Guyton-Klinger Withdrawal Decison Rules

Non-investing personal finance issues including insurance, credit, real estate, taxes, employment and legal issues such as trusts and wills
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

While tinkering with the cFIREsim retirement income calculator, I came across the Guyton-Klinger decision rules for retirement withdrawals. I have gotten several of the original articles and the rules seem quite sensible and useful. The authors provide various backtested results, with various asset allocations and probabilities of failures. A key period they have back tested is a 40 year retirement beginning in 1973, when returns were terrible for a number of years while inflation was very high. Their rules succeeded in that environment. In Monte Carlo simulations, they choose probabilities of > 90% or >95% success.

The advantage of their rules is that they permit a more aggressive rate of initial withdrawals that is also safe because of the way the rules limit vulnerabilities to the damage done by poor investment returns. They permit withdrawals of significantly higher than 4% initially. The rules seem to require multi-asset classes, but nothing even slightly exotic or unusual.

These rules are a bit more complicated than most, but really no more difficult to implement than most strategies are. I would say the rules are similar to bucket strategies in terms of difficulty of implementation.

Does anyone have any particular knowledge of these rules, an informed critique of them, or an informed endorsement of them?

Thanks for any help you can offer.
Leeraar
Posts: 4109
Joined: Tue Dec 10, 2013 8:41 pm
Location: Nowhere

Re: Guyton-Klinger Withdrawal Decison Rules

Post by Leeraar »

You might ask this question in one of the threads on VPW, Variable Percentage Withdrawal. There is also a Wiki entry on VPW.

L.
You can get what you want, or you can just get old. (Billy Joel, "Vienna")
kolea
Posts: 1322
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Guyton-Klinger Withdrawal Decison Rules

Post by kolea »

Their scheme is basically a constant percentage with a floor and ceiling based on the initial withdrawal and whether the returns for the year were positive. They back tested it with a Monte Carlo simulation which used historical returns. I think the latter is a point of concern since bonds and dividends aren't what they used to be. But I like the model - I use a fixed percentage with floor and ceiling also, but I calculate those differently than this study does.
Kolea (pron. ko-lay-uh). Golden plover.
Beliavsky
Posts: 973
Joined: Sun Jun 29, 2014 10:21 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by Beliavsky »

TwoByFour wrote:Their scheme is basically a constant percentage with a floor and ceiling based on the initial withdrawal and whether the returns for the year were positive. They back tested it with a Monte Carlo simulation which used historical returns. I think the latter is a point of concern since bonds and dividends aren't what they used to be. But I like the model - I use a fixed percentage with floor and ceiling also, but I calculate those differently than this study does.
Sounds like what I suggested in the thread 4% rule with % of current value ceiling, except for the annual return conditionality.

In that thread someone pointed out a similar 2013 Vanguard study "A more dynamic approach to spending for investors in retirement".

Googling "Guyton Klinger" gives relevant links. There was a paper

Using Decision Rules to Create Retirement Withdrawal Profiles
by William J. Klinger
FPA Journal
August 2007
BahamaMan
Posts: 896
Joined: Wed Oct 01, 2014 5:52 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by BahamaMan »

Another Vote for VPW. Much Simpler and Safer!

http://www.bogleheads.org/wiki/Variable ... withdrawal
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

BahamaMan wrote:Another Vote for VPW. Much Simpler and Safer!

http://www.bogleheads.org/wiki/Variable ... withdrawal
Simpler, yes for sure. Although I don't think G-K is overly complex to implement

What is the basis for suggesting it's safer?
Dave
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

TwoByFour wrote:Their scheme is basically a constant percentage with a floor and ceiling based on the initial withdrawal and whether the returns for the year were positive. They back tested it with a Monte Carlo simulation which used historical returns. I think the latter is a point of concern since bonds and dividends aren't what they used to be. But I like the model - I use a fixed percentage with floor and ceiling also, but I calculate those differently than this study does.
G-K is really somewhat more complex than this. The basic trajectory is to allow portfolio withdrawals to increase annually to keep pace with inflation, but they impose some limitations on withdrawals to compensate for both poor portfolio performance and better than expected portfolio performance. They apply a capital preservation rule that limits how high the withdrawal rate can get by imposing withdrawal limitations under conditions of poor portfolio performance. Conversely, they allow a prosperity rule that allows absolute withdrawals to increase above the inflation rate in cases where good portfolio performance has the affect of lowering the withdrawal rate below a threshold. The floor and ceiling parameters function as additional guardrails that limit how much of an excursion in absolute withdrawals the capital preservation rule and prosperity rules will allow.

I think it's hard to make broad generalizations about the G-K approach because it can be so many different things depending on how you set the parameters governing the capital preservation and prosperity rules, as well as where you set the upper and lower guardrails.
Dave
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

DFrank wrote:
G-K is really somewhat more complex than this. The basic trajectory is to allow portfolio withdrawals to increase annually to keep pace with inflation, but they impose some limitations on withdrawals to compensate for both poor portfolio performance and better than expected portfolio performance. They apply a capital preservation rule that limits how high the withdrawal rate can get by imposing withdrawal limitations under conditions of poor portfolio performance. Conversely, they allow a prosperity rule that allows absolute withdrawals to increase above the inflation rate in cases where good portfolio performance has the affect of lowering the withdrawal rate below a threshold. The floor and ceiling parameters function as additional guardrails that limit how much of an excursion in absolute withdrawals the capital preservation rule and prosperity rules will allow.

