Test of constant % spending model

[Lsbcal]

Test of constant % spending model + reserve fund

I wanted to test out how a constant percentage spending model would work under a worst case historical sequence. I used VPW to show the historical simulation results and to suggest a way to deploy a short term reserve account. I'm showing some results here because I'd like to hear if someone sees flaws in the approach. This is pretty detailed and the example is commented to help out a bit. Sorry about this but please ask questions if I can help to clear up confusion.

What I did was to use the VPW withdrawal algorithm to compute allowed spending each year. VPW by itself results in some very high withdrawals with the possibility one would have to cut spending in some later years until the portfolio recovers. It does have the advantage of never entirely running out of money over the sequence (up to the Depletion Years number). Since we personally do not need so much to spend I assumed we would put the excess into a short term reserve fund that just keeps up with inflation. This reserve then fills in the needs further into a very bad sequence and would be available should the bad sequence not happen at all. This reserve fund approach is used by some of our esteemed forum members and I like the idea.

Below is an example which shows how the short term reserve grows and then shrinks. The blue columns are my additions to show the way the reserve fund works. This is for a fictitious couple who are 65 and draw social security ($34000) with a 60/40 portfolio and a $1M starting value. The portfolio must last to age 110 i.e. there will be some money left over for heirs and no danger of serious worrisome depletion. The worst case postwar sequence starting in 1966 was used. The couple was assumed to spend $65000 inflation adjusted each year. This works out to a 3.1% withdrawal rate from the portfolio which happens to be close to our personal withdrawal rate for the last 3 years.

The detailed results are shown below. I've removed charts from the VPW backtest sheet and added some columns to show the reserve fund and the total portfolio value each year. In this simulation you can see the years when the reserve fund was needed (red negative values in the "unspent, goes to reserves" column). The worst case inflation corrected portfolio value occurred 17 years into the simulation at 44% of the start value. By age 90 the portfolio recovers to 55% of the start value and by age 100 the portfolio recovers to 84% of the start value.

Clearly there is some room to dynamically adjust spending if the bad sequence is not showing up or if the sequence is even worse then 1966. The sequence starting in 1964 has a much more comforting outcome. The 16 years since the year 2000 also is much better.

The fact that a 3.1% withdrawal rate works in a 60/40 portfolio will probably surprise nobody here. But I like the fact that I can see how the sequence unfolds and adjust some input parameters at will.

 

[rodi]

Interesting. But the title suggests constant percentage (vs variable percentage).

So this is VPW with a slush fund? (Unlike AudreyH's constant percentage with a slush fund.)

 

[rodi]

Nevermind - the blue shows the constant percentage...

Question - (the spreadsheet is making my eyes hurt) - is this a percentage of starting portfolio, adjusted for inflation (ala Trinity Study)? Or is this percentage of remaining portfolio (like what Audrey does.)

 

[Lsbcal]

Nevermind - the blue shows the constant percentage...

Question - (the spreadsheet is making my eyes hurt) - is this a percentage of starting portfolio, adjusted for inflation (ala Trinity Study)? Or is this percentage of remaining portfolio (like what Audrey does.)

Sorry about the size of the spreadsheet. It looks OK on my wide screen setup.

This would be the percentage of the (inflation adjusted) remaining portfolio. I think it is basically what Audrey does. VPW is used to suggest the short term reserve fund size i.e. what is left over after the 3.1% is spent in the example. Of course, if the 3.1% is too small to give the same spending level ($65000 in the example) then the reserve fund makes up the rest. In such a year the spending is a higher percentage of the portfolio (that does not include the reserve fund) as shown by the second column in the blue area labled "% of portfolio spent".

I am unsure if I'm using the most accepted terminology here. Perhaps this should be called a constant inflation corrected dollar spending model? For our spending I'd like to make sure that the ride is fairly smooth as DW would not like surprises as in "sorry dear, we have to stay home this year".

 

[big-papa]

Sorry about the size of the spreadsheet. It looks OK on my wide screen setup.

This would be the percentage of the (inflation adjusted) remaining portfolio. I think it is basically what Audrey does. VPW is used to suggest the short term reserve fund size i.e. what is left over after the 3.1% is spent in the example. Of course, if the 3.1% is too small to give the same spending level ($65000 in the example) then the reserve fund makes up the rest. In such a year the spending is a higher percentage of the portfolio (that does not include the reserve fund) as shown by the second column in the blue area labled "% of portfolio spent".

I am unsure if I'm using the most accepted terminology here. Perhaps this should be called a constant inflation corrected dollar spending model? For our spending I'd like to make sure that the ride is fairly smooth as DW would not like surprises as in "sorry dear, we have to stay home this year".

Can't you get something similar by using a pure VPW but with a more conservative asset allocation? Withdrawals will be lower but smoother as well.

 

[Lsbcal]

Can't you get something similar by using a pure VPW but with a more conservative asset allocation? Withdrawals will be lower but smoother as well.

I think for this period 1966 through 1980, there is not much difference in the results for a 60/40 or a 40/60 portfolio. Both will fall short of spending needs in the 1970's because too much was spent at the start of the simulation (no reserve fund). In the simulation at age 80 and year 1981, the 60/40 portfolio is $468k and the 40/60 is $480k.

