the hard work. I'm now digging through a bunch of the Simba backtesting threads tr3ying to find all the smaller asset classes data back to 1927. Fun times.
35 - Married - Aiming for FI/ER in early 40s.

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Mon Aug 05, 2013 11:49 am

Many thanks for your hard work, longinvest.

Your latest Excel sheet is indeed very enlightening. I had a perception that there was a skew towards steep increases at the end of the time period (notably when using a more aggressive portfolio), and this was really bothering me, but this was actually a consequence of the relatively short time period being looked at (having this Simba worksheet be extended before 1972 will be *so* useful!). When using the Shiller data, there are indeed many more starting dates not having such a skew. And where the spending cap does prove useful indeed (in addition of just being common sense) as peaks occur much earlier.

As to my issue with the lack of a modicum of capital being protected (for those of us who have such a goal), it is very easily implemented by just subtracting the inflation-adjusted equivalent of such low cap to the value of the portfolio being subject to the VPW algorithm. No more complicated than that, and I played around, this works fine (actually it protects more than I perceived, because it cumulates with your other protection mechanism, the 20% rule). Of course, this reduces the withdrawals to start with, but that's fair. So the issue is entirely and easily solvable, albeit only making sense for a relatively low cap.

Now, as most variable methods, withdrawals do vary quite a lot... And the spending cap only mildly helps in this respect. I understand this is the intrinsic nature of the approach, but I don't know if this makes it practical enough for many of us. And I still am quite wary of betting on life expectancy not changing dramatically in the coming 30 to 50 years. Still... nice work. Very interesting.

PS. in the VPW calculation sheet, you may want to make this 20% cap an assumption changeable in a single cell. Same thing for the age '100'.

User avatar
Cb
Posts: 273
Joined: Wed Feb 21, 2007 12:39 am

Re: Variable Percentage Withdrawal

Post by Cb » Mon Aug 05, 2013 2:04 pm

longinvest wrote:I don't know if there's much interest for all this as nobody is commenting.


I'm keeping an eye on this thread. I welcome your work on variable withdrawals. I've previously modeled variable withdrawals using FIREcalc, simulating various degrees of % of remaining portfolio WD schemes by pumping up FIREcalc's Annual Portfolio Expense Ratio input field. I'd input a "normal" WD of say, 2.0%, then add another 2% to my Portfolio's ER to reflect the % of remaining portfolio portion.

Unfortunately, the current owner of the E-R board & FIREcalc is non-responsive to questions regarding FIREcalc, and all development and data updates seem to have halted a few years back.

So I'm looking fwd to what you wind with,

Cb 8-)

User avatar
boknows
Posts: 122
Joined: Mon Jan 28, 2013 3:11 pm
Location: Fairfax, VA
Contact:

Re: Variable Percentage Withdrawal

Post by boknows » Mon Aug 05, 2013 2:16 pm

Cb wrote:
longinvest wrote:I don't know if there's much interest for all this as nobody is commenting.


I'm keeping an eye on this thread. I welcome your work on variable withdrawals. I've previously modeled variable withdrawals using FIREcalc, simulating various degrees of % of remaining portfolio WD schemes by pumping up FIREcalc's Annual Portfolio Expense Ratio input field. I'd input a "normal" WD of say, 2.0%, then add another 2% to my Portfolio's ER to reflect the % of remaining portfolio portion.

Unfortunately, the current owner of the E-R board & FIREcalc is non-responsive to questions regarding FIREcalc, and all development and data updates seem to have halted a few years back.

So I'm looking fwd to what you wind with,

Cb 8-)


I don't want to hijack the thread, because I already hinted at it before, but I'm actively developing a Retirement Calculator that is very much like FireCalc. I'm actively soliciting ideas for feature improvements, and already have a number of things that FireCalc does not. Head over to [link removed by admin LadyGeek] and register in the forums. I have the mods approval to start my own thread in the BH forums about my project... I'm just really trying to get the smaller asset class data into my database before I post that (I know how Bogleheads love their asset class choices).

[Update: 06-Aug-13: The Bogleheads Advisory Board has since rescinded this approval. --admin LadyGeek]
35 - Married - Aiming for FI/ER in early 40s.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 3:19 pm

siamond, thank for the comments.

siamond wrote:PS. in the VPW calculation sheet, you may want to make this 20% cap an assumption changeable in a single cell. Same thing for the age '100'.


OK, I've done that in the newest version.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 3:26 pm

boknows wrote:I'm now digging through a bunch of the Simba backtesting threads tr3ying to find all the smaller asset classes data back to 1927. Fun times.


Thanks, boknows. I'm keeping an eye, too, on Simba's Backtesting. It would be less work for me.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Mon Aug 05, 2013 3:35 pm

Cb wrote:I'm keeping an eye on this thread. I welcome your work on variable withdrawals. I've previously modeled variable withdrawals using FIREcalc, simulating various degrees of % of remaining portfolio WD schemes by pumping up FIREcalc's Annual Portfolio Expense Ratio input field. I'd input a "normal" WD of say, 2.0%, then add another 2% to my Portfolio's ER to reflect the % of remaining portfolio portion.

Unfortunately, the current owner of the E-R board & FIREcalc is non-responsive to questions regarding FIREcalc, and all development and data updates seem to have halted a few years back.

