Variable Percentage Withdrawal (VPW)

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
Cut-Throat
Posts: 2011
Joined: Sun Oct 17, 2010 9:46 am

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Wed Dec 25, 2013 10:49 am

longinvest wrote:I use a "scatter plot" (thanks to LadyGeek for the trick) which automatically adjusts the width of the graph. So, it doesn't make a difference. :beer

(It's 150 years, actually, just enough to fit in all of Shiller's historical data, when I want to do other investigations.)


I have no talent at all when it comes to Excel Spreadsheets, but I was in Software development for 30 years. From a 'design' point of view, if you display
your withdrawal amounts, and balance amounts in detail (Which is very nice BTW), the graphs on the same page should match the length of withdrawal period.
IOW - Everything on the page is tracking the same Time Period.

If it cannot be done within the confines of Excel, that is quite understandable.

Nice work you have done there all in all! :beer

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Wed Dec 25, 2013 11:21 am

Hi Cut-Throat,

Cut-Throat wrote:I have no talent at all when it comes to Excel Spreadsheets, but I was in Software development for 30 years. From a 'design' point of view, if you display
your withdrawal amounts, and balance amounts in detail (Which is very nice BTW), the graphs on the same page should match the length of withdrawal period.
IOW - Everything on the page is tracking the same Time Period.

If it cannot be done within the confines of Excel, that is quite understandable.


I think that I don't understand what you are suggesting. Currently, the graphs automatically adjust for the selected period.

For example, if you select a 35-years withdrawal period, you get:
Image

If, instead, you select a 45-years withdrawal period, you get:
Image

Isn't that what you are asking for?

Nice work you have done there all in all! :beer


Thanks!
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 (VPW)

Post by Cut-Throat » Wed Dec 25, 2013 11:33 am

Yes, yours looks perfect!

The problem again is in Excel. I remember having this same problem when you first did VPW.
I opened it with Open Office and everything looks great.

The Excel Version has Time periods of 160 years, so everything is scrunched to one end....
Not sure why.... Thanks!

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Wed Dec 25, 2013 2:54 pm

Hey longinvest... Merry Christmas! :sharebeer

So you ended up removing all the math about Bonds/Equities AA... No big deal, it's easy to experiment. And yes, it gets a bit trickier with the Int'l factor in the AA.
=> I would actually suggest to add REITs in the mix (for the Simba data set), since so many people use them even in very simple portfolios (cf. decorrelated returns).

It took me a little while to locate where you moved the all important start rate (in the Table tab). Frankly, I much preferred the way you had it set up in the N-1th version, with the 'recommended' rate displayed, and the user being allowed to override the cell above, which was used as the start rate for the actual VPW math.
=> I would suggest to restore this approach, so that all important parameters the end user will typically want to play with are in the primary tab.
=> (while leaving the 'long term growth trend' guesses where they are - including the ability to override - this is lower level logic).

I agree with cut-throat that showing the inflation on the primary tab (for the simulation table) would be a nice improvement to better understand the dynamics of the simulation. Just narrow a bit some columns (e.g. first two), and you'll easily create space for it.

Finally, the overall layout is somewhat confusing, as the data related to the single cycle (start year) is mixed up with the multi-year data. Maybe regroup the multi-year data on the top rows (aligned with the various input parameters), and regroup the year-in-focus data & charts in the rows below (aligned with the simulation table)? Or maybe do it the reverse way around, leftmost columns = input params + multi-years data; rightmost columns = year-in-focus data. Maybe remove the worst-drop chart to make space (it isn't very interesting, pretty much flat for all cycles including a really big crisis). There is probably more cosmetic stuff to do, but this seems the main thing imho.

===
Then the main missing thing for me is the ability to add a recurring income (e.g. SS/Pension) in the model. It's a full discussion in itself, but when I started to factor that in as simple portfolio addition every year (which works well with CVPW; it's -usually- a withdrawal reduction in truth, but same thing) in my own Excel sheet, this totally opened my eyes. Give it a try...

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Wed Dec 25, 2013 6:49 pm

siamond wrote:Hey longinvest... Merry Christmas! :sharebeer


Merry Christmas, siamond!

So you ended up removing all the math about Bonds/Equities AA... No big deal, it's easy to experiment. And yes, it gets a bit trickier with the Int'l factor in the AA.
=> I would actually suggest to add REITs in the mix (for the Simba data set), since so many people use them even in very simple portfolios (cf. decorrelated returns).


If people want to Slice and Dice, good for them. They can modify the spreadsheet. Personally, I have no intention to spend time maintaining this data in my spreadsheet. The Three Fund Portfolio is sufficient for 90% of people, in my humble opinion.

It took me a little while to locate where you moved the all important start rate (in the Table tab). Frankly, I much preferred the way you had it set up in the N-1th version, with the 'recommended' rate displayed, and the user being allowed to override the cell above, which was used as the start rate for the actual VPW math.
=> I would suggest to restore this approach, so that all important parameters the end user will typically want to play with are in the primary tab.
=> (while leaving the 'long term growth trend' guesses where they are - including the ability to override - this is lower level logic).


Why would a user of my spreadsheet want to fiddle with these parameters escapes me. Just tell me: What Start Percentage have you discovered that is better than the one provided by VPW?

Do you really expect a user that is unable to find the Table sheet to have the ability to come up with a reasonable explanation for fiddling with these numbers?

I think that the current design is OK. VPW's numbers are as good as they can get. They shouldn't be based on 10/20-year forward predictions; this is just not how VPW works. (e.g. Mr. Bogle suggests using yield to maturity for bonds and dividend yield and earning growth for stocks, in The Clash of the Cultures). Such rates are, unfortunately, not what VPW needs.

Mr. Bogle writes about speculative returns and long-tern returns. VPW needs something that goes way beyond the time-frame of Mr. Bogle's long-term returns (which are intended to be within one's lifetime). The parameters I've put, in there, are the best we can do, based on world history. If you've got better numbers to propose, please do here, in this forum, so that we can debate about them, instead of trying to make every VPW user "market-time" these values.

I agree with cut-throat that showing the inflation on the primary tab (for the simulation table) would be a nice improvement to better understand the dynamics of the simulation. Just narrow a bit some columns (e.g. first two), and you'll easily create space for it.


If I can't show the data graphically, I don't see the point. Users that want to really understand what's happening need to go and look into the Selection sheet that shows the detailed nominal returns and inflation. There's just no place for all this on the main sheet.

Finally, the overall layout is somewhat confusing, as the data related to the single cycle (start year) is mixed up with the multi-year data.
Maybe regroup the multi-year data on the top rows (aligned with the various input parameters), and regroup the year-in-focus data & charts in the rows below (aligned with the simulation table)? Or maybe do it the reverse way around, leftmost columns = input params + multi-years data; rightmost columns = year-in-focus data. Maybe remove the worst-drop chart to make space (it isn't very interesting, pretty much flat for all cycles including a really big crisis). There is probably more cosmetic stuff to do, but this seems the main thing imho.


I feel that you are criticizing just for the sake of it. :happy The titles are clear. I don't see how somebody can be confused. The reason I've put it that way was so that the user gets a hint that there are more data/graphs below. Perfect is the enemy of good enough. I'll settle for good enough as perfection is illusive. What will look perfect to siamond will look imperfect to somebody else.

===
Then the main missing thing for me is the ability to add a recurring income (e.g. SS/Pension) in the model. It's a full discussion in itself, but when I started to factor that in as simple portfolio addition every year (which works well with CVPW; it's -usually- a withdrawal reduction in truth, but same thing) in my own Excel sheet, this totally opened my eyes. Give it a try...


You like making people spend their wealth as late as possible, ideally past 85-years old, do you? I completely disagree with your suggested combined approach!

The right approach for SS/Pension is trivial. Assuming you'll get a $15,000 pension in 5 years; Take $15,000 off your portfolio to spend this year. Take another $60,000 dollars off your portfolio (for the next 4 years until the pension kicks in) and put it in CDs or in a short-term bond fund. Apply VPW to the rest of your portfolio. Actually, combining this with delayed Social Security is highly recommended if you are healthy and likely to live long. See the Delay Social Security to age 70 and Spend more money at 62 thread.


I don't see the point of making the spreadsheet any more complex. Simplicity is beautiful. I'd like to preserve it. VPW is just a tool, not a full retirement package.

Cheers. :sharebeer
Last edited by longinvest on Wed Dec 25, 2013 8:56 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

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Wed Dec 25, 2013 8:02 pm

longinvest wrote:I don't see the point of making the spreadsheet any more complex. Simplicity is beautiful. I'd like to preserve it. VPW is just a tool, not a full retirement package.


+1

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Wed Dec 25, 2013 8:08 pm

siamond wrote:[...] Maybe remove the worst-drop chart to make space (it isn't very interesting, pretty much flat for all cycles including a really big crisis). [...]


I reserved a separate answer to this suggestion. You are dismissing one of the most important parts of the new VPW spreadsheet.

This information is pretty critical. It is all the more useful when you select the Shiller data set which covers a little less than 150 years of history. What this graph shows you is that you MUST expect a big drop in withdrawals, and it quantifies this drop. Historically, very few retirements, using a 40/60 allocation didn't suffer through a 35% drop (maybe more, I'd have to look) in withdrawals.