I think it's hard to make broad generalizations about the G-K approach because it can be so many different things depending on how you set the parameters governing the capital preservation and prosperity rules, as well as where you set the upper and lower guardrails.
There are a couple of other things. They use multi-asset classes with decision rules about where to take and put funds depending on performance of the overall portfolio and the different asset classes. They also provide useful guidelines on where to place the guardrails and the probable consequences of these choices. They emphasize safety, setting their real recommendations at the 95% success level. They also speak to the issues of the allocation to equities, generally finding that relatively higher levels of equity holdings are acceptable and enhance success in their model.

I have to run right now (in NYC, where we just had our coldest February in 85 years and the snow hasn't been off the ground for more than a month!). But I will try to post a synopsis of their rules this evening.
Last edited by duffer on Mon Jan 30, 2017 8:47 am, edited 1 time in total.
BahamaMan
Posts: 896
Joined: Wed Oct 01, 2014 5:52 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by BahamaMan »

DFrank wrote:What is the basis for suggesting it's safer?
VPW is 100% Safe. When you start discussing 90-95% Success rates, that is Less than 100% Safe. But as previous posters indicate, it's hard to 'pin down' GK, as you can 'fiddle' with a lot of parameters as well as asset allocation.
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

So, listed below are Guyton-Klinger's withdrawal decision rules, as summarized in an excellent November, 2013 piece by Wade Pfau.

The link to his very useful article is: http://www.advisorperspectives.com/news ... ement2.php

Pfau likes these rules with a specific caveat: because the rules will reduce withdrawals in some years of poor market performance, retirees using this method must be able to live with years in which their income is reduced. So, the rules may be best for people who have their essential expenses covered by stable sources or who are otherwise able to vary their budgets downward without undue distress.

Inasmuch as my wife and I are fortunate in having two social securities and a joint life pension that will cover our essential expenses, I think these rules are likely to work well for us. Others might combine these rules with annuities to cover the margin between average withdrawals and the withdrawals in the occasional years of reduced withdrawals.

The decision rules are designed to work with multiple asset classes (of very ordinary type) so that withdrawals come from different sources in different years depending on market performance.

Pfau's summary of the rules is as follows:
Guyton and Klinger's decision rules

Jonathan Guyton’s decision rules are quite popular with advisors, and the most modern form of these rules is outlined in a March 2006 article from the JFP written with William Klinger. Guyton introduced his approach in an earlier 2004 article, asking in its subtitle, “Is the ‘Safe’ Initial Withdrawal Rate Too Safe?” The fundamental conclusion of his analysis was that clients who are willing to make provisions for spending cuts, should the need arise, can confidently start with a higher withdrawal rate than deemed appropriate when using a constant inflation-adjusted strategy. Guyton’s rules allow clients to begin retirement with a higher withdrawal rate, understanding that future spending may not always increase with inflation and may need to be cut in certain circumstances.

Guyton identified four decision rules that advisors can apply to client portfolios:

The portfolio management rule : The focus of this rule is on the asset classes from which withdrawals are taken and how the portfolio is rebalanced. The idea is to take withdrawals from the assets that had the greatest growth in the previous year and to move excess portfolio gains (beyond what is needed for the withdrawal) into a cash account dedicated to future withdrawals. For example, equities are not sold in years following a negative return as long as there are sufficient assets from other sources to meet withdrawal needs.
The withdrawal rule: Each year’s withdrawal amount increases by the previous year’s inflation rate, unless the previous year’s portfolio total return was negative and caused the current withdrawal rate to exceed the initial withdrawal rate. In those cases, the withdrawal amount is frozen with no inflation adjustment. There are no make-ups for missed withdrawal increases.
The capital preservation rule: If the current withdrawal rate has risen by more than 20% above the initial withdrawal rate (e.g., if the initial withdrawal rate was 5%, then the threshold is 6%), then current year spending is reduced by 10%. The rule is no longer applied within 15 years of the maximum planning age.
The prosperity rule: Conversely, if the current withdrawal rate has fallen by more than 20% below the initial withdrawal rate (e.g., if the initial withdrawal rate was 5%, then the threshold is 4%), current year spending is increased by 10%. Otherwise, retirees might miss out on being able to increase spending in a sustainable manner when markets are doing well.
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

BahamaMan wrote:
DFrank wrote:What is the basis for suggesting it's safer?
VPW is 100% Safe. When you start discussing 90-95% Success rates, that is Less than 100% Safe. But as previous posters indicate. It's hard to 'pin down' GK, as you can 'fiddle' with a lot of parameters as well as asset allocation.
Not to put too fine a point in it, but I don't think any withdrawal strategy is 100% safe. Over a 30+ year retirement period there is always the probability of a black swan event. That said, perhaps I don't fully understand it, but it seems as though the high success rate of VPW is achieved at the expense of accepting potentially large variations in spending over time.

What I like about G-K is that it establishes some boundaries on spending. There is nothing about G-K that inherently limits success rates to 90-95%. If you set up a G-K withdrawal strategy with aggressive capital preservation rule parameters and a very low lower guardrail parameter (in the extreme this could even be $0) I think you would see very high success rates. I set my lower guardrail equal to our expected non-discretionary retirement budget, and in my personal scenarios I obtain 100% success rates for G-K using cFiresim's backtesting tool. I also use a high ER as a means of discounting past market performance in the backtesting scenarios.
Last edited by DFrank on Wed Mar 04, 2015 12:29 pm, edited 1 time in total.
Dave
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

I've tinkered with Guyton-Klinger (and multiple other withdrawal strategies) quite a lot in my own Excel spreadsheets. I don't believe much in Monte-Carlo, my approach is based on solid backtesting, while playing with a drag on returns to simulate possible rosier or drearier average returns.