So I'm looking fwd to what you wind with,


Cb, thanks. My main objective is the development/improvement/discussion of a Variable Percentage Withdrawal strategy. The back testing spreadsheet is by no way meant to replace Firecalc and similar tools. It is way too primitive. But, until these tools add VPW (hopefully, one day) as an option, I need a way to back test it.

Already, spreadsheet has allowed me to see the actual behavior of VPW on historical data. This is a great eye opener. When I thought about the VPW formula, I had no idea how it would actually behave in real life.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Mon Aug 05, 2013 4:13 pm

longinvest wrote: The back testing spreadsheet is by no way meant to replace Firecalc and similar tools. It is way too primitive. But, until these tools add VPW (hopefully, one day) as an option, I need a way to back test it.
Already, spreadsheet has allowed me to see the actual behavior of VPW on historical data. This is a great eye opener. When I thought about the VPW formula, I had no idea how it would actually behave in real life.


Nor Should VPW try to Replace FireCalc. VPW is a Flexible Withdrawal Tool, which has been sorely needed. Most every other tool was always a Fixed Withdrawal under Historical Worst Case Conditions (Which is not how we actually Spend Money).

I have been convinced for some time that the "Take 4% Inflation Adjusted Amount from your Portfolio" was largely a Fantasy. I was a proponent of "Variable Percent of Portfolio Balance " for some time, but have never had a tool to provide a reasonable formula and Backtest against Historical Data.

This Tool Fits the Bill !

umfundi
Posts: 3361
Joined: Tue Jun 07, 2011 5:26 pm

Re: Variable Percentage Withdrawal

Post by umfundi » Mon Aug 05, 2013 4:34 pm

Cut-Throat wrote:
longinvest wrote: The back testing spreadsheet is by no way meant to replace Firecalc and similar tools. It is way too primitive. But, until these tools add VPW (hopefully, one day) as an option, I need a way to back test it.
Already, spreadsheet has allowed me to see the actual behavior of VPW on historical data. This is a great eye opener. When I thought about the VPW formula, I had no idea how it would actually behave in real life.


Nor Should VPW try to Replace FireCalc. VPW is a Flexible Withdrawal Tool, which has been sorely needed. Most every other tool was always a Fixed Withdrawal under Historical Worst Case Conditions (Which is not how we actually Spend Money).

I have been convinced for some time that the "Take 4% Inflation Adjusted Amount from your Portfolio" was largely a Fantasy. I was a proponent of "Variable Percent of Portfolio Balance " for some time, but have never had a tool to provide a reasonable formula and Backtest against Historical Data.

This Tool Fits the Bill !

I am going to echo Cut-Throat. VPW is simply a different, and complementary, way to look at the problem. Fess up:

1. You don't know how long you will live.
2. You don't know what future market returns will be, not on average, never mind year by year.

Unfortunately, these are two critical factors in making a precise plan.

I think this VPW tool does a good job of optimizing a probable scenario. It is not insurance for the worst possible case, as many other withdrawal strategies try to be.

Keith
Déjà Vu is not a prediction

stevelb
Posts: 12
Joined: Sat Aug 11, 2007 11:21 am

Re: Variable Percentage Withdrawal

Post by stevelb » Mon Aug 05, 2013 4:55 pm

Thank you for this excellent tool and your continued work and modifications. Very much appreciated.

User avatar
LadyGeek
Site Admin
Posts: 39221
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Variable Percentage Withdrawal

Post by LadyGeek » Sun Aug 18, 2013 12:21 pm

Yes, it's appreciated. This spreadsheet is now in the wiki: Savings/Withdrawals Over Time

The spreadsheet itself isn't displayed because the wiki can only do that for Google Drive files. If you move to Google Drive, let me know.

A few comments:

- The charts' X-axis scaling can be improved by setting the chart type to scatter (vs. line). You'll get some extension on the low and high side, but I think it's better than what's currently shown.

I used this technique in this post of JonoJono1: Improve, Revise, Update, Debug Simba's Backtester. It might also fix the Excel vs. LibreOffice compatibility so you can have one file for both. You'll need to save the file as .xls from within LibreOffice, then open MS Excel and see how it looks.

- Instructions
Consider adding a line (after row 10) that the user should then compare the Variable Withdrawal Percentage to the other methods, which are available in the CDW..., CPW... , 1NW... tabs.

It's also important to note that the variable percentages are calculated from the 'VPW Calculation' tab and that the expected return and standard deviation can be changed.
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.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 12:51 pm

LadyGeek wrote:Yes, it's appreciated. This spreadsheet is now in the wiki: Savings/Withdrawals Over Time


Thank you, LadyGeek.


The spreadsheet itself isn't displayed because the wiki can only do that for Google Drive files. If you move to Google Drive, let me know.

A few comments:

- The charts' X-axis scaling can be improved by setting the chart type to scatter (vs. line). You'll get some extension on the low and high side, but I think it's better than what's currently shown.



OK. I'll try that.


I used this technique in this post of JonoJono1: Improve, Revise, Update, Debug Simba's Backtester. It might also fix the Excel vs. LibreOffice compatibility so you can have one file for both. You'll need to save the file as .xls from within LibreOffice, then open MS Excel and see how it looks.



I'll install LibreOffice and see how it goes.


- Instructions
Consider adding a line (after row 10) that the user should then compare the Variable Withdrawal Percentage to the other methods, which are available in the CDW..., CPW... , 1NW... tabs.