If you're planning for retirement, this gives you a guideline for how much more money you must have, based on the start withdrawal and your absolute minimal spending floor.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Thu Dec 26, 2013 11:31 am

Weren't you looking for feedback a few posts ago? Oh well.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Thu Dec 26, 2013 10:58 pm

I think that I am getting way too emotional about something that was suposed to be a fun project. I'll take a little while away from the computer.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

garypeterson
Posts: 1
Joined: Sat Dec 28, 2013 4:51 pm

Re: Variable Percentage Withdrawal (VPW)

Post by garypeterson » Sat Dec 28, 2013 6:38 pm

I've found the VPW method discussion and backtesting fascinating and helpful -- both the spreadsheet pointed to in this thread as well as www.cfiresim.com implementation.

After looking at it a bit, I realized the withdrawal amounts exactly matched the 'actuarial process' described in http://howmuchcaniaffordtospendinretirement.webs.com/ and associated spreadsheets. The difference is that VPW seems to advocate a static calculation of the withdrawal percentages whereas the 'actuarial process' advocates a recalculation of the withdrawal amount annually, updating the parameters as appropriate. I point this out as I find some of the insights on the 'actuarial process' webpage helped put VPW in context. That page also has some insight into how this would interact w/annuities and social security -- some topics some have brought up here.

The other thing I realized is that if the inflation and return amounts match those used when calculating the VPW per-year withdrawal percentages, the result will be exactly the same as a constant withdrawal method. That may be obvious to some, but took me a while to sort through.

I'm wrestling w/whether this approach -- a statically-calculated percentage -- would result in predictable enough withdrawals and, if not, what minimal smoothing function would be needed.

Thanks for taking the time to develop this and share w/the community!

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Dec 29, 2013 9:59 am

siamond wrote:Weren't you looking for feedback a few posts ago? Oh well.


siamond, I needed to take a step back. I apologize. The points you raised are sensible.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Dec 29, 2013 10:51 am

garypeterson wrote:I've found the VPW method discussion and backtesting fascinating and helpful -- both the spreadsheet pointed to in this thread as well as http://www.cfiresim.com implementation.

After looking at it a bit, I realized the withdrawal amounts exactly matched the 'actuarial process' described in http://howmuchcaniaffordtospendinretirement.webs.com/ and associated spreadsheets.


I didn't know about the Blanchett actuarial approach, but I suspected that the VPW idea wasn't new. Now, I know that it wasn't. Thanks for the link.

The difference is that VPW seems to advocate a static calculation of the withdrawal percentages whereas the 'actuarial process' advocates a recalculation of the withdrawal amount annually, updating the parameters as appropriate. I point this out as I find some of the insights on the 'actuarial process' webpage helped put VPW in context. That page also has some insight into how this would interact w/annuities and social security -- some topics some have brought up here.


The Actuarial method advocates a new calculation every year with updated:
  1. Nominal market return predictions based on current valuations.
  2. Inflation prediction.
  3. Plan length = current life expectancy according to mortality tables + 2 years.

It allows to specify a bequest amount, which is used instead of $0 as final value in the calculations.

It does not automatically try to smooth withdrawals, but the paper lists a smoothing method (e.g. average of 2 or 3 years).

Pensions/SS are not really taken into account other than being added to the calculated amount.

The other thing I realized is that if the inflation and return amounts match those used when calculating the VPW per-year withdrawal percentages, the result will be exactly the same as a constant withdrawal method. That may be obvious to some, but took me a while to sort through.

I'm wrestling w/whether this approach -- a statically-calculated percentage -- would result in predictable enough withdrawals and, if not, what minimal smoothing function would be needed.

Thanks for taking the time to develop this and share w/the community!


I find the Actuarial withdrawal method quite interesting.

It's a question of compromise. The principal problem is uncertainty: we don't know what inflation, market returns, and our date of death will be. Is using yearly predictions any better than using static values and revising the plan when life conditions change? I just don't know.

The biggest drawbacks that I see, relative to VPW, is:
  • The need to predict the market and inflation every year. I try to imagine myself at 85 years old having to do so. I just don't like it.
  • I can't backtest this method. How do I find what market and inflation predictions were in January 1929?
  • As life expectancy increases every year you get older, the calculated withdrawal rate is higher than with a conservative fixed plan length (like 35 years for a 65 years-old person).
  • I think that (the proposed) smoothing will just get things worse when markets drop sharply. I think that smoothing is better done through proper asset allocation.

I am uploading a new version of the spreadsheet version that includes "worst nominal drop" statistics. Using a bond-heavy portfolio does make a significant difference, in nominal terms. What happens is that when CPI-U increases quickly, real withdrawals drop fast for both bond- and stock-heavy portfolios. In nominal terms, bond-heavy portfolios (but not too heavy) don't drop much. Maybe something like 15% to 20%. There is additional purchase power lost to inflation, but that might not be immediately felt by the retiree. A retiree often has the choice of changing his spending habits and select items that are not as affected by inflation (in the short term). After a while real withdrawals increase back. That's when nominal withdrawals catch back with CPI-U.

I really believe that smoothing formula tricks can't replace asset allocation for smoothing withdrawals.

As an example, here are the statistics for 1966 for a 40/60 portfolio:

Code: Select all

Simulation Statistics      
Median      $40 323
95% Above      $26 306
Minimum      $24 168
Worst Drop      -47,1%
Worst Nominal Drop      -12.3%

In nominal terms, withdrawals never regressed more than 12.3% (current withdrawal relative to highest preceding withdrawal). There were years with very high inflation. Nominal withdrawals were slow to catch up with inflation, but they didn't drop in any significant way. That can be quite comforting for a retiree.

Anyway, thanks. I think that VPW users should be aware of the Actuarial method and choose whatever method they prefer.
Last edited by longinvest on Sun Dec 29, 2013 6:22 pm, edited 2 times in total.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Dec 29, 2013 11:25 am

New answers.

siamond wrote:=> I would actually suggest to add REITs in the mix (for the Simba data set), since so many people use them even in very simple portfolios (cf. decorrelated returns).


I don't see how to add REITs, Gold, Small-Value, and all the various other asset classes. How do you find the long-term growth trend of these asset classes?

Do you propose, instead, to do something similar to the Actuarial approach, making market predictions every year? How do you predict the return of REITs?

=> I would suggest to restore this approach, so that all important parameters the end user will typically want to play with are in the primary tab.


I guess that this is your solution to the market prediction problem. That would explain things. I see the direct selection of a start rate as simply an indirect way to make a growth prediction. But there is still an implicit prediction. VPW simply recovers this prediction by reversing the calculation based on plan length. Do we make that the default approach? I'm not convinced.

=> (while leaving the 'long term growth trend' guesses where they are - including the ability to override - this is lower level logic).


One could argue that historical growth trends won't necessarily repeat in the future. I agree. But, will they be higher or lower? I'm not good at making predictions.

The problem compounds if we add alternate asset classes. I don't know an equivalent to the Credit Suisse Global Investment Yearbook that covers them.

I agree with cut-throat that showing the inflation on the primary tab (for the simulation table) would be a nice improvement to better understand the dynamics of the simulation. Just narrow a bit some columns (e.g. first two), and you'll easily create space for it.


OK. I've added worst nominal drop statistics, which is probably what we are trying to estimate here. Showing raw nominal withdrawals was not very helpful (in the previous version). The real withdrawal graph is more useful when looking at a complete withdrawal path. When you also consider the worst nominal drop, you get a hint of whether important drops were due to market drops or inflation.
Last edited by longinvest on Sun Dec 29, 2013 11:49 am, edited 2 times in total.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Dec 29, 2013 11:41 am

siamond wrote:Then the main missing thing for me is the ability to add a recurring income (e.g. SS/Pension) in the model.


It is difficult to model complete retirement scenarios in a single spreadsheet.

On the other hand, a complete software like cfiresim can handle:
  • Social Security (early / late, one or two different start dates)
  • Pensions (one or two different start date)
  • Planned expenses (big withdrawals)
Unfortunately, cfiresim does not seem to implement the recommended VPW calculation. It implements CVPW.


On a separate note, one could use VPW along with the worst-drop statistics to plan the bridge to pensions and Social Security instead of using 100% fixed income. It's a risk tradeoff. At least, the VPW spreadsheet allows backtesting it.
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 (VPW)

Post by Cut-Throat » Sun Dec 29, 2013 5:47 pm

longinvest wrote:I am uploading a new version of the spreadsheet version that includes "worst nominal drop" statistics. Using a bond-heavy portfolio does make a significant difference, in nominal terms. What happens is that when CPI-U increases quickly, real withdrawals drop fast for both bond- and stock-heavy portfolios. In nominal terms, bond-heavy portfolios (but not too heavy) don't drop much. Maybe something like 15% to 20%. There is additional purchase power lost to inflation, but that might not be immediately felt by the retiree. A retiree often has the choice of changing his spending habits and select items that are not as affected by inflation (in the short term). After a while real withdrawals increase back. That's when nominal withdrawals catch back with CPI-U.


+1 !!!!

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Sun Dec 29, 2013 6:15 pm

longinvest wrote:In nominal terms, withdrawals never regressed more than 12.3% (current withdrawal relative to highest preceding withdrawal). There were years with very high inflation. Nominal withdrawals were slow to catch up with inflation, but they didn't drop in any significant way. That can be quite comforting for a retiree.

Anyway, thanks. I think that VPW users should be aware of the Actuarial method and choose whatever method they prefer.