The way it is explained in the articles does make it look a tad complicated, but it is actually downright simple. You pick a reasonable expected rate of return for your portfolio (which could be any AA, really), in real terms. Using the VPW default rate could be a good choice for that. Then you allow the actual rate of return to change within reason (say up to +/- 20%) - they call it "guardrails" - while using a regular CPI adjustment on your annual withdrawal. And it's only if you get out of bounds (e.g. due to your portfolio fast increasing or fast decreasing in presence of severe market vagaries) that you adjust your actual withdrawal, say by adding/subtracting 10% from the previous year (in real terms).

To take an example, say you pick 5% as your expected rate of return. Then your withdrawals will stay stable in real terms if the actual rate stays between 4% and 6%. If not, then a 10% adjustment will be in order. Simple as that. Yes, the article suggests a little more complexity, with the MWR rule, but I didn't find it useful. I believe Guyton gave up on it in recent articles. Note that the concepts of floor and ceiling were added by cFIREsim and are not part of the original G-K rules.

Pros:
- very simple once you understand the mindset behind it - just takes a couple of simple Excel formulas to implement
- quite safe if you're minimally sensible about the expected rate of return (long-term adjustments work very well in backtesting)
- short-term volatility (year over year) is excellent; does NOT overreact due to the vagaries of the market
- on average, this tends to be a perpetual scheme, will last forever (hence preserving a solid bequest for heirs and/or charities)
- allows to withdraw significantly more money on average than any SWR scheme
- works with any AA

Cons:
- If the actual average rate of return on the long-term goes out of the guardrails, then things start to drift. So an extreme situation (say returns become 3% on average and you selected 5% +/- 20%) will not go very well, and may fail (in the sense of the portfolio going down to zero). This is why I said it is quite safe, but not 100% safe.
- Not a great idea to make the guardrails very wide (i.e. more than 20% or 25%), as this reduces the adaptability of the algorithm. Not a great idea either to pick an overly conservative expected rate of return (will lose on average income). So you need to make a compromise. The default parameters work quite well...
- This is a variable scheme, so you need a dual-budget, and stomach for some level of volatility, notably decade over decade. Still, this is very contained, and seems very realistic.

Overall, I think this is one of the best approaches around, notably if you'd like low short-term volatility, improved average income, and a perpetual scheme. Still, the possible drift if average future returns get way out of bounds is a tad troublesome. We really can't say that VPW is better than G-K, or the reverse way around. Both of them tend to correspond to a rather different set of goals.
Last edited by siamond on Wed Mar 04, 2015 12:25 pm, edited 2 times in total.
Beliavsky
Posts: 973
Joined: Sun Jun 29, 2014 10:21 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by Beliavsky »

BahamaMan wrote:
DFrank wrote:What is the basis for suggesting it's safer?
VPW is 100% Safe.
How do you define safe? Even spending 50% of your money each year is "safe" in the sense that it will always be mathematically possible. If you are down to $1000 in savings you can spend $500. Spending plans should be judged by the expected utility of consumption they provide.
Beliavsky
Posts: 973
Joined: Sun Jun 29, 2014 10:21 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by Beliavsky »

siamond wrote:I've tinkered with Guyton-Klinger (and multiple other withdrawal strategies) quite a lot in my own Excel spreadsheets. I don't believe in Monte-Carlo, my approach is based on solid backtesting, while playing with a drag on returns to simulate possible rosier or drearier average returns.
That will give similar results to using Monte Carlo with historical means and covariances estimated from historical data. You can adjust the mean downwards if desired. I don't see why you would believe in backtesting and not Monte Carlo.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

Beliavsky wrote:
siamond wrote:I've tinkered with Guyton-Klinger (and multiple other withdrawal strategies) quite a lot in my own Excel spreadsheets. I don't believe in Monte-Carlo, my approach is based on solid backtesting, while playing with a drag on returns to simulate possible rosier or drearier average returns.
That will give similar results to using Monte Carlo with historical means and covariances estimated from historical data. You can adjust the mean downwards if desired. I don't see why you would believe in backtesting and not Monte Carlo.
Sorry, my statement was maybe too black & white. I just don't believe that the real world of investing can be summarized by a neat simple equation. There are loose correlations all over the place, patterns that develop, then disappear, etc. I am much more comfortable running a scenario equal or close to something that actually happened than running a very theoretical model. But I acknowledge this is just an opinion, and I can see that people would prefer otherwise. And you're right, I suspect the assessment of G-K using both methods would provide very similar results.
Beliavsky
Posts: 973
Joined: Sun Jun 29, 2014 10:21 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by Beliavsky »

duffer wrote: The withdrawal rule: Each year’s withdrawal amount increases by the previous year’s inflation rate, unless the previous year’s portfolio total return was negative and caused the current withdrawal rate to exceed the initial withdrawal rate. In those cases, the withdrawal amount is frozen with no inflation adjustment. There are no make-ups for missed withdrawal increases.
This is silly. If inflation is 6%, you let real spending fall by 6% if returns are poor, but if inflation is 0% you don't let real spending fall. It's also silly because it is discontinuous for portfolio returns near 0%. Why should your spending be substantially different this year if your return last year was -0.1% vs. 0.1%?
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

Beliavsky wrote:
duffer wrote: The withdrawal rule: Each year’s withdrawal amount increases by the previous year’s inflation rate, unless the previous year’s portfolio total return was negative and caused the current withdrawal rate to exceed the initial withdrawal rate. In those cases, the withdrawal amount is frozen with no inflation adjustment. There are no make-ups for missed withdrawal increases.
This is silly. If inflation is 6%, you let real spending fall by 6% if returns are poor, but if inflation is 0% you don't let real spending fall. It's also silly because it is discontinuous for portfolio returns near 0%. Why should your spending be substantially different this year if your return last year was -0.1% vs. 0.1%?
Yes, this is the MWR rule. I agree, seems rather silly. And really does NOT make any meaningful difference in my backtesting. cFIREsim didn't implement it, by the way. Nor I think anybody should. Stick to the CPR/PR rules, this is the good part.
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