I'll do that.


It's also important to note that the variable percentages are calculated from the 'VPW Calculation' tab and that the expected return and standard deviation can be changed.


I'm planning some improvements to VPW. I'll possibly reorganize things and possibly rename a few parameters to eliminate ambiguities.

Thanks for the comments and for the addition into the Wiki.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 2:01 pm

I have received some private and public feedback on VPW. In particular, I'd like to thank Cut-Throat, boknows, and siamond.

After back-testing the behavior of VPW for various Asset Allocations and retirement lengths, mostly on Shiller data in the VPW spreadsheet, and also simulating random returns, my feeling is that:
  • VPW is pretty successful at achieving the goals of spending, in general, more than the other usual withdrawal methods (Constant-Dollar, Constant-Percentage, 1/N) in early retirement years.
  • VPW allows the retiree to spend more of his wealth than Constant-Percentage.
  • VPW doesn't fail like Constant-Dollar.
  • VPW is, in most cases, smoother than 1/N.

Unfortunately, VPW suggests highly varying payments, from year to year in inflation-adjusted dollars. Another way to say that is: VPW tracks the markets too closely.

In the back-testing tab of the spreadsheet, there is an option to cap on payments (in inflation-adjusted dollars). If one is lucky in choosing this cap, it leads to stable withdrawals for the whole retirement length, but also to a big pile of unspent money. If one is not so lucky (cap too high), payments will drop with the markets and will only be capped if the markets recover enough later in retirement, leading to an unnecessary late cap and unspent money.

I've thought about an improvement. I am seeking feedback before going ahead and implementing it into the spreadsheet. To avoid confusion, I'll call it VPW 2.0. 8-)

I got my inspiration, in part, from siamond's BoW method ( viewtopic.php?f=1&t=121138 ).

Here's the idea: The VPW table would be used as a guide to make sure withdrawals are not too high (adding safety against premature depletion) or not too low (adding protection against too much unspent wealth). Otherwise, the objective would be to keep payments relatively stable.

VPW 2.0 would work as follows. Using our typical 65-years old person, 60% bonds / 40% stocks $1,000,000 portfolio:
  • At age 65, VPW says 4.6%. Compute 90% of the proposed withdrawal: $1,000,000 X 4.6 % X 90% = $41,000. That's the initial payment.
  • At age 66, VPW says 4.7%. Compute 90% and 110% of the proposed withdrawal. Assuming the market dropped 5%, this gives us: ($1,000,000 - $41,000) - 5% = $911,050. So, $911,050 X 4.7% X 90% = $39,000, and $911,050 X 4.7% X 110% = $47,000. This gives us the payment boundaries: $39,000 < Payment < $47,000. If last year's payment is within the computed bounds, keep it as is (in inflation-adjusted dollars, of course). If it is below the lower bound, the new payment is the lower bound. If it is above the upper bound, the new payment is the upper bound. In this case, the payment remains $41,000 for our 66 years-old person.
  • At age 67, VPW says 4.8%. Assuming the market increased 10%, this gives us: ($911,050 - $41,000) + 10% = $957,055. So, $957,055 X 4.7% X 90% = $40,000, and $957,055 X 4.7% X 110% = $49,000. In this case, the payment remains $41,000 for our 67 years-old person.
  • And so on.

As you can see, I keep the VPW tables as is, but I add a tolerance interval around the calculated payment to reduce, as much as possible, year-to-year volatility in payments. Doing so, I do not sacrifice VPW's protections against premature depletion or under-spending.

What do you think of this idea?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 2:08 pm

longinvest wrote:I have received some private and public feedback on VPW. In particular, I'd like to thank Cut-Throat, boknows, and siamond.

After back-testing the behavior of VPW for various Asset Allocations and retirement lengths, mostly on Shiller data in the VPW spreadsheet, and also simulating random returns, my feeling is that:
  • VPW is pretty successful at achieving the goals of spending, in general, more than the other usual withdrawal methods (Constant-Dollar, Constant-Percentage, 1/N) in early retirement years.
  • VPW allows the retiree to spend more of his wealth than Constant-Percentage.
  • VPW doesn't fail like Constant-Dollar.
  • VPW is, in most cases, smoother than 1/N.

Unfortunately, VPW suggests highly varying payments, from year to year in inflation-adjusted dollars. Another way to say that is: VPW tracks the markets too closely.

In the back-testing tab of the spreadsheet, there is an option to cap on payments (in inflation-adjusted dollars). If one is lucky in choosing this cap, it leads to stable withdrawals for the whole retirement length, but also to a big pile of unspent money. If one is not so lucky (cap too high), payments will drop with the markets and will only be capped if the markets recover enough later in retirement, leading to an unnecessary late cap and unspent money.

I've thought about an improvement. I am seeking feedback before going ahead and implementing it into the spreadsheet. To avoid confusion, I'll call it VPW 2.0. 8-)

I got my inspiration, in part, from siamond's BoW method ( viewtopic.php?f=1&t=121138 ).

Here's the idea: The VPW table would be used as a guide to make sure withdrawals are not too high (adding safety against premature depletion) or not too low (adding protection against too much unspent wealth). Otherwise, the objective would be to keep payments relatively stable.