Yes, the inflation number is very important, especially to a heavy bond allocation holder.

With no backtesting of the Actuarial method, this makes VPW a 'slam dunk' !!

thanks again longinvest!

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Sun Dec 29, 2013 8:52 pm

Thanks for taking another pass at my comments, longinvest. Appreciated. You are spending long hours on this project, thank you for your dedication & patience.

longinvest wrote:I don't see how to add REITs, Gold, Small-Value, and all the various other asset classes. How do you find the long-term growth trend of these asset classes?

I was only suggesting REITs (no other asset class) since a) many people do use them in a 4-funds portfolio of sorts b) I thought it would be interesting to see how introducing one more asset class with strong historical decorrelation would affect the outcome (including the undertones of equity/bonds balance that we're both pondering about, and the 'worst drop' effects).

I didn't think this would imply to make a long-term prediction (which would be unfortunate, agreed). I don't believe you included such a prediction for International Stocks in the VPW math? Couldn't you just use the same principle for REITs (i.e. same 'guess-timate' predictions as for regular stocks)? No big deal, I am just curious...

longinvest wrote:I guess that this is your solution to the market prediction problem. That would explain things. I see the direct selection of a start rate as simply an indirect way to make a growth prediction. But there is still an implicit prediction. VPW simply recovers this prediction by reversing the calculation based on plan length. Do we make that the default approach? I'm not convinced.

You are quite right, I am quite wary about the automated rate computation, as it is a prediction that I have troubles to take at face value, so I really like the ability to play by myself and see what goes in backtesting... Plus I like the ability to play with a higher or lower rate, to make the early years more or less aggressive (as we discussed when suggesting CVPW). Anyhoo, the way you presented things in the previous version wasn't making a judgment call if VPW or CVPW is better. It was just making a rate recommendation, and letting the user override and see the outcome on the charts & tables right away. Seemed very clear to me. Anyhoo, no big deal, we still have a similar ability in the new version. Peace! ;-)

longinvest wrote:OK. I've added worst nominal drop statistics, which is probably what we are trying to estimate here. Showing raw nominal withdrawals was not very helpful (in the previous version). The real withdrawal graph is more useful when looking at a complete withdrawal path. When you also consider the worst nominal drop, you get a hint of whether important drops were due to market drops or inflation.

Cool. I'm quite curious to see the result. Did you post a link to your new Excel sheet already (if yes, I missed it).

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Sun Dec 29, 2013 9:09 pm

longinvest wrote:
siamond wrote:Then the main missing thing for me is the ability to add a recurring income (e.g. SS/Pension) in the model.


It is difficult to model complete retirement scenarios in a single spreadsheet.

On the other hand, a complete software like cfiresim can handle:
  • Social Security (early / late, one or two different start dates)
  • Pensions (one or two different start date)
  • Planned expenses (big withdrawals)
Unfortunately, cfiresim does not seem to implement the recommended VPW calculation. It implements CVPW.

You're right, I was trying to cram too much in your worksheet. cFIREsim is indeed a much more complete tool, and this does address my point. The author is a nice guy, very open to suggestions, by the way, maybe we could ask him to display a recommended initial spend when selecting the CVPW method (he will then have to maintain the corresponding math with the latest Credit Suisse predictions though).

CVPW fits very well in the general cFIREsim presentation, asking for the 1st year spend & portfolio as the primary parameters (hence an implicit initial withdrawal rate). Also once you factor in additional income (SS, inheritance, etc), you will probably feel more need to play with the initial rate to get a sensible trajectory. And yes, I had a say in the matter, so I'm guilty as charged! :wink:

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Mon Dec 30, 2013 8:22 pm

siamond wrote:You're right, I was trying to cram too much in your worksheet. cFIREsim is indeed a much more complete tool, and this does address my point. The author is a nice guy, very open to suggestions, by the way, maybe we could ask him to display a recommended initial spend when selecting the CVPW method (he will then have to maintain the corresponding math with the latest Credit Suisse predictions though).


Thanks for talking me up Siamond. lol.

longinvest wrote:Unfortunately, cfiresim does not seem to implement the recommended VPW calculation. It implements CVPW.


So, because I seem to have to relearn the VPW method every time I talk about it, can you tell me the difference between the CVPW (that I implement in cFIREsim), and the VPW that you suggest?

VPW does the yearly table based on some "Expected Returns" and "Standard Deviation" of Stocks/Bonds, where the CVPW is a straight PMT of the initial spending? Is the Credit Suisse "predictions" you're talking about the source of these numbers? I actually had initially implemented the normal VPW method, but I think one of us shot it down for some reason or another for use on the site :oops:
35 - Married - Aiming for FI/ER in early 40s.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Mon Dec 30, 2013 11:16 pm

siamond wrote:Thanks for taking another pass at my comments, longinvest. Appreciated. You are spending long hours on this project, thank you for your dedication & patience.

longinvest wrote:I don't see how to add REITs, Gold, Small-Value, and all the various other asset classes. How do you find the long-term growth trend of these asset classes?

I was only suggesting REITs (no other asset class) since a) many people do use them in a 4-funds portfolio of sorts b) I thought it would be interesting to see how introducing one more asset class with strong historical decorrelation would affect the outcome (including the undertones of equity/bonds balance that we're both pondering about, and the 'worst drop' effects).

I didn't think this would imply to make a long-term prediction (which would be unfortunate, agreed). I don't believe you included such a prediction for International Stocks in the VPW math? Couldn't you just use the same principle for REITs (i.e. same 'guess-timate' predictions as for regular stocks)? No big deal, I am just curious...


US and International stocks are stocks. So, the trend should be relatively identical. Anyway, I'm using world stock growth (and the same for bonds) over 1900-2012 in the spreadsheet.

After looking into it, I think that it would be OK to consider REITs as stocks. They are part of the total stock index.

The problem remains: if I add REITs, I'm sure I'll get requests for every tilt under the sun. This will really complicate the spreadsheet.

Furthermore, I personally do not believe that this would help with anything; either the VPW approach makes sense, or it doesn't. The existing back-testing spreadsheet has convinced me that it does. There's no point trying to "calibrate" withdrawals with 1972-2012 data for specific tilts. It just makes no sense.

According to most of the serious writings (e.g. Bernstein, Bogle, etc.), there are only two broad category of investment assets: fixed income and stocks. In my opinion, commodities are, by nature, speculative (they produce no income), so you can't apply the dividend discount model to them.

So, if someone has a tilted portfolio, they should be able to determine the equivalent "normal" stock/bond ratio, enter it in VPW and get the appropriate percentages.

longinvest wrote:I guess that this is your solution to the market prediction problem. That would explain things. I see the direct selection of a start rate as simply an indirect way to make a growth prediction. But there is still an implicit prediction. VPW simply recovers this prediction by reversing the calculation based on plan length. Do we make that the default approach? I'm not convinced.

You are quite right, I am quite wary about the automated rate computation, as it is a prediction that I have troubles to take at face value, so I really like the ability to play by myself and see what goes in backtesting...

Plus I like the ability to play with a higher or lower rate, to make the early years more or less aggressive (as we discussed when suggesting CVPW). Anyhoo, the way you presented things in the previous version wasn't making a judgment call if VPW or CVPW is better. It was just making a rate recommendation, and letting the user override and see the outcome on the charts & tables right away. Seemed very clear to me. Anyhoo, no big deal, we still have a similar ability in the new version. Peace! ;-)


I think that calibrating the aggressiveness (if we really want that*) should be done through a dedicated parameter, not by letting users guess the Start Percentage. It really makes no sense, to me, to require users to guess a Start Percentage.

VPW needs a general growth trend for bonds and for stocks. This trend is not the "ahead from now prediction", and VPW doesn't, in any way, try to hide speculative (e.g. short term) returns. VPW does adjust withdrawals, every year, on the current portfolio value.

[edited:] Think about it. Let say that you start with a long-term "ahead from now" prediction. Assume that you are at the top of a huge bubble. It would make sense that the long-term "ahead from now" prediction would be quite low. If you use that prediction as internal rate, you'll calculate relatively low percentages. Assuming that the bubble deflates during the first year, you'll get "too low" percentages for the rest of your retirement. You can repeat with the reverse scenario, starting retirement during a market crash.

What VPW needs is a growth trend that spans bubbles and crashes. It doesn't need to be precise; VPW adjusts withdrawals every year. It shouldn't be too low, otherwise you'll get the same skew delaying spending to old age as 1/N.

The current growth trends of VPW (based on historical world markets) seem OK to me, unless you can convince me otherwise. You haven't, so far. :)

*: I don't know of other methods that actively try to increase constant-dollar withdrawals with time. As for reducing them with time, I think that the variable nature of VPW has sufficient provisions to reduce (temporarily) withdrawals due to market misbehavior.

longinvest wrote:OK. I've added worst nominal drop statistics, which is probably what we are trying to estimate here. Showing raw nominal withdrawals was not very helpful (in the previous version). The real withdrawal graph is more useful when looking at a complete withdrawal path. When you also consider the worst nominal drop, you get a hint of whether important drops were due to market drops or inflation.

Cool. I'm quite curious to see the result. Did you post a link to your new Excel sheet already (if yes, I missed it).