If one reduces the Depletion Age to maybe 90 then pure VPW probably covers the spending in this example. Personally I figure on leaving an estate because skating on very thin ice is not my idea of a good ride. Hence the selection of Depletion Age = 110.

 

[audreyh1]

I haven't tried to model my unspent funds as my spending is variable too, and I don't know how much will be left over any given year. I'm hoping most of my unspent funds go into the "can spend any time we want to" ever increasing slush fund.

Although I have done some very rough modeling to help me gage how much of a "buffer" might need to be set aside to help during a long bad spell, and how I might have to manage it and my spending. I modeled the size I might need for this buffer (based on our current "max" budget versus current portfolio size) so that I had an idea of how much of our current short term reserves could be spent right away versus be saved for a rainy day. I have actually set these funds aside in 3% CDs not to be tapped until after our tax income drops below 80% of our current "max" budget.

We already have sizable short-term reserves, so I didn't need to model building one. In fact when we retired we had already set some short-term reserves aside.

The "max" budget is the the most I really expect us to spend any given year, but it doesn't include one time large items like buying a new car, or some splurge because we have the extra funds. This budget has a high % of discretionary expenses. It was also recently increased (i.e, gave ourselves a raise) since we finally exceeded the old budget last year. Our spending is quite variable, but I use this generous "max" budget for planning purposes.

Another thing worth mentioning is that where our portfolio is today (knock on wood) our current withdrawal rate is exceeding our max after-tax spending budget by 20-25%. So right at day one of withdrawal we have excess funds we can set aside, and potentially more unspent by the end of the year, unless we use it for one-off items.

This indicates we could tolerate a pretty big hit to our portfolio and income before we actually see our "max" budget impacted and need to draw on short term funds. And our budget being heavily padded, we could could easily cut some of the discretionary spending before we need to draw on short term reserves to supplement our income, if we choose to go that route.

It will take me a while to digest your spreadsheet. I think there are things I don't understand about VPW.

One of the ways I compared schemes was looking worst case income and min, average and max terminal portfolio values. I found that the worst case years were actually like 1906 and several other years pre 1920. 1966 wasn't quite as bad - at least not for % remaining portfolio.

3.1% is a fairly low rate. My models with % remaining portfolio indicated I could go above 4.25% and still do fine as long as my expense ratios were really low - as your rate increases, the income drop you might have to deal with widens, but you're starting out at higher income from the get go. We're at 3.5% now but probably will increase as we get older.

 

[Lsbcal]

...
One of the ways I compared schemes was looking worst case income and min, average and max terminal portfolio values. I found that the worst case years were actually like 1906 and several other years pre 1920. 1966 wasn't quite as bad - at least not for % remaining portfolio.

I tried 1906 in the example and it showed similar results as the 1966 simulation. In year 16 of the simulation (1921) the total portfolio (including reserves) was 45% of the start.

3.1% is a fairly low rate. My models with % remaining portfolio indicated I could go above 4.25% and still do fine as long as my expense ratios were really low - as your rate increases, the income drop you might have to deal with widens, but you're starting out at higher income from the get go. We're at 3.5% now but probably will increase as we get older.

Our withdrawal rate over the last 3 years has been 3.2% and is plenty for several vacations and discretionary house spending. Keep in mind that this simulation includes SS which helps to reduce portfolio spending.

 

[big-papa]

I think for this period 1966 through 1980, there is not much difference in the results for a 60/40 or a 40/60 portfolio. Both will fall short of spending needs in the 1970's because too much was spent at the start of the simulation (no reserve fund). In the simulation at age 80 and year 1981, the 60/40 portfolio is $468k and the 40/60 is $480k.

If one reduces the Depletion Age to maybe 90 then pure VPW probably covers the spending in this example. Personally I figure on leaving an estate because skating on very thin ice is not my idea of a good ride. Hence the selection of Depletion Age = 110.

A lot of us add extra years to VPW for the same reason. I think one reason that sequence withdraws a lot early on during that sequence is because the expected returns used for VPW are high. It's a variable you get to choose and if it's too low, VPW will adjust with larger withdrawals later. Vice versa if you choose too high. It also doesn't have to be constant. You can do some interesting things for the first few years, for example, before returning to a standard VPW annual percentage calculation.

There was a lot of discussions over at bogleheads on smoothing methods. I've mentioned a couple on the other thread you're subscribed to. I'll add yours to my list of things to try out. I have my own VPW spreadsheet and use the PMT function in Excel so I can more easily experiment with these sort of ideas as well as different AA's with various tilts, etc. using data copied and pasted from Simba's spreadsheet.

 

[Lsbcal]

If anyone wants the spreadsheet example in Excel, I'll mail it to them. Just PM me with an email address.

 

[Lsbcal]

A lot of us add extra years to VPW for the same reason. I think one reason that sequence withdraws a lot early on during that sequence is because the expected returns used for VPW are high. It's a variable you get to choose and if it's too low, VPW will adjust with larger withdrawals later. Vice versa if you choose too high. It also doesn't have to be constant. You can do some interesting things for the first few years, for example, before returning to a standard VPW annual percentage calculation.
...