VPW 2.0 would work as follows. Using our typical 65-years old person, 60% bonds / 40% stocks $1,000,000 portfolio:
  • At age 65, VPW says 4.6%. Compute 90% of the proposed withdrawal: $1,000,000 X 4.6 % X 90% = $41,000. That's the initial payment.
  • At age 66, VPW says 4.7%. Compute 90% and 110% of the proposed withdrawal. Assuming the market dropped 5%, this gives us: ($1,000,000 - $41,000) - 5% = $911,050. So, $911,050 X 4.7% X 90% = $39,000, and $911,050 X 4.7% X 110% = $47,000. This gives us the payment boundaries: $39,000 < Payment < $47,000. If last year's payment is within the computed bounds, keep it as is (in inflation-adjusted dollars, of course). If it is below the lower bound, the new payment is the lower bound. If it is above the upper bound, the new payment is the upper bound. In this case, the payment remains $41,000 for our 66 years-old person.
  • At age 67, VPW says 4.8%. Assuming the market increased 10%, this gives us: ($911,050 - $41,000) + 10% = $957,055. So, $957,055 X 4.7% X 90% = $40,000, and $957,055 X 4.7% X 110% = $49,000. In this case, the payment remains $41,000 for our 67 years-old person.
  • And so on.

As you can see, I keep the VPW tables as is, but I add a tolerance interval around the calculated payment to reduce, as much as possible, year-to-year volatility in payments. Doing so, I do not sacrifice VPW's protections against premature depletion or under-spending.

What do you think of this idea?


It sounds fine, but I would like to back-test it, to see what it would actually do.

I have seen other withdrawal methods called Floor and ceiling which reminds me of this.

Can you make it an option, in your current spreadsheet, so that we can turn it off and on?

EDIT : I've changed my mind and I don't care for this..Explanation in Below Post.
Last edited by Cut-Throat on Sun Aug 18, 2013 3:00 pm, edited 1 time in total.

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 2:19 pm

longinvest wrote:As you can see, I keep the VPW tables as is, but I add a tolerance interval around the calculated payment to reduce, as much as possible, year-to-year volatility in payments. Doing so, I do not sacrifice VPW's protections against premature depletion or under-spending.

What do you think of this idea?


Actually, in my own big ugly Excel sheet (not the derivative of yours, my own stuff), I implemented VPW plus some safety measures... Hence something similar and yet different from your proposal... I did define 90%/100% guardrails (ala Guyton-Klinger), but I accepted the VPW math if inside such guardrails (and went to the lower/upper guardrail otherwise). While you're suggesting more stability in the spend. Now I see the analogy you've drawn from BoW. Interesting idea. Well, we should give it a try. Something I've learned by now is that the best intent is often defeated by unexpected dynamics of the market returns...

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 2:27 pm

While you are thinking hard... :wink:

I have two lingering issues with VPW:

1. even with some form of guardrails, it often goes crazy high after a while, notably at the end of the retirement time. Or sometimes earlier. This could be viewed as ok depending on your goals (e.g. donate extra money when possible), but this does make me cringe a tad. I added an absolute (constant dollars) upper cap in my Excel sheet, but this isn't quite consistent with the general mindset of the model.
=> Not sure we should do something about this point, all things considered.

2. it continues to bug me that the whole thing is making such a big bet on life expectancy. If the point is to spend a portion of your portfolio in a finite amount of time, and you have another portion of your portfolio (+ SS/Pension) for safety, then fine, but otherwise, this is just not reasonable. And the 20% cap at the end is really not such a good safety and... well... a bit of a hack. :wink:
=> Maybe try to do something with re-adjusting the math every year based on life expectancy models (like the one SS uses)?

I am just sharing thoughts here, not pushing for a specific change.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 2:43 pm

siamond wrote:While you are thinking hard... :wink:

I have two lingering issues with VPW:

1. even with some form of guardrails, it often goes crazy high after a while, notably at the end of the retirement time. Or sometimes earlier. This could be viewed as ok depending on your goals (e.g. donate extra money when possible), but this does make me cringe a tad. I added an absolute (constant dollars) upper cap in my Excel sheet, but this isn't quite consistent with the general mindset of the model.
=> Not sure we should do something about this point, all things considered.


The problem remains that future returns are uncertain. If the markets go up quickly, at 90 years old, VPW will produce big payments. There just isn't time left to spread the spending. Spending earlier (before we know the markets go up) might be unwise.

(On a side note: A SPIA has the same feature. If you buy at 85 years old, you'll get a very high payment, regardless of interest rates).

2. it continues to bug me that the whole thing is making such a big bet on life expectancy. If the point is to spend a portion of your portfolio in a finite amount of time, and you have another portion of your portfolio (+ SS/Pension) for safety, then fine, but otherwise, this is just not reasonable.


I don't expect VPW to replace the safety of an insurance product pooling life expectancies. If you don't have a pension or enough SS, maybe you need to buy a SPIA (inflation-adjusted or not: there are schools of thought on it).

And the 20% cap at the end is really not such a good safety and... well... a bit of a hack. :wink:
=> Maybe try to do something with re-adjusting the math every year based on life expectancy models (like the one SS uses)?

I am just sharing thoughts here, not pushing for a specific change.


Maybe you're expecting to live forever; there are science-fiction novels about that, and maybe science is getting there. I don't know.