It's already in the official spreadsheet: viewtopic.php?f=10&t=120430&p=1761580#p1761563
Last edited by longinvest on Wed Jan 01, 2014 4:15 pm, edited 3 times in total.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Mon Dec 30, 2013 11:51 pm

Hi boknows,

boknows wrote:
longinvest wrote:Unfortunately, cfiresim does not seem to implement the recommended VPW calculation. It implements CVPW.


So, because I seem to have to relearn the VPW method every time I talk about it, can you tell me the difference between the CVPW (that I implement in cFIREsim), and the VPW that you suggest?

VPW does the yearly table based on some "Expected Returns" and "Standard Deviation" of Stocks/Bonds, where the CVPW is a straight PMT of the initial spending? Is the Credit Suisse "predictions" you're talking about the source of these numbers? I actually had initially implemented the normal VPW method, but I think one of us shot it down for some reason or another for use on the site :oops:


Nice software. Congratulations!

If you download the latest spreadsheet (viewtopic.php?f=10&t=120430&p=1761580#p1761563) you'll see that on the main VPW sheet, the user only has to specify the following values to obtain a Variable Percentage Withdrawal Table:
  • his age
  • the number of years to deplete the portfolio
  • his asset allocation
This should be the default VPW configuration. The user shouldn't need to provide any start rate or internal rate. Of course, for simulation, he should be asked for a portfolio size (but not for a start amount).

For advanced users, the current spreadsheet allows two ways to override internal rates. These advanced settings are on the Table sheet. They are not intended for normal VPW usage; their use is mostly meant for experimentation. The current setup allowed to combine VPW and CVPW. You get CVPW by overriding the Start Percentage on the Table sheet.

siamond is fond of the CVPW approach, but I am not as much as he is, specially for normal users. I have explained why in viewtopic.php?f=10&t=120430&start=250#p1902927. I am not opposed to allow users to play with the advanced parameters, but they certainly should not appear as parameters that a user is expected to play with. This is why they are located in a separate sheet and they require to use an "override" setting.

If you'd like, I could spend some time looking at cfiresim to better understand its user interface and maybe make some recommendations on how to integrate VPW. In the short term, maybe the simplest would be to have 2 distinct withdrawal methods: VPW and CVPW (with a warning, maybe, that this is not the officially recommended approach?).

You've put a lot of interesting options in cFIREsim: pensions, social security, additional spending. Integrating these with VPW without breaking its "never fails" property will probably require some additional thinking and discussion. It would be nice if you agreed that we discuss about it here.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Tue Dec 31, 2013 1:24 pm

longinvest wrote:Hi boknows,

boknows wrote:
longinvest wrote:Unfortunately, cfiresim does not seem to implement the recommended VPW calculation. It implements CVPW.


So, because I seem to have to relearn the VPW method every time I talk about it, can you tell me the difference between the CVPW (that I implement in cFIREsim), and the VPW that you suggest?

VPW does the yearly table based on some "Expected Returns" and "Standard Deviation" of Stocks/Bonds, where the CVPW is a straight PMT of the initial spending? Is the Credit Suisse "predictions" you're talking about the source of these numbers? I actually had initially implemented the normal VPW method, but I think one of us shot it down for some reason or another for use on the site :oops:


Nice software. Congratulations!

If you download the latest spreadsheet (viewtopic.php?f=10&t=120430&p=1761580#p1761563) you'll see that on the main VPW sheet, the user only has to specify the following values to obtain a Variable Percentage Withdrawal Table:
  • his age
  • the number of years to deplete the portfolio
  • his asset allocation
This should be the default VPW configuration. The user shouldn't need to provide any start rate or internal rate. Of course, for simulation, he should be asked for a portfolio size (but not for a start amount).

For advanced users, the current spreadsheet allows two ways to override internal rates. These advanced settings are on the Table sheet. They are not intended for normal VPW usage; their use is mostly meant for experimentation. The current setup allowed to combine VPW and CVPW. You get CVPW by overriding the Start Percentage on the Table sheet.

siamond is fond of the CVPW approach, but I am not as much as he is, specially for normal users. I have explained why in viewtopic.php?f=10&t=120430&start=250#p1902927. I am not opposed to allow users to play with the advanced parameters, but they certainly should not appear as parameters that a user is expected to play with. This is why they are located in a separate sheet and they require to use an "override" setting.

If you'd like, I could spend some time looking at cfiresim to better understand its user interface and maybe make some recommendations on how to integrate VPW. In the short term, maybe the simplest would be to have 2 distinct withdrawal methods: VPW and CVPW (with a warning, maybe, that this is not the officially recommended approach?).

You've put a lot of interesting options in cFIREsim: pensions, social security, additional spending. Integrating these with VPW without breaking its "never fails" property will probably require some additional thinking and discussion. It would be nice if you agreed that we discuss about it here.


Yeah, like I said, I think I implemented this before on the site, before scrapping it for one reason or another. I think that it would be easy to implement.

What is the "never fails" property? It seems that the calculations give a very specific percent withdrawal (of current portfolio) for each year of the simulation, and inherently, as long as that is not 100%, it will never fail. Right?

I can provide a basic version of what I think the VPW is on the development part of the site, and you can see what you think (by looking at the CSV output).
35 - Married - Aiming for FI/ER in early 40s.

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Tue Dec 31, 2013 3:35 pm

boknows wrote:What is the "never fails" property? It seems that the calculations give a very specific percent withdrawal (of current portfolio) for each year of the simulation, and inherently, as long as that is not 100%, it will never fail. Right?


Not really, it is much more than that. The Constant dollar withdrawal plans fail when the retiree continues to withdraw the initial withdrawal amount plus the inflation adjustment into the teeth of a bear market or inflation. In reality this is not the natural reaction or the intelligent one. After 10 years of this kind of spending, the portfolio is damaged so bad that it cannot recover when the turnaround finally comes.

VPW insures that you are withdrawing more during market upturns and less during market downturns. You don't have to wait until 8-10 years down the road to realize that you are in trouble. With VPW an adjustment of about 10-20% less at the start of a bear market will insure portfolio survivability. This is just like buying low and selling high in reverse.

Download the VPW tool and backtest it against a poor year to retire and see how a little adjustment at the right time, saves your retirement spending. It also lets you spend far more money in the early years than a constant dollar plus inflation plan...... And it usually spends more money and leaves less on the table than other withdrawal plans.

I start my VPW withdrawal Plan tomorrow !! As it is a New Year and the first of my Retirement withdrawal Plan.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Tue Dec 31, 2013 5:36 pm

boknows wrote:Yeah, like I said, I think I implemented this before on the site, before scrapping it for one reason or another. I think that it would be easy to implement.


Yes, it should be easy. The formula is pretty simple.

What is the "never fails" property?


It means that VPW will not deplete the portfolio before the chosen year.

It seems that the calculations give a very specific percent withdrawal (of current portfolio) for each year of the simulation, and inherently, as long as that is not 100%, it will never fail. Right?


Yes, this is right. But, if you allow for specific portfolio withdrawals in addition to the percentage withdrawals, you can break this property. I am concerned with the integration of the "Income/Savings/SS/Pension" and "Spending Inputs" sections with VPW withdrawals on the cfiresim form.

Mainly, I'd like to discuss how one should plan for bridging retirement to social security and pensions, as well as planned extra expenses, without breaking the no-failure property. I do not know how (or if) you do this, currently, in cfiresim, when VPW withdrawal is selected.

I can provide a basic version of what I think the VPW is on the development part of the site, and you can see what you think (by looking at the CSV output).


That would be neat.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Tue Dec 31, 2013 5:52 pm

Cut-Throat wrote:I start my VPW withdrawal Plan tomorrow !! As it is a New Year and the first of my Retirement withdrawal Plan.


A real life test of VPW! I really didn't anticipate that someone would adopt the idea so soon, when I started this thread. I was just trying to get feedback from the Bogleheads on a method I was developing for my own retirement, many years from now.

I would like to thank you, Cut-Throat, for all the improvements that you have suggested. Thanks to your feedback and to the feedback of other Bogleheads members, VPW is now simpler and only requires 3 pieces of information to generate a personalized table: start age, depletion years, and asset allocation.

Happy New Year 2014!

longinvest
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 (VPW)

Post by Cut-Throat » Wed Jan 01, 2014 8:53 am

longinvest wrote:A real life test of VPW! I really didn't anticipate that someone would adopt the idea so soon, when I started this thread. I was just trying to get feedback from the Bogleheads on a method I was developing for my own retirement, many years from now.

I would like to thank you, Cut-Throat, for all the improvements that you have suggested. Thanks to your feedback and to the feedback of other Bogleheads members, VPW is now simpler and only requires 3 pieces of information to generate a personalized table: start age, depletion years, and asset allocation.

Happy New Year 2014!

longinvest


Actually, I was committed to a Variable Withdrawal Method for the last 10 years or so. I had seen some VW Plans, but never one that had backtesting incorporated into it. The VPW tool that you have developed is the cream of the crop!.....The backtesting feature has given me the confidence to employ it to a 'T''. It has also upped my withdrawal percentage to allow me to spend more in the early years.

And the real 'eye opener' was the knowledge that pulling more money out of the markets in Up Years, helped your portfolio much more than 'letting it ride' into a downturn. That was my 'real epiphany of 2013!

Thanks again for a great tool and Happy 2014 to you as well.

Retirement is off to a great start as it will be 81 degrees today here in Florida! 8-) 8-)

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Wed Jan 01, 2014 10:11 am

boknows wrote:Yeah, like I said, I think I implemented this before on the site, before scrapping it for one reason or another. I think that it would be easy to implement.