Beliavsky wrote:
duffer wrote: The withdrawal rule: Each year’s withdrawal amount increases by the previous year’s inflation rate, unless the previous year’s portfolio total return was negative and caused the current withdrawal rate to exceed the initial withdrawal rate. In those cases, the withdrawal amount is frozen with no inflation adjustment. There are no make-ups for missed withdrawal increases.
This is silly. If inflation is 6%, you let real spending fall by 6% if returns are poor, but if inflation is 0% you don't let real spending fall. It's also silly because it is discontinuous for portfolio returns near 0%. Why should your spending be substantially different this year if your return last year was -0.1% vs. 0.1%?
duffer was quoting from a Wade Pfau article, and that quote does not accurately describe the G-K rules. I'd refer you back to the original paper that you linked earlier.
Dave
User avatar
ResearchMed
Posts: 10767
Joined: Fri Dec 26, 2008 11:25 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by ResearchMed »

siamond wrote:I've tinkered with Guyton-Klinger (and multiple other withdrawal strategies) quite a lot in my own Excel spreadsheets. I don't believe much in Monte-Carlo, my approach is based on solid backtesting, while playing with a drag on returns to simulate possible rosier or drearier average returns.

The way it is explained in the articles does make it look a tad complicated, but it is actually downright simple. You pick a reasonable expected rate of return for your portfolio (which could be any AA, really), in real terms. Using the VPW default rate could be a good choice for that. Then you allow the actual rate of return to change within reason (say up to +/- 20%) - they call it "guardrails" - while using a regular CPI adjustment on your annual withdrawal. And it's only if you get out of bounds (e.g. due to your portfolio fast increasing or fast decreasing in presence of severe market vagaries) that you adjust your actual withdrawal, say by adding/subtracting 10% from the previous year (in real terms).

To take an example, say you pick 5% as your expected rate of return. Then your withdrawals will stay stable in real terms if the actual rate stays between 4% and 6%. If not, then a 10% adjustment will be in order. Simple as that. Yes, the article suggests a little more complexity, with the MWR rule, but I didn't find it useful. I believe Guyton gave up on it in recent articles. Note that the concepts of floor and ceiling were added by cFIREsim and are not part of the original G-K rules.

Pros:
- very simple once you understand the mindset behind it - just takes a couple of simple Excel formulas to implement
- quite safe if you're minimally sensible about the expected rate of return (long-term adjustments work very well in backtesting)
- short-term volatility (year over year) is excellent; does NOT overreact due to the vagaries of the market
- on average, this tends to be a perpetual scheme, will last forever (hence preserving a solid bequest for heirs and/or charities)
- allows to withdraw significantly more money on average than any SWR scheme
- works with any AA

Cons:
- If the actual average rate of return on the long-term goes out of the guardrails, then things start to drift. So an extreme situation (say returns become 3% on average and you selected 5% +/- 20%) will not go very well, and may fail (in the sense of the portfolio going down to zero). This is why I said it is quite safe, but not 100% safe.
- Not a great idea to make the guardrails very wide (i.e. more than 20% or 25%), as this reduces the adaptability of the algorithm. Not a great idea either to pick an overly conservative expected rate of return (will lose on average income). So you need to make a compromise. The default parameters work quite well...
- This is a variable scheme, so you need a dual-budget, and stomach for some level of volatility, notably decade over decade. Still, this is very contained, and seems very realistic.

Overall, I think this is one of the best approaches around, notably if you'd like low short-term volatility, improved average income, and a perpetual scheme. Still, the possible drift if average future returns get way out of bounds is a tad troublesome. We really can't say that VPW is better than G-K, or the reverse way around. Both of them tend to correspond to a rather different set of goals.
Is there a version of Guyton-Klinger that is not perpetual, for those without major legacy desires?

Perhaps using something like the percentages used for life-expectancy and required minimum distributions?
The main problem I worry about with this is that whereas it might work well early on, once one gets into very advanced ages, it might not work, as pointed out by Beliavsky (yes, it's "safe", but not particularly practical at certain points).

RM
This signature is a placebo. You are in the control group.
kolea
Posts: 1322
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Guyton-Klinger Withdrawal Decison Rules

Post by kolea »

A I mentioned above, I like the basic scheme of a constant percent withdrawal that is limited at top and bottom. G-K has one way of calculating the top and bottom limits, and others have other ways. Pick the way you like best. If you read the recent thread on this forum about the Yale endowment fund, you may have noticed that Yale has a variation of the constant-percent with floor and ceiling withdrawal scheme, but with a smoothing function. There are endless variations you can come up with.

BTW - "Capital Preservation Rule" is just a fancy name for a ceiling, and "Prosperity Rule" is just a fancy name for a floor.
Kolea (pron. ko-lay-uh). Golden plover.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

ResearchMed wrote:Is there a version of Guyton-Klinger that is not perpetual, for those without major legacy desires?

Perhaps using something like the percentages used for life-expectancy and required minimum distributions?
The main problem I worry about with this is that whereas it might work well early on, once one gets into very advanced ages, it might not work, as pointed out by Beliavsky (yes, it's "safe", but not particularly practical at certain points).

RM
I am not quite sure to see why it 'might not work' at an advanced age. Nor did I see any such concern in my own backtesting. Personally, when I get old and into (self-insured) LTC years, I would much rather see my portfolio stay in shape instead of seeing it slowly go to zero like VPW would do. But that's just me, again, we all have different goals.

That being said, to answer your question, yes, there are more advanced uses of G-K rules, where you may elect to change the key parameters for a given period of your retirement. I didn't keep a pointer, but I remember reading an article from Guyton doing exactly that. But I don't know, this seems to lose the simple and feel-safe appeal of G-K. If you have such a goal of no-legacy, then maybe using G-K for a decade or two, then switching to VPW is for you. Or straight VPW with some level of smoothing year over year.