The solution would be similar to what I have already suggested: for the part you don't want to deplete, you use Constant-Percentage Withdrawal (CPW). Now, you could try to improve it with the same idea as VPW 2.0. Maybe a CPW 2.0 with a payment tolerance interval?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
LadyGeek
Site Admin
Posts: 39221
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Variable Percentage Withdrawal

Post by LadyGeek » Sun Aug 18, 2013 2:44 pm

Consider showing the remaining balance in the 'Variable-Percentage Withdrawal' tab. IMO, you've left that exercise to the reader, but showing the numbers in one spot will bring the point home. It will also show (close to) 0.00 at the end to confirm the analysis. The portfolio size is entered in 'VPW Backtest', which makes the UI a bit disconnected.

- In 'Variable-Percentage Withdrawal', move all of the data back into a single column from Age 40 to 99. Then, add the balance remaining ('VPW Calculation' Balance * portfolio size) in a new column.
- Consider moving the portfolio size into 'Variable-Percentage Withdrawal' and link to the cell from 'VPW Backtest'

An additional benefit of one column is the ease of pasting the data elsewhere (multiple columns are more work and prone to error).

================
In 'VPW Backtest' there are 4 conditional formatting rules which change the color of various cells (white on black). Are they used for debugging, or is there some other purpose? An explanation should be provided.

Minor nit: 'Historical Returns 2' the term "long bonds" refers to Canadian bonds which have a maturity of 11-30 years. For the US crowd, I suggest changing "long" to "10-Year" bonds to avoid confusion.
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.

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 2:59 pm

Actually I've changed my mind on your 'Floor and Ceiling' approach. It destroys the simplicity of your VPW tool.
And Any tool that is too complex to explain will not enjoy longevity.

VPW currently gives a Constant or rising increase of Percentage of Portfolio Balance as you age. Easy to Understand Why and How.

Somehow, I'd like to see the starting percentage being smaller in times of Higher Stock Valuations and Lower interest rates.
that is what I am currently doing by changing 'Expected returns'

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 3:08 pm

*** POST UPDATED AS I HAD ORIGINALLY QUOTED IMPROPERLY PARAMETERIZED CHARTS ***

Well, I was intrigued and I couldn't resist... It just took a couple of changes to implement VPW2.0 in my own Excel sheet.

Overall, it seems to be a pretty good improvement. I quickly tried quite a few years between 1917 and 1980, and I couldn't find any undesirable side-effects. I had to switch to 20% guardrails to make it meaningful though (VPW varies a lot!). I did look at the impact on the portfolio value trajectory, and this is relatively minimal, as one might expect.

Here are two examples, 1978 and the dreaded 1965. I use my own AA (quite diversified, and rather aggressive towards stocks) with a mix of Shiller (1915 to 1971) and Simba (1972+) returns in my own Excel sheet, so you are not going to find the same exact results, but I think it shows well what happens.

Image

Image

Give it a try with your own Excel sheet, this does seem worth it. And it's trivial to make it optional.
Last edited by siamond on Sun Aug 18, 2013 7:00 pm, edited 2 times in total.

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 3:18 pm

longinvest wrote:Maybe you're expecting to live forever; there are science-fiction novels about that, and maybe science is getting there. I don't know.

The solution would be similar to what I have already suggested: for the part you don't want to deplete, you use Constant-Percentage Withdrawal (CPW). Now, you could try to improve it with the same idea as VPW 2.0. Maybe a CPW 2.0 with a payment tolerance interval?


This is the science thing which does bug me a bit. This is exactly the point, we don't know! Things may change a LOT 20 years from now. Cancer might be just a bad memory, etc. I agree that a form of CPW (or BOW) could partly address the issue, but the point is nevertheless quite intrisic to VPW (I think?). Hence my suggestion to try to play with a dynamic adjustment of life expectancy. But maybe this breaks the PMT idea, I didn't dive in the math behind this function enough to appreciate if that is the case, or not.

Anyhoo, as I said, just a thought... :wink:

PS. and yes, the spirit of your new proposal is indeed broader than VPW. The Guyton-Klinger model might benefit from it... and yes, the CPW model too. Interesting.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 3:34 pm

Cut-Throat wrote:Actually I've changed my mind on your 'Floor and Ceiling' approach. It destroys the simplicity of your VPW tool.
And Any tool that is too complex to explain will not enjoy longevity.

VPW currently gives a Constant or rising increase of Percentage of Portfolio Balance as you age. Easy to Understand Why and How.

Somehow, I'd like to see the starting percentage being smaller in times of Higher Stock Valuations and Lower interest rates.
that is what I am currently doing by changing 'Expected returns'


I could make that simpler: each year, compute 2 percentages: low and high.

The withdrawal rule: You can withdraw as much as you like within the low/high boundaries.

So, instead of:
65 4.6%
66 4.7%
68 4.8%

VPW 2.0 would say:
65 3.8% to 4.6%
66 3.8% to 4.7%
67 3.9% to 4.8%

Back-testing would start in the middle of the first interval and try to keep payments stable.

What do you think?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 4:04 pm

longinvest wrote:
Cut-Throat wrote:Actually I've changed my mind on your 'Floor and Ceiling' approach. It destroys the simplicity of your VPW tool.
And Any tool that is too complex to explain will not enjoy longevity.

VPW currently gives a Constant or rising increase of Percentage of Portfolio Balance as you age. Easy to Understand Why and How.