I think VPW sets the suggested withdrawal rates based on the historical real stock and bond growth rates. If not overridden these rates are 5% for stocks and 1.8% for bonds. Of course, the simulations use the actual stock and bond performances for each year.

I've not fiddled with overriding those stock/bond real historical returns. But I just tried the suggested override numbers of 7% for stocks and 0.5% for bonds. The results with a short term reserve fund were not much changed although this upped the first year pure VPW withdrawal to 4.9% from 4.4% in this example.

 

[big-papa]

I think VPW sets the suggested withdrawal rates based on the historical real stock and bond growth rates. If not overridden these rates are 5% for stocks and 1.8% for bonds. Of course, the simulations use the actual stock and bond performances for each year.

I've not fiddled with overriding those stock/bond real historical returns. But I just tried the suggested override numbers of 7% for stocks and 0.5% for bonds. The results with a short term reserve fund were not much changed although this upped the first year pure VPW withdrawal to 4.9% from 4.4% in this example.

It does. There was a question of validity regarding using the time frame for a backtest as part of the long term calculations for stock returns since those returns hadn't yet happened. But with enough history it won't move the needle much.

The addition of smoothing to VPW does help though.
A few people suggested recalculating the long term returns every year based on the CAPE at the end of the previous year. Somewhere between 0.8/CAPE and 0.9/CAPE seemed to give the best results. I've tried it and it helps. This is forward looking returns instead of backwards looking historical returns.

I tried short term smoothing by using a modified Clatchett algorithm: take the larger of the VPW withdrawal or 95% of the previous year's withdrawal. I found that 95% of the previous year's inflation adjusted withdrawal works best. You can use higher than 95% by the way. The tradeoff is between smoothing the withdrawals and how long it takes to come out of the hole especially for the late 60's starting date.

Lots of options that can be explored and put in Excel.

 

[Lsbcal]

A few have asked for the spreadsheet. If anyone has any comments either problems with my math or a better way to do this, I'd like to hear about it here. Thanks.

 

[Chuckanut]

A few have asked for the spreadsheet. If anyone has any comments either problems with my math or a better way to do this, I'd like to hear about it here. Thanks.

Thanks for the SS info. The concept of how to manage one's withdrawal rate fascinates me.

 

[Lsbcal]

Sorry, I made a mistake in the OP. This does not really change the basic outcome of the simulation but it clarifies the method. This should have been called a constant inflation adjusted spending model, I think. I would change the title but can not apparently do so.

Anyway, the "percent of portfolio spent" column has been corrected below. You can see that the spending reaches as high as 7.0% of the total portfolio. It then goes down from there. By age 100 (not shown) the portfolio is 84% of the start portfolio and spending is at a 3.7% rate.

As you see the total portfolio value (far right) and the short term reserves column is not changed.

.

 

[audreyh1]

Well - I think it has been generally understood that constant income/spending implies inflation adjusted because that is what the Trinity study and SAFEMAX papers model. And FIRECALC calls it constant spending.

On VPW - I think the name, Variable Percentage Withdrawal, is a little confusing. Because it's really Increasing Percentage Withdrawal, when the percentage increases with age kind of like the RMD. There is no case in which the percentage withdrawn decreases.

I guess I always imagined Variable meant the percentage went up and down.

 

[big-papa]

Well - I think it has been generally understood that constant income/spending implies inflation adjusted because that is what the Trinity study and SAFEMAX papers model. And FIRECALC calls it constant spending.

On VPW - I think the name, Variable Percentage Withdrawal, is a little confusing. Because it's really Increasing Percentage Withdrawal, when the percentage increases with age kind of like the RMD. There is not case in which the percentage withdrawn decreases.

I guess I always imagined Variable meant the percentage went up and down.

Correct - VPW increases the % withdrawn each year. It does not do so based on inflation but, rather, the expected return of the portfolio which is an input parameter. As an aside, if you set that value to 0%, it then becomes identical to a 1/N method where N=number of years remaining, which is subset of an RMD method that assumes you know exactly the year you're going to croak.

 

[Lsbcal]

I sure am glad I don't write papers for a living.

Anyway, this exercise really helped me because it shows that having a short term reserves squirreled away in a low risk account will probably work in even the worst forward sequences.

 

[audreyh1]

I sure am glad I don't write papers for a living.

Anyway, this exercise really helped me because it shows that having a short term reserves squirreled away in a low risk account will probably work in even the worst forward sequences.

Yep - that's why I hoard unspent funds instead of reinvesting them.

 

[Lsbcal]

Yep - that's why I hoard unspent funds instead of reinvesting them.

I knew you were right in past discussions but I wanted a model just for future what if's and the comfort factor.

 

[audreyh1]

I knew you were right in past discussions but I wanted a model just for future what if's and the comfort factor.

Well a lot of it has been instinctive for me (especially my conviction to not reinvest unspent funds), but I also had to go do the analysis to see how my "system" might be stressed in extreme cases, and evaluate what amount really should be set aside just in case income dropped to quite low levels over a multi-year period.