Happy new year, everybody!

The only difference between VPW and CVPW is the automated or custom way to compute the initial withdrawal rate (or the initial spend).

Well, given the way the cFIREsim UI is designed, I guess you probably have two main options:

1. Unified method: when the users selects [C]VPW, then display a 'recommended' initial spending amount right below the corresponding field (and let the end user decide to follow the advice or not). Careful, if the user modifies the AA, then you need to update the displayed recommendation.
2. Separate methods: add a new withdrawal method dubbed 'regular VPW', have it work exactly like CVPW, except that the field when the user typically enters the initial spending amount is superseded with the (mandatory!) recommended amount, and greyed out.

Personally, the first approach seems more sensible to me, but hey, both approaches should work fine.

PS. initial spend and initial withdrawal rate are congruent, of course. If one is used to interact with end users, the other isn't necessary.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Wed Jan 01, 2014 4:13 pm

siamond wrote:
boknows wrote:Yeah, like I said, I think I implemented this before on the site, before scrapping it for one reason or another. I think that it would be easy to implement.


Happy new year, everybody!



Happy New Year, too!


The only difference between VPW and CVPW is the automated or custom way to compute the initial withdrawal rate (or the initial spend).

Well, given the way the cFIREsim UI is designed, I guess you probably have two main options:

1. Unified method: when the users selects [C]VPW, then display a 'recommended' initial spending amount right below the corresponding field (and let the end user decide to follow the advice or not). Careful, if the user modifies the AA, then you need to update the displayed recommendation.


This would lead the user to think that CVPW is a recommended approach. It is not recommended by its designer (myself), based on the arguments presented in the following post: viewtopic.php?f=10&t=120430&start=250#p1902927.


2. Separate methods: add a new withdrawal method dubbed 'regular VPW', have it work exactly like CVPW, except that the field when the user typically enters the initial spending amount is superseded with the (mandatory!) recommended amount, and greyed out.

Personally, the first approach seems more sensible to me, but hey, both approaches should work fine.


@boknows: I think that, if you still wish to provide CVPW as a withdrawal method, it should be done separately from VPW.


PS. initial spend and initial withdrawal rate are congruent, of course. If one is used to interact with end users, the other isn't necessary.


@siamond: Of course, everyone is entitled to his opinion. I have provided my arguments (in viewtopic.php?f=10&t=120430&start=250#p1902927) as to why I think that the internal rates of VPW should not be tampered with by normal users (directly or through CVPW). I would really enjoy reading your counter arguments to better understand why you insist on exposing casual users to CVPW.

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

jaj2276
Posts: 307
Joined: Sat Apr 16, 2011 5:13 pm

Re: Variable Percentage Withdrawal (VPW)

Post by jaj2276 » Wed Jan 01, 2014 7:12 pm

First of all, thanks for everyone who's contributed to this thread. Thanks of course to longinvest for actually making (and continuing to enhance) the spreadsheet.

My one comment is that this spreadsheet makes me think that I'm being too conservative in my asset allocation as I'm making my way towards retirement. It seems in many scenarios, increasing % of bonds doesn't change the minimum yearly withdrawal that much AND the total withdrawals from the portfolio don't increase (and likely decrease). Take a bad starting year (Shiller 1966) as an example. A $1M portfolio split 60/40 leads to $1.2M of withdrawals with a minimum yearly withdraw of $25k (1982). If you make the portfolio 100% US stocks, you get $1.2M of withdrawals and a minimum yearly withdrawal of $25k (also in 1982).

Now take a good starting year (Shiller 1982). 100% stocks leads to total withdrawals of $4.55M and a minimum yearly withdrawal of $62k. A 60/40 leads to a minimum yearly withdrawal of $53k (worse) and total withdrawals of $3.5M (much worse).

I've looked at other starting years and they seem to follow the same pattern.

So what am I missing here? Am I making the cardinal sin of assuming that future markets (over tens of years) will behave similarly?

User avatar
siamond
Posts: 3332
Joined: Mon May 28, 2012 5:50 am

Re: Variable Percentage Withdrawal (VPW)

Post by siamond » Wed Jan 01, 2014 7:32 pm

Ok, I think I am done with this thread. Sorry if I unintentionally upset some people. :?

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Thu Jan 02, 2014 12:13 am

siamond wrote:Ok, I think I am done with this thread. Sorry if I unintentionally upset some people. :?


siamond, you have positively contributed to this thread. It would be a loss if you quit.

I do sincerely have trouble understanding your position on CVPW. I have expressed my reasoning. I guess that you are not interested in providing counter arguments. OK, I won't insist more.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Thu Jan 02, 2014 1:02 am

jaj2276 wrote:First of all, thanks for everyone who's contributed to this thread. Thanks of course to longinvest for actually making (and continuing to enhance) the spreadsheet.

My one comment is that this spreadsheet makes me think that I'm being too conservative in my asset allocation as I'm making my way towards retirement. It seems in many scenarios, increasing % of bonds doesn't change the minimum yearly withdrawal that much AND the total withdrawals from the portfolio don't increase (and likely decrease). Take a bad starting year (Shiller 1966) as an example. A $1M portfolio split 60/40 leads to $1.2M of withdrawals with a minimum yearly withdraw of $25k (1982). If you make the portfolio 100% US stocks, you get $1.2M of withdrawals and a minimum yearly withdrawal of $25k (also in 1982).

Now take a good starting year (Shiller 1982). 100% stocks leads to total withdrawals of $4.55M and a minimum yearly withdrawal of $62k. A 60/40 leads to a minimum yearly withdrawal of $53k (worse) and total withdrawals of $3.5M (much worse).

I've looked at other starting years and they seem to follow the same pattern.

So what am I missing here? Am I making the cardinal sin of assuming that future markets (over tens of years) will behave similarly?


Hi jaj2276,

A higher allocation to equities brings higher volatility. This is why the spreadsheet provides some important information:
  • Minimum withdrawal during the simulation.
  • Worst withdrawal drop during the simulation.
  • Worst nominal withdrawal drop during the simulation.
The spreadsheet also shows rolling amounts for up to 1871-2012, using the Shiller data set.

Yes, you are likely to get more money using a more aggressive portfolio, but, you'll also get much bigger nominal withdrawal drops, and even, sometimes, lower minimum withdrawals, when returns are poor. Usually, people that depend on their portfolio to live like withdrawals to stay above a certain floor, and not to drop too dramatically in nominal terms; it helps with budgeting/planning.

When you look at "good" retirement years, such as 1982, you feel like you'd like a 100/0 portfolio. But, when you look at a year like 1966, a 100/0 is much uglier to live with than a 40/60 portfolio. To see that, you have to put yourself in the shoes of the retiree that does not know, in 1966, that valuations are high and high inflation is coming. He starts withdrawing. Look at the worst withdrawals drops in nominal terms:
  • 100/0: 39.7%
  • 40/60: 12.3%
Now, I don't know about you, but me, I would have an easier time reducing my budget by 12.3% than by 39.7%.

It is even more revealing to look at the actual nominal withdrawal numbers (on the Path sheet):
Year 100/0 40/60
1966 $58,501 $45,672
1967 $52,108 $44,533
1968 $57,539 $46,426
1969 $60,576 $48,061
1970 $52,822 $45,487
1971 $53,801 $48,982
1972 $58,271 $52,116
1973 $65,280 $55,223
1974 $52,209 $51,905
1975 $39,383 $48,405
1976 $51,946 $56,065
1977 $55,065 $59,612
1978 $47,929 $57,203
...

Look at the drops of 1975 and 1978. Don't forget that inflation is high and eating away your spending power. The 100/0 portfolio is really difficult to budget for, even though many years, you have higher withdrawals than with 40/60.

The 40/60 portfolio, on the other hand, edges slowly a little higher almost every year, and even though it goes at it slower than inflation, it does not expose you to high volatility. It is easy to depend on it for budgeting. Eventually, it catches up to inflation (thanks to the 40% stocks).

Remember that CPI-U is only an indirect measure of inflation. Your personal inflation rate can be different and somewhat controlled by making different spending decisions.

longinvest
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 (VPW)

Post by Cut-Throat » Thu Jan 02, 2014 8:16 am

Excellent Explanation Longinvest !

You can't do anything about the markets, but you do have 'some' control over your personal inflation rate.

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Thu Jan 02, 2014 1:06 pm

longinvest wrote:Yes, this is right. But, if you allow for specific portfolio withdrawals in addition to the percentage withdrawals, you can break this property. I am concerned with the integration of the "Income/Savings/SS/Pension" and "Spending Inputs" sections with VPW withdrawals on the cfiresim form.

Mainly, I'd like to discuss how one should plan for bridging retirement to social security and pensions, as well as planned extra expenses, without breaking the




So, the way that cFIREsim works (similar to FireCalc), is that any adjustments to the portfolio (one-time or recurring spending, one-time or recurring savings, pension, social security etc) are applied to the portfolio first as an adjustment, THEN the "spending" amount is calculated (no matter which withdrawal method you're using).

So, I suppose if you have $100k in your portfolio at a given point, and you have $101k 1-time spending adjustment, it will fail the portfolio. But, that is sort-of the nature of these 1-time spending/savings adjustments, no?