Somehow, I'd like to see the starting percentage being smaller in times of Higher Stock Valuations and Lower interest rates.
that is what I am currently doing by changing 'Expected returns'


I could make that simpler: each year, compute 2 percentages: low and high.

The withdrawal rule: You can withdraw as much as you like within the low/high boundaries.

So, instead of:
65 4.6%
66 4.7%
68 4.8%

VPW 2.0 would say:
65 3.8% to 4.6%
66 3.8% to 4.7%
67 3.9% to 4.8%

Back-testing would start in the middle of the first interval and try to keep payments stable.

What do you think?


What I think, is don't ruin the good in search of the perfect. It's still more Complex than your original Idea. And I am a Big Proponent of the KISS Principle.
I very much like the idea that you came up with of a Single Percentage of Portfolio Balance that rises or remain constant based on Age. Easy to say and understand and it makes perfect Sense.

I am still in favor of some formula that would change the Starting Percentage Based on the Back Testing Years Valuations... Throughout Historical Data, we have all seen where Low Valuations have led to increased Withdrawal Rates and High Valuations have led to lower Withdrawal Rates.

Maybe someone else would chime in with idea of what to look at for Stock Valuations in Past History and Interest rate History for Bonds.

Until then, I am using VPW 1.0 and Manipulating Expected Return Percentages to give me pretty smooth data. My Example from 1937, Age 63....70% Bonds 30% Stocks .... .5% For Bonds 1.5% from Stocks ...Portfolio gives .8% real.
Even this Starting in 1937 with the $1 Million Portfolio starts out with a 3.2% at $32K, but by 1948 it drops to $22K. Doable, but I would not want to start out higher than 3.2% ..The portfolio is also pretty much gone at age 100 with about $70K left over. These are Real World Numbers. So the Expected Return was Probably less than 1%....So, it would make sense for VPW try to do this on a backtest.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 4:27 pm

Cut-Throat wrote:What I think, is don't ruin the good in search of the perfect. It's still more Complex than your original Idea. And I am a Big Proponent of the KISS Principle.
I very much like the idea that you came up with of a Single Percentage of Portfolio Balance that rises or remain constant based on Age. Easy to say and understand and it makes perfect Sense.


I guess you're right. Anyway, I tested the idea in my spreadsheet; it doesn't make a real difference unless you make the tolerance band wider. The problem is that, the wider the tolerance band, the less you end up spending in the earlier years (and more in the later). Not good.

In the same vein, removing the artificial 20% cap seems like a good idea. It doesn't accomplish much, anyway, and without it, the withdrawals are smoother in the later years. If course, at the end of the depletion period, you're left with $0, but this is like with 1/N, but in a smoother manner.

I am still in favor of some formula that would change the Starting Percentage Based on the Back Testing Years Valuations... Throughout Historical Data, we have all seen where Low Valuations have led to increased Withdrawal Rates and High Valuations have led to lower Withdrawal Rates.

Maybe someone else would chime in with idea of what to look at for Stock Valuations in Past History and Interest rate History for Bonds.

Until then, I am using VPW 1.0 and Manipulating Expected Return Percentages to give me pretty smooth data. My Example from 1937, Age 63....70% Bonds 30% Stocks .... .5% For Bonds 1.5% from Stocks ...Portfolio gives .8% real.
Even this Starting in 1937 with the $1 Million Portfolio starts out with a 3.2% at $32K, but by 1948 it drops to $22K. Doable, but I would not want to start out higher than 3.2% ..The portfolio is also pretty much gone at age 100 with about $70K left over. These are Real World Numbers.


How about getting rid of the Asset Allocation and simply adding a "growth slope" cell to be provided by the user?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 4:34 pm

longinvest wrote:How about getting rid of the Asset Allocation and simply adding a "growth slope" cell to be provided by the user?


Not sure what you mean by this? Did you mean 'Expected returns'?

I think Asset Allocation is a good idea, because some people have pensions and are more comfortable with a Higher Stock Allocation.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 4:36 pm

siamond wrote:Well, I was intrigued and I couldn't resist... It just took a couple of changes to implement VPW2.0 in my own Excel sheet.

Overall, it seems to be a pretty good improvement. I quickly tried quite a few years between 1917 and 1980, and I couldn't find any undesirable side-effects. I had to switch to 20% guardrails to make it meaningful though (VPW varies a lot!). I did look at the impact on the portfolio value trajectory, and this is relatively minimal, as one might expect.

Here are two examples, 1978 and the dreaded 1965. I use my own AA (quite diversified, and rather aggressive towards stocks) with a mix of Shiller (1915 to 1971) and Simba (1972+) returns in my own Excel sheet, so you are not going to find the same exact results, but I think it shows well what happens.

Image

Image

Give it a try with your own Excel sheet, this does seem worth it. And it's trivial to deactivate, just use 0% guardrails.


To me, the tolerance bands don't seem to make much of a difference. Of course, the minor variations disappear, but the major ones remain intact. So, as Cut-Throat says: why add the complexity?
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

Call_Me_Op
Posts: 6482
Joined: Mon Sep 07, 2009 2:57 pm
Location: Milky Way

Re: Variable Percentage Withdrawal

Post by Call_Me_Op » Sun Aug 18, 2013 4:39 pm

What happens if you are still alive after 30 years?
Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 4:41 pm

Call_Me_Op wrote:What happens if you are still alive after 30 years?


Read the Threads, this tool has nothing to do with 30 years. You set the Plan length.