(edited: need to research more how this worked - can't remember off the top of my head)
User avatar
ResearchMed
Posts: 10767
Joined: Fri Dec 26, 2008 11:25 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by ResearchMed »

siamond wrote:
ResearchMed wrote:Is there a version of Guyton-Klinger that is not perpetual, for those without major legacy desires?

Perhaps using something like the percentages used for life-expectancy and required minimum distributions?
The main problem I worry about with this is that whereas it might work well early on, once one gets into very advanced ages, it might not work, as pointed out by Beliavsky (yes, it's "safe", but not particularly practical at certain points).

RM
I am not quite sure to see why it 'might not work' at an advanced age. Nor did I see any such concern in my own backtesting. Personally, when I get old and into (self-insured) LTC years, I would much rather see my portfolio stay in shape instead of seeing it slowly go to zero like VPW would do. But that's just me, again, we all have different goals.

That being said, to answer your question, yes, there are more advanced use of G-K rules, where you may elect to change the key parameters for a given period of your retirement. I didn't keep a pointer, but I remember reading an article from Guyton doing exactly that. One idea is that if you increase the expected rate of return beyond what is truly expected, well, you increase your spend and slowly bring your portfolio down (well, at least in average). Sensible guardrails will prevent you to get overboard. But I don't know, this seems to lose the simple appeal of G-K. If you have such a goal of no-legacy, then maybe using G-K for a decade or two, then switching to VPW is for you. Or straight VPW with some level of smoothing year over year.
Thanks.

We'd have Social Security and some of our assets annuitized, so the "going to zero" isn't quite as scary. It would be for those 'extras', and at that point, it's less likely that we'd need a new car or roof, or be traveling much.
All of this depends upon what type of "assisted care" situation we are in or anticipating, obviously.
We'd always want enough extra (from SS & annuity income OR from remaining investments) for any extra "care" that would make us more comfortable as necessary.

There's really nothing other than annuitizing that really deals with this, and we'd never annuitize everything.
So we recognize that we'll no doubt have *some* legacy issues, and we have made provisions for "what is left". And while we are still competent, we can adjust that as we wish, of course.

But that's quite different from those who do have some explicit legacy desires, especially for heirs.
We can imagine that being a difficult "tug" at times, if we faced it.
This also makes it easier to annuitize more rather than less.
We'll see how the interest rates go, and look into a few small deferred annuities for inflation, or even extra care.
At this point, it's all still at the planning stage. We can't touch the main 403b monies for now anyway.

We also plan to "front load" some spending, mostly for travel, "while we can", realizing that it will leave less for later.
Our health problems seem to have suddenly jumped up, unfortunately, leading us to bump up some travel that we had sort of "scheduled" for future years.

As long as we've got the basics covered, we can adjust to everything else.
More travel? Less travel? Fewer "special" meals, whatever...
Clothing costs are already noticeably less, more so than we would have expected.

RM
This signature is a placebo. You are in the control group.
kolea
Posts: 1322
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Guyton-Klinger Withdrawal Decison Rules

Post by kolea »

ResearchMed wrote:
Is there a version of Guyton-Klinger that is not perpetual, for those without major legacy desires?

Perhaps using something like the percentages used for life-expectancy and required minimum distributions?
The main problem I worry about with this is that whereas it might work well early on, once one gets into very advanced ages, it might not work, as pointed out by Beliavsky (yes, it's "safe", but not particularly practical at certain points).

RM
It is very easy to make it finite in scope - change the calculation for the floor. For the floor, I determined what my bare-bones budget is, adjust it for inflation, and that is the floor. Since that is a essentially a constant, you run the risk of exhausting your assets if you hit it enough. But in my back testing I had 98% success rate. My floor came out to be 2.9% (of starting assets).
Kolea (pron. ko-lay-uh). Golden plover.
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

TwoByFour wrote:BTW - "Capital Preservation Rule" is just a fancy name for a ceiling, and "Prosperity Rule" is just a fancy name for a floor.
In G-K these are the rules that govern how much your withdrawal is adjusted based on the performance of your portfolio. So, they are more than just a ceiling/floor.
Dave
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

TwoByFour wrote:BTW - "Capital Preservation Rule" is just a fancy name for a ceiling, and "Prosperity Rule" is just a fancy name for a floor.
TwoByFour wrote:It is very easy to make it finite in scope - change the calculation for the floor. For the floor, I determined what my bare-bones budget is, adjust it for inflation, and that is the floor. Since that is a essentially a constant, you run the risk of exhausting your assets if you hit it enough. But in my back testing I had 98% success rate. My floor came out to be 2.9% (of starting assets).
Not too sure to follow you, at least with the 'classic' definition of the CPR and PR rules. They could be viewed as a floor/ceiling of sorts, but those are *relative* numbers, i.e. a percentage of your current portfolio (i.e a withdrawal rate). While a floor/ceiling is usually defined in absolute terms (well, in real $), like a bare bone budget of $5K a month. Those are two different things, and this is why cFIREsim supports both types of parameters.

This being said, I agree that if one starts ignoring the CPR rule at a given age, the portfolio is more likely to start decreasing on average. Actually, in the first G-K write-ups, I remember the authors suggested to deactivate the CPR rule when you're 85 and beyond. I think the primary idea is that, by then, nothing much you can do will destroy your portfolio anyway, and it becomes more important (psychologically speaking) to get a very stable withdrawal inflow year over year than to preserve a portfolio. And this is all circling around the concept of a floor, I guess. This always seemed a tad 'ad hoc' to me, but I can see the logic.