Do you have thoughts on how to change the way things are calculated to keep the "never fail" property intact here?
35 - Married - Aiming for FI/ER in early 40s.

jaj2276
Posts: 307
Joined: Sat Apr 16, 2011 5:13 pm

Re: Variable Percentage Withdrawal (VPW)

Post by jaj2276 » Thu Jan 02, 2014 8:54 pm

longinvest wrote:
jaj2276 wrote: ... Stuff deleted ...


Hi jaj2276,

A higher allocation to equities brings higher volatility. This is why the spreadsheet provides some important information:
  • Minimum withdrawal during the simulation.
  • Worst withdrawal drop during the simulation.
  • Worst nominal withdrawal drop during the simulation.

... Even better stuff deleted ...

longinvest


Yes, this all makes sense. I guess I'm viewing this through a slightly different lens.

If I believe that my living expenses won't increase to match the good withdrawals and the bad withdrawal years are still "good enough" to cover my living expenses, then I should be able to handle large nominal drops and could choose a portfolio that maximizes my total withdrawals. Speaking only to my situation, I feel it's likely that even during the good withdrawal years, my discretionary spending won't fill the gap between living expenses and withdrawal amount. Then during bad withdrawal years, I'll have previous year's "savings" to supplement my withdrawals if need be.

I'm currently in Age-10 in bonds. By the time I hope to retire (age 50), I hope to be at a 60/40 allocation. Even though I'm sort of advocating for a 100% equities portfolio given your spreadsheet, I'm not likely to then go all in on stocks (i.e. sell my 40% bonds to buy stocks) at retirement. However, I've read research recently that mentioned that one should INCREASE their stock allocation during retirement (basically be at your most conservative right at retirement and then have the portfolio become less conservative throughout retirement). What are your thoughts on adding a variable to increase/decrease bonds (or decrease/increase stocks, whichever makes more sense) over the depletion years?

So if I was 60/40 at retirement and had depletion years set to 40, I might say -1% for bonds (in some new cell/variable). Then for each year, my portfolio isn't 60/40 but rather (.60 - (-.01 * (1 - year))) / (.40 + (-.01 * (1 - year)))?

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Fri Jan 03, 2014 10:21 pm

longinvest wrote:
boknows wrote:I can provide a basic version of what I think the VPW is on the development part of the site, and you can see what you think (by looking at the CSV output).

That would be neat.


www.cfiresim.com/dev/input.php is the "development" version of the site. The beginnings of the "Original VPW" (we can always change the name) method in the "Adjust Spending" menu. It calculates the WR correctly, according to your spreadsheet. However, there are a few things that can break it, that I'd like to discuss:

1) What "Long Term Growth Rate" does the Credit Suisse provide for Cash allocations? Gold?
2) Is the way that SS/Pensions/Extra Spending that I mentioned in a previous post an alright exception to the "no fail" rule?
35 - Married - Aiming for FI/ER in early 40s.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sat Jan 04, 2014 11:47 am

jaj2276 wrote:I'm currently in Age-10 in bonds. By the time I hope to retire (age 50), I hope to be at a 60/40 allocation. Even though I'm sort of advocating for a 100% equities portfolio given your spreadsheet, I'm not likely to then go all in on stocks (i.e. sell my 40% bonds to buy stocks) at retirement. However, I've read research recently that mentioned that one should INCREASE their stock allocation during retirement (basically be at your most conservative right at retirement and then have the portfolio become less conservative throughout retirement). What are your thoughts on adding a variable to increase/decrease bonds (or decrease/increase stocks, whichever makes more sense) over the depletion years?

So if I was 60/40 at retirement and had depletion years set to 40, I might say -1% for bonds (in some new cell/variable). Then for each year, my portfolio isn't 60/40 but rather (.60 - (-.01 * (1 - year))) / (.40 + (-.01 * (1 - year)))?


I'm back to work, so I won't really have time to work on the spreadsheet until the Summer. I'll think about how I could try to model this, but it would complicate the user interface.

If someone wants to actually use a sliding AA during retirement, he only needs to look up the withdrawal percentage in a VPW Table specific to his current AA and age at the time of withdrawal. As his equity ratio and age change, he'll get adjusted withdrawal percentages.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sat Jan 04, 2014 12:27 pm

boknows wrote:http://www.cfiresim.com/dev/input.php is the "development" version of the site. The beginnings of the "Original VPW" (we can always change the name) method in the "Adjust Spending" menu. It calculates the WR correctly, according to your spreadsheet. However, there are a few things that can break it, that I'd like to discuss:

1) What "Long Term Growth Rate" does the Credit Suisse provide for Cash allocations? Gold?


Credit Suisse shoes the data for US Bills, but not for their international equivalents. Their return was near 0%. (0.9% 1900-2012, -0.3% 2000-2012). In my opinion, cash is not really an investment; it's savings. I would probably give it a growth rate of 0% or maybe a little less.

As for gold, I think that it is purely speculative: it doesn't generate any kind of income; it only generates costs (storage). I personally have a moral issue with cautioning its use as "normal" in a portfolio. So, I won't make any rate recommendation. Instead, I'll go as far as to say that it wouldn't be a loss if you didn't model it.

2) Is the way that SS/Pensions/Extra Spending that I mentioned in a previous post an alright exception to the "no fail" rule?


If I understand well, the current cfiresim VPW implementation models additional or reduced spending as follows:
There is a single portfolio. Each year:
  • The withdrawal mandated by the VPW percentage is made.
  • Any additional discretionary withdrawal is made.
  • Any pension income received is invested back into the portfolio (once the pension kicks in). Same for social security.
Have I got this right? If yes, then this is really not how it should be done!

The portfolio on which VPW is applied should be left alone during the whole retirement period. Any non-VPW withdrawal should be done PRIOR to starting the retirement (for modelling).

In other words, in case of pensions and social security, an appropriate withdrawal should be made prior to starting retirement to put aside funds for bridging the gap until the pension (or social security) begins. Similarly, an appropriate withdrawal must be made prior to starting retirement to fund discretionary spending.

The question that remains is how to invest the bridge and discretionary spending money. There are many schools of thought:
  • One approach is to buy TIPS strips which mature at the time money is needed. (Difficult, if even possible to do).
  • Buy CDs maturing at the time money is needed, for any money needed with the first 5 years of retirement. Invest the rest in a bond fund.
  • Invest pension/social security money in a separate portfolio and use VPW (with the depletion period set to the bridge length) to extract variable bridge withdrawals until the pension or social security kicks in.
  • Invest "fuzzy" future discretionary spending, in more than 5 years, in a separate portfolio. Spend whatever amount of money this portfolio evolves into when the time comes.

In the above, there are no reason for separate portfolios to share the same AA as the main retirement portfolio.

Modelling this is a challenge, but I see no good alternative. A naive implementation, as I think is currently implemented in cfiresim, is completely inappropriate, as it breaks so many VPW features!
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Sat Jan 04, 2014 1:51 pm

longinvest wrote: In my opinion, cash is not really an investment; it's savings. I would probably give it a growth rate of 0% or maybe a little less.


I agree, but people DO put it as part of their AA, so it should be accounted for to include as many people as possible. Currently, it allows for users to put the cash return % in themselves, since it varies from country to country so much. But, how do we include that in the VPW table formula?

As for gold, I think that it is purely speculative: it doesn't generate any kind of income; it only generates costs (storage). I personally have a moral issue with cautioning its use as "normal" in a portfolio. So, I won't make any rate recommendation. Instead, I'll go as far as to say that it wouldn't be a loss if you didn't model it.


Again, I'm pretty close to agreement with how you think of gold.... but it is a major driving factor for a lot of people's AA's. To include it in the VPW formula, should we just use the historical average? Should be make the user able to input it?

boknows wrote:2) Is the way that SS/Pensions/Extra Spending that I mentioned in a previous post an alright exception to the "no fail" rule?


longinvest wrote:If I understand well, the current cfiresim VPW implementation models additional or reduced spending as follows:
There is a single portfolio. Each year:
  • The withdrawal mandated by the VPW percentage is made.
  • Any additional discretionary withdrawal is made.
  • Any pension income received is invested back into the portfolio (once the pension kicks in). Same for social security.
Have I got this right? If yes, then this is really not how it should be done!


Pretty close. For each year, EndingPortfolio = StartingPortfolio+AnySpendingOrSavingsAdjustments-Spending.

longinvest wrote:The portfolio on which VPW is applied should be left alone during the whole retirement period. Any non-VPW withdrawal should be done PRIOR to starting the retirement (for modelling).

In other words, in case of pensions and social security, an appropriate withdrawal should be made prior to starting retirement to put aside funds for bridging the gap until the pension (or social security) begins. Similarly, an appropriate withdrawal must be made prior to starting retirement to fund discretionary spending.


Is this really practical, though? I can think of a few ways to break this. 1) What if SS+Pension is greater than the allowed withdrawal according to VPW? Doesn't that throw things for a loop? What if you plan on leaving your current residence during retirement, and move to a cheaper area, thus incurring a lump sum addition of $100k+ to your portfolio (more than your VPW spend)?