User avatar
LadyGeek
Site Admin
Posts: 39221
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Variable Percentage Withdrawal

Post by LadyGeek » Sun Aug 18, 2013 4:44 pm

Keeping things simple is good, as you then wonder how to interpret the results. Complexity assumes the user understands the implications of changing the numbers.

Not a suggestion, but an observation: You can change the starting point of the historical distributions as another way to (somewhat) randomize the results - a random point on a random pattern sort of thing (what Simba's back-tester does).

===================================
Suggested documentation cleanup:

'VPW Backtest' , 'CDW Comparison', 'CPW Comparison', '1NW Comparison' all have a column title of "Constant Dollars" but appear to have different meaning in each worksheet. Is constant dollars the reference for the comparison? Can you clarify?

(Perhaps this is the same as the previous question): In the comparison worksheets, it's not at all obvious what those comparisons are to. Are they constant dollars? Can you add some chart plot titles to help with the description?

'VPW Backtest' "Returns" heading should say "Historical Returns", perhaps concatenate with the selected data source, e.g. "Shiller Historical Returns".
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.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 4:53 pm

Cut-Throat wrote:
longinvest wrote:How about getting rid of the Asset Allocation and simply adding a "growth slope" cell to be provided by the user?


Not sure what you mean by this? Did you mean 'Expected returns'?


Yes.

I think Asset Allocation is a good idea, because some people have pensions and are more comfortable with a Higher Stock Allocation.


It seems to me that you don't like a static "Expected Real Return" basic on Asset Allocation without taking valuation into consideration. Yet, I don't see how to compute such valuation and how to use it to tweak the VPW tables. After reading The Four Pillars of Investing by William Bernstein, lately, I am growing uncomfortable suggesting any such number.

The basic idea of the VPW table is to combine 1/N with a growth rate, then compute withdrawal percentages. So, if I don't know how to compute this growth rate, why not let the user set it?

[Edited to fix a typo.]
Last edited by longinvest on Sun Aug 18, 2013 5:03 pm, edited 1 time in total.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 4:54 pm

Call_Me_Op wrote:What happens if you are still alive after 30 years?


You use SS, pensions, and SPIAs for your spending floor. It's in the F.A.Q.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 4:57 pm

Call_Me_Op wrote:What happens if you are still alive after 30 years?


Forgot to say: VPW is age based. Your question question should have been:
What happens if you are still alive at 100 years old?

The "target age" is configurable. If you want to leave a legacy, then use Constant-Percentage Withdrawal, instead.

VPW is a tool, not a retirement plan.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal

Post by Cut-Throat » Sun Aug 18, 2013 6:00 pm

longinvest wrote:It seems to me that you don't like a static "Expected Real Return" basic on Asset Allocation without taking valuation into consideration. Yet, I don't see how to compute such valuation and how to use it to tweak the VPW tables. After reading The Four Pillars of Investing by William Bernstein, lately, I am growing uncomfortable suggesting any such number.

The basic idea of the VPW table is to combine 1/N with a growth rate, then compute withdrawal percentages. So, if I don't know how to compute this growth rate, why not let the user set it?


We'll have to think on this some more....The problem I see with letting the 'user' set the Growth rate, is that some users don't have a clue.

You had a default plugged in... Maybe it was just too high ? I don't know. I do know it changes based on Valuations. I'm just not sure how to do it. It does not have to be Exact, as VPW compensates by always taking the percent against the remaining portfolio balance. That is the strength of this tool.

A powerful tool, could do this based on past history and then change it when back-testing against a certain retirement year.
Not something we have to come up with today, but it bears consideration. Let's sleep on it.

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 6:57 pm

Silly me, I just realized that when I quickly generated a few charts, I forgot to deactivate the guardrails on my implementation of the regular VPW. So I was comparing a 'smooth guardrails' model to a 'step-by-step guardrails' model, which wasn't the intent. Sorry about that, my bad.

I fixed my post with more proper charts here: http://www.bogleheads.org/forum/viewtopic.php?f=10&t=120430&p=1780577#p1780577

Personally, I agree with the intent that *IF* you plan to use a good deal of VPW as a primary source of income/spend, then you might seek some form of year-to-year stability. You may notice on the new charts that it actually does take a few years (often 4) before a change occurs. Now the problem is this change can be fairly dramatic.

The perceived complexity doesn't seem significant to me. Frankly, guardrails are trivial to understand. While the core math of VPW is NOT trivial to understand, and would be hard to implement without a worksheet of sorts. And well, we live in a world of Excel sheets anyway.

Now those dramatic cliff-like changes... Hm... Not too sure about it on second thought. Actually, maybe my smoother implementation of guardrails was better... :wink:

longinvest wrote:To me, the tolerance bands don't seem to make much of a difference. Of course, the minor variations disappear, but the major ones remain intact. So, as Cut-Throat says: why add the complexity?

Oh... Just saw this post, I missed it when catching up with you guys... I guess we reached a similar conclusion. The big cliffs aren't very nice indeed.

PS. and yes, removing the 20% cap mechanism seems the right thing to do. This form of cap was just not helping...

User avatar
siamond
Posts: 3160
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal

Post by siamond » Sun Aug 18, 2013 7:25 pm

Cut-Throat wrote:Read the Threads, this tool has nothing to do with 30 years. You set the Plan length.


longinvest wrote:
Call_Me_Op wrote:What happens if you are still alive after 30 years?