Still, the most efficient way to deplete the portfolio seems an increase in the primary rate of return (IWR in G-K's terminology). Even if VPW does seem more effective for such goal.
BahamaMan
Posts: 896
Joined: Wed Oct 01, 2014 5:52 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by BahamaMan »

DFrank wrote:Not to put too fine a point in it, but I don't think any withdrawal strategy is 100% safe. Over a 30+ year retirement period there is always the probability of a black swan event. That said, perhaps I don't fully understand it, but it seems as though the high success rate of VPW is achieved at the expense of accepting potentially large variations in spending over time.
VPW is 100% safe, as it would not fail long before the other methods being discussed would fail. But, the 'Large variations' that you speak of, are not as Large as you would think.

Spend some time to understand it and I'm sure that will see the benefits of it. I am actually 'Living it', currently from the Bahamas. So, it is not just a theoretical Withdrawal Method such as the '4%' Rule.
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

BahamaMan wrote:
DFrank wrote:Not to put too fine a point in it, but I don't think any withdrawal strategy is 100% safe. Over a 30+ year retirement period there is always the probability of a black swan event. That said, perhaps I don't fully understand it, but it seems as though the high success rate of VPW is achieved at the expense of accepting potentially large variations in spending over time.
VPW is 100% safe, as it would not fail long before the other methods being discussed would fail. But, the 'Large variations' that you speak of, are not as Large as you would think.

Spend some time to understand it and I'm sure that will see the benefits of it. I am actually 'Living it', currently from the Bahamas. So, it is not just a theoretical Withdrawal Method such as the '4%' Rule.

I've played around with the VPW spreadsheet, and I see minimum withdrawals that are as low as 64% of the initial withdrawal in the backtesting worksheet. That's a pretty large variation IMO.

When I run backtested GK simulations I don't see variations that large. By comparison, I can limit withdrawal variations to 15% using G-K and still have backtested 100% success rate while setting my investment costs to 2% as a way to put about a 1.8% drag on historic returns.
Dave
randomguy
Posts: 9208
Joined: Wed Sep 17, 2014 9:00 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by randomguy »

BahamaMan wrote:
VPW is 100% safe, as it would not fail long before the other methods being discussed would fail.
Nope. The 1966 VPW retiree fails after 9 years of retirement (technically year 5 was also a failure year by a couple hundered dollars) and continues to fail for the next 12 years. The 4% SWR fails after 20 years. Failing by running out of money is one failure case. Failing to generate enough money from the portfolio the second failure case. VPW insulates you against 1 while exposing you to 2. SWR insulates you from 2 but exposes you to 1. Guyton-Klinger and the various other options try to be a hybrid by preventing huge drops in spending ability while trying to also prevent running out of money. They are willing to cut spending in bad years but they are not as aggressive as doing as a VPW approach does.
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

DFrank wrote: I've played around with the VPW spreadsheet, and I see minimum withdrawals that are as low as 64% of the initial withdrawal in the backtesting worksheet. That's a pretty large variation IMO.
Are you saying that the withdrawals started at $48,000 and dropped to $30,720 (e.g. 64% of $48K)? That's barely 23% lower than $40,000 (as per the 4% SWR rule of thumb). And, this assumes that the retiree has 0$ in Social Security and pensions!

If the retiree had $20,000 in social security, we're talking about barely reducing withdrawals by 16% (before taxes) below the equivalent 4% SWR planning amount of $60,000 (e.g. ($30,720 + $20,000) / ($40,000 + $20,000)).

If I didn't have the flexibility in my budget to cut 16% during a market crisis, I wouldn't consider myself ready to retire! Actually, I plan to have much more flexibility than that, and I would recommend that the typical "VPW" retiree to have it too.

The advantage of VPW is that it does let you spend your money when times are good. It only reduces withdrawals when markets tank. Most other withdrawal methods don't let you benefit from a bull market at the start of your retirement and delay your spending until you're old or dead. But, VPW can't predict the future. So, if markets drop, it will lower withdrawals without assuming that markets will quickly recover. VPW would still work in a Japan-like scenario (1989-).

Let me note that VPW is no replacement for asset allocation. If one decides to retire using VPW and a 100% portfolio on the top of a bubble, then VPW will significantly lower withdrawals after the subsequent market crash, instead of letting the retiree deplete the portfolio. It is highly recommended to use a balanced portfolio with VPW (No more than 75%, no less than 25% in stocks). As usual, more bonds usually lead to less volatility. Vanguard Target Retirement funds have retirees in the 50/50 down to 30/70 (stocks/bonds) allocation ranges. VPW backtests very well (specially in nominal terms) within these allocation ranges.
Last edited by longinvest on Wed Mar 04, 2015 5:49 pm, edited 6 times in total.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

randomguy wrote:
BahamaMan wrote:
VPW is 100% safe, as it would not fail long before the other methods being discussed would fail.
Nope. The 1966 VPW retiree fails after 9 years of retirement (technically year 5 was also a failure year by a couple hundered dollars) and continues to fail for the next 12 years. The 4% SWR fails after 20 years. Failing by running out of money is one failure case. Failing to generate enough money from the portfolio the second failure case. VPW insulates you against 1 while exposing you to 2. SWR insulates you from 2 but exposes you to 1. Guyton-Klinger and the various other options try to be a hybrid by preventing huge drops in spending ability while trying to also prevent running out of money. They are willing to cut spending in bad years but they are not as aggressive as doing as a VPW approach does.
Randomguy,

We already had this discussion, and I replied with this:
viewtopic.php?f=10&t=157604&newpost=2364803#p2364847
viewtopic.php?f=10&t=157604&newpost=2364803#p2364928
viewtopic.php?f=10&t=157604&newpost=2364803#p2364944
and
viewtopic.php?f=10&t=157604&newpost=2364803#p2365021

I think that it would be futile to repeat this thread here.