I started typing an example of a large one-time expense, but the way you phrased it, people should set that money aside before retirement.

longinvest wrote:The question that remains is how to invest the bridge and discretionary spending money. There are many schools of thought:
  • One approach is to buy TIPS strips which mature at the time money is needed. (Difficult, if even possible to do).
  • Buy CDs maturing at the time money is needed, for any money needed with the first 5 years of retirement. Invest the rest in a bond fund.
  • Invest pension/social security money in a separate portfolio and use VPW (with the depletion period set to the bridge length) to extract variable bridge withdrawals until the pension or social security kicks in.,
  • Invest "fuzzy" future discretionary spending, in more than 5 years, in a separate portfolio. Spend whatever amount of money this portfolio evolves into when the time comes.

In the above, there are no reason for separate portfolios to share the same AA as the main retirement portfolio.

Modelling this is a challenge, but I see no good alternative. A naive implementation, as I think is currently implemented in cfiresim, is completely inappropriate, as it breaks so many VPW features!


So how do you envision this happening in your own retirement? If your home needs a roof replaced and a furnace replaced in the same year (lets call that $30k), do you have a completely separate portfolio for these incidences, so that it doesn't bog down the VPW logic?

You're right, this could be challenging. What if, in the case of a portfolio adjustment (let's give 2 examples. a 4-yr period of recurring college tuition, and a 1-time inheritance)... what if we recalculated the VPW table every single year (not hard to do in a programming language). So, if you have a 1-time $200k inheritance or something in year 5 of 35, VPW would recalculate the spend rate for the next 30 years with the inclusion of that inheritance. Conversely, this could cover any extraneous income (like SS or pensions), because each year it would take into account the extra funds in your portfolio.

Thoughts?
35 - Married - Aiming for FI/ER in early 40s.

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Sat Jan 04, 2014 2:59 pm

boknows wrote:
longinvest wrote:The portfolio on which VPW is applied should be left alone during the whole retirement period. Any non-VPW withdrawal should be done PRIOR to starting the retirement (for modelling).

In other words, in case of pensions and social security, an appropriate withdrawal should be made prior to starting retirement to put aside funds for bridging the gap until the pension (or social security) begins. Similarly, an appropriate withdrawal must be made prior to starting retirement to fund discretionary spending.


Is this really practical, though? I can think of a few ways to break this. 1) What if SS+Pension is greater than the allowed withdrawal according to VPW? Doesn't that throw things for a loop?


I am doing exactly that !.....My wife and I won't take S.S. for 8 years. We have set aside about $50 grand a year for 8 years or $400 Grand in Cash. When we run VPW we input our starting Portfolio number as VPW Portfolio = Total Portfolio - $400 Grand.

Not very complicated at all!

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Sat Jan 04, 2014 8:35 pm

Cut-Throat wrote:
I am doing exactly that !.....My wife and I won't take S.S. for 8 years. We have set aside about $50 grand a year for 8 years or $400 Grand in Cash. When we run VPW we input our starting Portfolio number as VPW Portfolio = Total Portfolio - $400 Grand.

Not very complicated at all!


What you describe is slightly different... you're just bridging the gap between now and when you run VPW. What I'm saying is: What if you have a portfolio of $1M, and VPW says that your first year of spend should be $46k. What if you have SS and Pensions that equal $50k? You just spend $96k that year, but ignore the $50k for the sake of the VPW calculation? That's something I can deal with, logically... but I think that people generally using a tool online would like to have all of their income streams and expenditures (recurring and one-time) integrated into 1 system. They don't want to put aside $400k and just not include it in the tool.
35 - Married - Aiming for FI/ER in early 40s.

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Sat Jan 04, 2014 9:36 pm

boknows wrote:
Cut-Throat wrote:
I am doing exactly that !.....My wife and I won't take S.S. for 8 years. We have set aside about $50 grand a year for 8 years or $400 Grand in Cash. When we run VPW we input our starting Portfolio number as VPW Portfolio = Total Portfolio - $400 Grand.

Not very complicated at all!


What you describe is slightly different... you're just bridging the gap between now and when you run VPW. What I'm saying is: What if you have a portfolio of $1M, and VPW says that your first year of spend should be $46k. What if you have SS and Pensions that equal $50k? You just spend $96k that year, but ignore the $50k for the sake of the VPW calculation? That's something I can deal with, logically... but I think that people generally using a tool online would like to have all of their income streams and expenditures (recurring and one-time) integrated into 1 system. They don't want to put aside $400k and just not include it in the tool.


Not sure we're communicating here. I am running VPW now. In Fact I ran it Jan 1 for my first withdrawal. I don't include S.S. at all. I take the output of VPW and add another $50K out of the $400K I have set aside for my Spending.

When S.S. kicks in 8 years from now, I will continue to withdraw from my portfolio whatever VPW is telling me in addition to my S.S...... VPW is a portfolio withdrawal tool. Not a tool that calculates all income streams. Income streams such as Pensions, SS etc. go on until death, so they do not need to be 'calculated'. A portfolio has a finite amount of money. Two completely different things.

A lot of folks here keep trying to make VPW a Complete Retirement Planning Tool. It is not, and In my opinion should not try to be.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Jan 05, 2014 10:14 am

boknows wrote:
longinvest wrote: In my opinion, cash is not really an investment; it's savings. I would probably give it a growth rate of 0% or maybe a little less.


I agree, but people DO put it as part of their AA, so it should be accounted for to include as many people as possible. Currently, it allows for users to put the cash return % in themselves, since it varies from country to country so much. But, how do we include that in the VPW table formula?


InternalRate = (stockAllocation x stockRate) + (bondAllocation x bondRate) + (cashAllocation x cashRate)

As for gold, I think that it is purely speculative: it doesn't generate any kind of income; it only generates costs (storage). I personally have a moral issue with cautioning its use as "normal" in a portfolio. So, I won't make any rate recommendation. Instead, I'll go as far as to say that it wouldn't be a loss if you didn't model it.


Again, I'm pretty close to agreement with how you think of gold.... but it is a major driving factor for a lot of people's AA's. To include it in the VPW formula, should we just use the historical average? Should be make the user able to input it?


Gold being speculative, it is like lottery tickets. The safest bet is to consider any gold allocation similarly to a lottery ticket allocation: the money has already been spent. You subtract it from the portfolio start value and you forget about it.

longinvest wrote:The portfolio on which VPW is applied should be left alone during the whole retirement period. Any non-VPW withdrawal should be done PRIOR to starting the retirement (for modelling).

In other words, in case of pensions and social security, an appropriate withdrawal should be made prior to starting retirement to put aside funds for bridging the gap until the pension (or social security) begins. Similarly, an appropriate withdrawal must be made prior to starting retirement to fund discretionary spending.


Is this really practical, though? I can think of a few ways to break this. 1) What if SS+Pension is greater than the allowed withdrawal according to VPW? Doesn't that throw things for a loop? What if you plan on leaving your current residence during retirement, and move to a cheaper area, thus incurring a lump sum addition of $100k+ to your portfolio (more than your VPW spend)?

I started typing an example of a large one-time expense, but the way you phrased it, people should set that money aside before retirement.


If you don't have enough money to cover the bridge to Social Security and pensions, and to fund your discretionary spending, then you are certainly not ready to retire. A retirement planning tool should definitely not hide the lack of funding!

Yes, the pot of money on which VPW is applied should be separated from money needed for other things. If after a few years of applying VPW the retiree's life circumstances change, then nothing prevents him from revising his plans. But, I don't think that this should be part of the original planning.

longinvest wrote:The question that remains is how to invest the bridge and discretionary spending money. There are many schools of thought:
  • One approach is to buy TIPS strips which mature at the time money is needed. (Difficult, if even possible to do).
  • Buy CDs maturing at the time money is needed, for any money needed with the first 5 years of retirement. Invest the rest in a bond fund.
  • Invest pension/social security money in a separate portfolio and use VPW (with the depletion period set to the bridge length) to extract variable bridge withdrawals until the pension or social security kicks in.,
  • Invest "fuzzy" future discretionary spending, in more than 5 years, in a separate portfolio. Spend whatever amount of money this portfolio evolves into when the time comes.

In the above, there are no reason for separate portfolios to share the same AA as the main retirement portfolio.

Modelling this is a challenge, but I see no good alternative. A naive implementation, as I think is currently implemented in cfiresim, is completely inappropriate, as it breaks so many VPW features!


So how do you envision this happening in your own retirement? If your home needs a roof replaced and a furnace replaced in the same year (lets call that $30k), do you have a completely separate portfolio for these incidences, so that it doesn't bog down the VPW logic?


Long term house and car expenses, travel expenses, health expenses (and insurance) and other expenses should be part of a normal budget. Personally, I simply pluck money away for these on every pay check. These amounts can be quantified and planned for. So, during retirement, a part of every withdrawal should be put aside for these expenses.

That's normal budgeting, in my opinion.

You're right, this could be challenging. What if, in the case of a portfolio adjustment (let's give 2 examples. a 4-yr period of recurring college tuition, and a 1-time inheritance)... what if we recalculated the VPW table every single year (not hard to do in a programming language). So, if you have a 1-time $200k inheritance or something in year 5 of 35, VPW would recalculate the spend rate for the next 30 years with the inclusion of that inheritance. Conversely, this could cover any extraneous income (like SS or pensions), because each year it would take into account the extra funds in your portfolio.

Thoughts?