Forgot to say: VPW is age based. Your question question should have been:
What happens if you are still alive at 100 years old?

The "target age" is configurable. If you want to leave a legacy, then use Constant-Percentage Withdrawal, instead.

VPW is a tool, not a retirement plan.


Actually... Maybe cut-throat's comment should be taken to heart. If this entire VPW concept was squarely described as a fixed-duration tool (i.e. how to spend a chunk of invested money in N years, irrespective of market gyrations), then we'd stop agonizing about "what if you're still alive", "what if a cure to cancer is found", and so on.

After all, there might be other reasons to select a given fixed duration, which have nothing to do with your age or your life expectancy.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 8:40 pm

I have just uploaded a new version of the VPW spreadsheet (see viewtopic.php?f=10&t=120430&p=1761580#p1761563 ). I've migrated it to LibreOffice and I now use .xls (Excel) as the native format.

In this version, I have made many interface improvements based on LadyGeek suggestions and tips. I also removed the mostly useless percentage and spending caps.

LadyGeek: Thanks a lot for your tips and suggestions.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
LadyGeek
Site Admin
Posts: 39221
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Variable Percentage Withdrawal

Post by LadyGeek » Sun Aug 18, 2013 9:40 pm

You're welcome. Getting the UI (User Interface) in a clear, understandable format is critical and is why I spend so much time on this aspect. All of your hard work (everyone here) goes for nothing if the user can't understand it. That being said...

'VPW Backtest' -
- Typo in G25, "Shiller HIstorical Returns" (should be lower case 'i')
- The historical return data should be formatted to 2 decimal places. Consider if this is necessary for your Historical Returns worksheets.
- The conditional formatting rules are still there. I'm not sure what they are for.
- The stock allocation section should be combined with the Asset Allocation in 'Variable-Percentage Withdrawal'. I see what you are doing, but someone is going to confuse the two. It's best to put everything in one spot so it's clear to see (US stocks / International stocks / bonds) together.

'CDW Comparison', 'CPW Comparison', '1NW Comparison' - In A1 of each worksheet, clarify that the comparison is to Variable Percentage Withdrawal. It's not obvious, I checked the chart data to be sure. For example: "Constant-Dollar Withdrawal (CDW) Comparison to Variable Percentage Withdrawal (VPW)"

Global suggestion: It's very easy for someone to accidentally overwrite a cell formula. Change the cells needing user input to bold blue. I use this formatting technique a lot.
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.

longinvest
Posts: 2395
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal

Post by longinvest » Sun Aug 18, 2013 11:40 pm

LadyGeek wrote:You're welcome. Getting the UI (User Interface) in a clear, understandable format is critical and is why I spend so much time on this aspect. All of your hard work (everyone here) goes for nothing if the user can't understand it. That being said...

'VPW Backtest' -
- Typo in G25, "Shiller HIstorical Returns" (should be lower case 'i')


OK.

- The historical return data should be formatted to 2 decimal places. Consider if this is necessary for your Historical Returns worksheets.


OK.

- The conditional formatting rules are still there. I'm not sure what they are for.


I used black conditional formatting because I don't know how to make the cells disappear. The idea is that:
  • Shiller's Historical Data doesn't have International returns. The US/International allocation is ignored when Shiller is selected. I use black formatting to show that this data is ignored.
  • The retirement start year has no meaning, when Random data is selected. I use black formatting to show that the retirement year is ignored.
  • There are a few cells that I change to red when the user inputs invalid values. Example: a Stock allocation higher than 100% or lower than 0%.

Do I really have to document this into the spreadsheet?

- The stock allocation section should be combined with the Asset Allocation in 'Variable-Percentage Withdrawal'. I see what you are doing, but someone is going to confuse the two. It's best to put everything in one spot so it's clear to see (US stocks / International stocks / bonds) together.


This is tricky. The VPW table does not take into account any difference in the stock allocation between US and International, so it is not needed there. Also, Shiller's data does not include international stocks. Even random returns do no take them into account.

I included the US/International split on the back-test sheet because Simba's data contains the historical returns of International stocks (VGTSX).

It doesn't seem appropriate to put this stock split on the VPW Table sheet. I made an exception for the portfolio size, because you asked for it and it didn't matter much. I should probably put it back on the back-testing sheet.

I think that it is better to keep all back-testing-specific inputs out of the VPW Table sheet. The VPW Table sheet should remain as simple as possible. It only needs 2 pieces of data: Stock percentage in Asset Allocation, and Portfolio Depletion Age.

'CDW Comparison', 'CPW Comparison', '1NW Comparison' - In A1 of each worksheet, clarify that the comparison is to Variable Percentage Withdrawal. It's not obvious, I checked the chart data to be sure. For example: "Constant-Dollar Withdrawal (CDW) Comparison to Variable Percentage Withdrawal (VPW)"


OK.

Global suggestion: It's very easy for someone to accidentally overwrite a cell formula. Change the cells needing user input to bold blue. I use this formatting technique a lot.


There's no danger of unwanted modifications; all cells are locked against modification, except the ones meant to be modified by the user. The modifiable cells have a green heading. I thought it became obvious once you start playing with it. It seems not.

I can try adding a special border to modifiable cells, just to make it even more obvious. Anyway, if you can't modify a cell, you can't overwrite its formula.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

Post Reply