VPW is a very good tool to use within an overall retirement plan that combines many things:
  • Delayed Social Security (to 70)
  • CD (or TIPS) ladder to bridge the gap between retirement and the start of Social Security payments.
  • Pension, if any.
  • Inflation-adjusted single premium immediate annuity, if necessary.
  • VPW for withdrawals from retirement portfolio (with a significant amount of bonds). (Revisit plan every few years).
  • Etc.
Last edited by longinvest on Wed Mar 04, 2015 7:23 pm, edited 1 time in total.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
BahamaMan
Posts: 896
Joined: Wed Oct 01, 2014 5:52 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by BahamaMan »

randomguy wrote:Nope. The 1966 VPW retiree fails after 9 years of retirement (technically year 5 was also a failure year by a couple hundered dollars) and continues to fail for the next 12 years. The 4% SWR fails after 20 years. Failing by running out of money is one failure case. Failing to generate enough money from the portfolio the second failure case. VPW insulates you against 1 while exposing you to 2. SWR insulates you from 2 but exposes you to 1. Guyton-Klinger and the various other options try to be a hybrid by preventing huge drops in spending ability while trying to also prevent running out of money. They are willing to cut spending in bad years but they are not as aggressive as doing as a VPW approach does.
This is so Spectacularly incorrect on so many levels, that you cannot possibly understand VPW.... Study up a bit and then you can start making comments and asking questions.
DFrank
Posts: 474
Joined: Thu Mar 20, 2014 10:44 pm
Location: North Idaho

Re: Guyton-Klinger Withdrawal Decison Rules

Post by DFrank »

longinvest wrote:
DFrank wrote: I've played around with the VPW spreadsheet, and I see minimum withdrawals that are as low as 64% of the initial withdrawal in the backtesting worksheet. That's a pretty large variation IMO.
Are you saying that the withdrawals started at $48,000 and dropped to $30,720 (e.g. 64% of $48K)? That's barely 23% lower than $40,000 (as per the 4% SWR rule of thumb). And, this assumes that the retiree has 0$ in Social Security and pensions!
Yes, that's what I'm pointing out. By comparison, using G-K I only see a 15% drop from the initial withdrawal rate. I'd also note that the G-K scenario represents handicapping the historical returns by about 1.8%. To be fair this probably isn't a completely apples to apples comparison because I used the VPW spreadsheet here for the first number, and CFIREsim for the other (The VPW spreadsheet doesn't allow me to model my scenario which is quite complex).

Don't get me wrong, I'm not suggesting VPW is a horrible approach. This dialog started with a question about a comment that VPW has a 100% success rate, and an implication that it had higher success rate compared to G-K. My observation is that VPW achieves it's high success rate by allowing large variations from the initial withdrawal amount. As it happens I don't think any approach has a 100% success rate (over a 30+ year period black swan events can happen) and it's not obvious to me that the success rate of VPW is better than G-K.
longinvest wrote:If I didn't have the flexibility in my budget to cut 16% during a market crisis, I wouldn't consider myself ready to retire! Actually, I plan to have much more flexibility than that, and I would recommend that the typical "VPW" retiree to have it too.
Agreed. None of us know how our retirements will play out until the end. Until that time, the flexibility, financial and otherwise, to react to unplanned events is critical.
Dave
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

Maybe it would be helpful to visualize what's going on...

Here, I used a starting portfolio of $1.5M, SS income of $30k a year (assuming it follows inflation), an AA of 30/30/40 (US/Int'l/Bonds), a long & happy retirement lasting up to 40 years, and compared G-K with VPW. In both cases, I used the (conservative) VPW expected rate of return (3.9%). G-K is parameterized with 20% guardrails and 10% withdrawal adjustments. And here is the trajectory starting in 1955 (with the oil crisis smack in the middle).

Image
Image

- the portfolio trajectory is obviously very different, VPW (2) aims at depleting the portfolio, G-K (1) aims at preserving it (in this case, with the conservative rate of return, it goes up)
- the spend trajectory, on average, is clearly to the advantage of VPW, and this is just a simple consequence of the previous point. Note that on both cases, it goes up on average, which is due again to the conservative rate of return
- now, as to volatility, we can see how flat and slowly adjusting the G-K trajectory is, while VPW goes up & down much more dynamically (note the 1973/74/75 and the 1985/86/87 data points)

Now, which one do you prefer? There are pros & cons either way. It totally depends on your personal goals...

PS. I actually ran a full series of such cycles starting from 1926 to 1980, and looked at the std deviation. More precisely the std deviation compared to a linear regression. VPW is at 14% while G-K is at 10%. And yet, visually, you don't get such perception, right? I don't like the std-dev metric very much.
PS2. longinvest, if you squint hard enough, the numbers are exactly the same as yours if... you remove this ROUND (x, 3) formula you have in your percentage computation (I do a much more direct PMT math, exact same outcome except for this annoying detail. Took me a little while to track it down!
randomguy
Posts: 9208
Joined: Wed Sep 17, 2014 9:00 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by randomguy »

longinvest wrote:
DFrank wrote: I've played around with the VPW spreadsheet, and I see minimum withdrawals that are as low as 64% of the initial withdrawal in the backtesting worksheet. That's a pretty large variation IMO.
Are you saying that the withdrawals started at $48,000 and dropped to $30,720 (e.g. 64% of $48K)? That's barely 23% lower than $40,000 (as per the 4% SWR rule of thumb). And, this assumes that the retiree has 0$ in Social Security and pensions!