Money for any fixed expense within 5 years should probably not be invested. Stocks are way too volatile, and the duration of the Total Bond Market is over 5 years. So, the money should be dumped into savings accounts, CDs or maybe a short-term bond fund for convenience, if one has the flexibility to experience a small loss.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Jan 05, 2014 10:29 am

Cut-Throat wrote:When S.S. kicks in 8 years from now, I will continue to withdraw from my portfolio whatever VPW is telling me in addition to my S.S...... VPW is a portfolio withdrawal tool. Not a tool that calculates all income streams. Income streams such as Pensions, SS etc. go on until death, so they do not need to be 'calculated'. A portfolio has a finite amount of money. Two completely different things.

Thanks Cut-Throat for sharing your real-life experience using VPW. This is very valuable!
A lot of folks here keep trying to make VPW a Complete Retirement Planning Tool. It is not, and In my opinion should not try to be.


To be fair to boknows, he is asking about how to integrate VPW into his cfiresim retirement planning tool; he is not asking me to change the VPW spreadsheet. I did criticize the current VPW integration within cfiresim in respect to social security, pensions and additional spending.

I do think that discussing the larger picture of retirement planning with VPW as a component can be useful to readers of this thread.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

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

Re: Variable Percentage Withdrawal (VPW)

Post by boknows » Sun Jan 05, 2014 1:48 pm

longinvest wrote:
To be fair to boknows, he is asking about how to integrate VPW into his cfiresim retirement planning tool; he is not asking me to change the VPW spreadsheet. I did criticize the current VPW integration within cfiresim in respect to social security, pensions and additional spending.

I do think that discussing the larger picture of retirement planning with VPW as a component can be useful to readers of this thread.


Absolutely. I'm not trying to change VPW! Just trying to understand if it's possible to merge that method into something else.

So, would it be completely fallacious to represent "Spending" on cFIREsim (using VPW) as the following, using Cut-Throat's real life example of waiting on SS:

- Take your portfolio amount, calculate VPW table.
- "Spending" for any given year would equal = VPW withdrawal amount + Pensions + SS + Any other income stream listed. Since pensions and SS aren't coming out of the portfolio, but are income streams, they'll cancel eachother out. We're just figuring out the semantics of "spending" in relation to SS and pensions.

So, again, my only concern is for 1-time "expenditures" that folks might list that are large enough to throw things off (upgrading a house, college tuition, etc). I am not proposing changing VPW... just trying to figure out how to integrate it here. Is there no intuitive way to keep these in the fray? Should we simply set up an error for when people enter extra "spending" inputs, that says "VPW is a withdrawal calculation method, not a total retirement planning tool. Extra spending inputs are not compatible with this withdrawal method."? Or something to that effect.
35 - Married - Aiming for FI/ER in early 40s.

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

Re: Variable Percentage Withdrawal (VPW)

Post by Cut-Throat » Sun Jan 05, 2014 3:42 pm

OK, I follow what you are trying to do here. Sorry about that!

I think if someone inputs a one-time spending amount that may drive VPW below a certain level, the portfolio would fail. So you could add a 'Floor Level for VPW' and disallow the 1 time spending amount or just Let it fail. Lots of different options.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Jan 05, 2014 4:21 pm

Cut-Throat wrote:I think if someone inputs a one-time spending amount that may drive VPW below a certain level, the portfolio would fail. So you could add a 'Floor Level for VPW' and disallow the 1 time spending amount or just Let it fail. Lots of different options.


I disagree. I've been trying to explain that this is plain wrong in my previous posts. No additional one-time spending or other fixed spending should be taken out of VPW's dedicated portfolio.

I don't know how to be clearer.
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 (VPW)

Post by Cut-Throat » Sun Jan 05, 2014 4:35 pm

longinvest wrote:I disagree. I've been trying to explain that this is plain wrong in my previous posts. No additional one-time spending or other fixed spending should be taken out of VPW's dedicated portfolio.

I don't know how to be clearer.


Well, I thought we weren't modifying VPW. I was talking about modifying boknows retirement planner.

As I said before, I don't think VPW should be anything other than a Portfolio Withdrawal Tool.

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Jan 05, 2014 5:00 pm

boknows wrote:
longinvest wrote:
To be fair to boknows, he is asking about how to integrate VPW into his cfiresim retirement planning tool; he is not asking me to change the VPW spreadsheet. I did criticize the current VPW integration within cfiresim in respect to social security, pensions and additional spending.

I do think that discussing the larger picture of retirement planning with VPW as a component can be useful to readers of this thread.


Absolutely. I'm not trying to change VPW! Just trying to understand if it's possible to merge that method into something else.

So, would it be completely fallacious to represent "Spending" on cFIREsim (using VPW) as the following, using Cut-Throat's real life example of waiting on SS:

- Take your portfolio amount, calculate VPW table.
- "Spending" for any given year would equal = VPW withdrawal amount + Pensions + SS + Any other income stream listed. Since pensions and SS aren't coming out of the portfolio, but are income streams, they'll cancel eachother out. We're just figuring out the semantics of "spending" in relation to SS and pensions.


I disagree with your proposed approach. Nothing is cancelled out. It would be contrary to VPW's philosophy, for example, to delay social security until 70 and live on less money from 62 (or earlier) to 69. It is one of VPW's objectives to allow you to spend your portfolio while you are alive and healthy. VPW's philosophy requires bridging social security and pensions. This requires separate bridging funds managed in a liability-matching manner.

Here's a quick solution for you. To bridge the gap until social security and pensions kick in, just total up the missing payments and remove the total from VPW's dedicated portfolio.

Example:

age: 60
Asset Allocation: 40 stocks / 60 bonds
Depletion: 35 years
Total portfolio: $1,000,000
Social Security: $20,000/year starting at 70
Pension: none


Provision for bridging Social Security: $200,000
VPW Portfolio: $800,000

The $200,000 is for liability matching. The user should be informed that he must invest 180,000 in time-matched fixed income so that $20,000 becomes available every year from 61 to 69. Of course, the additional $20,000 is meant for spending immediately at 60 and need not be invested.

VPW's withdrawals rates are applied on the $800,000 portfolio, invested into 40% stocks / 60% bonds.

Every year, the user should get:
VPW withdrawal + $20,000 = available for spending

Starting at age 70, the $20,000 will come from social security, of course.

This is a simplified approach, as it assumes a 0% real growth (matching inflation) in the time-matched fixed income.

A non-indexed pension would actually require a little less money than this quick solution.

So, again, my only concern is for 1-time "expenditures" that folks might list that are large enough to throw things off (upgrading a house, college tuition, etc). I am not proposing changing VPW... just trying to figure out how to integrate it here. Is there no intuitive way to keep these in the fray? Should we simply set up an error for when people enter extra "spending" inputs, that says "VPW is a withdrawal calculation method, not a total retirement planning tool. Extra spending inputs are not compatible with this withdrawal method."? Or something to that effect.


Again, the simplest solution is to put money aside and assume 0% real growth.

Example:

age: 60
Asset Allocation: 40 stocks / 60 bonds
Depletion: 35 years
Total portfolio: $1,000,000
Social Security: $20,000/year starting at 70
Pension: none
One-time Spending: $25,000 at 65
One-time Spending: $25,000 at 70
One-time Spending: $25,000 at 75
One-time Spending: $25,000 at 80

Provision for bridging Social Security: $200,000
Provision for one-time spendings: $100,000
VPW Portfolio: $700,000

Again, it is important to inform the user that $300,000 should be invested in a liability-matching manner. That means fixed income.

A more precise approach would query the user for fixed-income rates for the required time frames (TIPS for planning inflation-adjusted future amounts, CDs and equivalents for nominal future amounts). Example:
CD Rates:
1 year:
2 years:
3 years:
etc.
Then, the provision funds would be discounted accordingly. Precise planning for future inflation-adjusted amounts at precise times is tricky as there are very few products offering both inflation and time guarantees (such as a TIPS strip maturing at the right time). The crude approximation of 0% real growth and investing money in CDs should be good enough in most cases.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2467
Joined: Sat Aug 11, 2012 8:44 am

Re: Variable Percentage Withdrawal (VPW)

Post by longinvest » Sun Jan 05, 2014 5:40 pm

Cut-Throat wrote:
longinvest wrote:I disagree. I've been trying to explain that this is plain wrong in my previous posts. No additional one-time spending or other fixed spending should be taken out of VPW's dedicated portfolio.

I don't know how to be clearer.


Well, I thought we weren't modifying VPW. I was talking about modifying boknows retirement planner.

As I said before, I don't think VPW should be anything other than a Portfolio Withdrawal Tool.


The problem is that using VPW within a larger retirement plan should be done correctly, otherwise, the most basic assumptions of VPW will be broken. The naive approach of using VPW on a portfolio to drive annual withdrawal but also taking occasional additional spending from this same portfolio is just wrong. Actually, it's not only wrong for VPW, but also for other common withdrawal approaches (constant-dollar withdrawal, fixed-percentage withdrawal, and 1/N withdrawal).

As even the author of a new retirement planning tool got into that trap, I thought that there is a possibility that other readers of this thread might fall into it.

To prevent this incorrect use of withdrawal methods, I am trying to explain how to do things correctly: one should divide the total portfolio in dedicated parts, one part for VPW, and one part for each other type of fixed expense.


It is just plain wrong to extract additional money from VPW's portfolio than the amount dictated by the withdrawal percentage. Such an extraction will simply lower all future withdrawals below what they should have been.

Similarly, it is contrary to VPW's philosophy not to bridge the gap until social security and pensions kick in. Such an omission would impose lean early retirement years and delay spending to older years.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

Post Reply