You actually bottom out at 28k.:) SS and pensions don't factor in. You wanted 40k from your portfolio and you get 30k. You have failed. How bad that failure is up to you and how tight you budget is. For the couple living on 15k of SS and 10k from investments, dropping from 25k to 22k might be really painful. You might prefer the VPW to the 4% SWR (i.e. passing up the trips to europe versus living on SS when your 85) failure case but it is still failure.
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

Thanks, Siamond.

Your charts expose very clearly the trade-offs in volatility-of-withdrawals and spending-level of G-K and VPW.

Let me add that, in a worse than worst-historical scenario, G-K still carries a risk of premature depletion. VPW doesn't carry such risk, but it will lower withdrawals more-or-less proportionally to the loss in value of the portfolio; it can seem brutal, but that's still better than dropping withdrawals to zero a few years later (e.g. premature depletion).

It is my personal opinion that the most effective tool, for lowering withdrawal volatility, is the addition of bonds in a portfolio (VPW), instead of using smoothing methods (G-K). Most smoothing methods require markets to recover quickly enough, as they tend to hurt the portfolio in bad times by keeping withdrawals higher than the PMT function would indicate to allow the portfolio to survive. (I know that you worked on using valuation metrics to adapt the PMT's internal growth rate based on market conditions, to help smooth things a little more than plain VPW). Of course, more bonds usually leads to somewhat less returns (we're definitely not talking about a 100% bonds portfolio, here!).

As I already wrote: I wouldn't use a variable withdrawal method alone (e.g. VPW or G-K). I think that it is important to establish some additional income floor in the form of delayed Social Security (and inflation-adjusted SPIA, only if necessary). This floor does not need to be completely sufficient in itself, but it should be high enough so that if withdrawals drop significantly, one doesn't need to start eating Alpo.
Last edited by longinvest on Wed Mar 04, 2015 9:35 pm, edited 1 time in total.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
randomguy
Posts: 9208
Joined: Wed Sep 17, 2014 9:00 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by randomguy »

siamond wrote:Maybe it would be helpful to visualize what's going on...

Here, I used a starting portfolio of $1.5M, SS income of $30k a year (assuming it follows inflation), an AA of 30/30/40 (US/Int'l/Bonds), a long & happy retirement lasting up to 40 years, and compared G-K with VPW. In both cases, I used the (conservative) VPW expected rate of return (3.9%). G-K is parameterized with 20% guardrails and 10% withdrawal adjustments. And here is the trajectory starting in 1955 (with the oil crisis smack in the middle).
Nice charts. I think starting in 66 or 73 might make them more dramatic. Those are pretty much the worst years for any scheme and to some extent the ones we care about.

But looking at the chart I would say that G-K is much too concerned with perserving capital for my tastes. Might have to code it up and see what happens when you put a 30% downside or only 5%/yr adjustments in.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

randomguy wrote:Nice charts. I think starting in 66 or 73 might make them more dramatic. Those are pretty much the worst years for any scheme and to some extent the ones we care about.

But looking at the chart I would say that G-K is much too concerned with preserving capital for my tastes. Might have to code it up and see what happens when you put a 30% downside or only 5%/yr adjustments in.
Well, it depends. To try to compare apple to apple, I plugged the VPW expected rate of return in the previous charts. But remember, it is based on an historical average of world wide returns, and is quite conservative for scenarios with a major US component (well, at least when backtesting). This works ok-ish with VPW as you do end up spending it all (well, if you live long enough!), while it may not be wise to be that conservative with G-K. If I had used 5% in the G-K IWR, you'd see more spend and less portfolio growth. This is really a key point I made on another thread. Those withdrawal methods are really quite sensitive to the IWR/rate-of-return you inject in them to begin with. And that's a tough guess to make. Here is the G-K at 5% vs VPW at 3.9% trajectory.

Image
Image

Something I like about PMT-centric methods is that you can aim at a specific final value (VPW uses $0, I would personally be much less aggressive), hence have a direct handle on how much of the capital you want to preserve vs how much you want to spend. With something like G-K, it is much harder to do something like that, you have to make guesses about rates-of-return and be more or less aggressive, and this is VERY error-prone. Personally, I would primarily use G-K if you have a personal goal of preserving the full portfolio, but it isn't very well suited otherwise.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

longinvest wrote:Let me add that, in a worse than worst-historical scenario, G-K still carries a risk of premature depletion. VPW doesn't carry such risk, but it will lower withdrawals more-or-less proportionally to the loss in value of the portfolio; it can seem brutal, but that's still better than dropping withdrawals to zero a few years later (e.g. premature depletion).
Yes, I would somewhat agree, which is why I stated in a previous post that G-K is 'quite safe', but not perfect in this respect. Although I don't know that we can actually make G-K fail in a realistic scenario (if failure means full portfolio depletion). At its core, this is a % of the current portfolio scheme (which cannot fail). So the main risk is that the 10% adjustments are not fast enough and well, this is a stretch. I just tried using an extremely aggressive drag on returns (-5% on equities, -1.5% on bonds), I ran all cycles from 1926, and I just couldn't make it fail. Sure enough, the withdrawal average became pretty bad and the heirs bequest wouldn't be exactly pretty, but the same would be true with VPW.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

longinvest wrote:It is my personal opinion that the most effective tool, for lowering withdrawal volatility, is the addition of bonds in a portfolio (VPW), instead of using smoothing methods (G-K). Most smoothing methods require markets to recover quickly enough, as they tend to hurt the portfolio in bad times by keeping withdrawals higher than the PMT function would indicate to allow the portfolio to survive. (I know that you worked on using valuation metrics to adapt the PMT's internal growth rate based on market conditions, to help smooth things a little more than plain VPW). Of course, more bonds usually leads to somewhat less returns (we're definitely not talking about a 100% bonds portfolio, here!).
Yes, this is THE interesting question. Ok, so let's test this assumption a little bit. Let's use G-K at 5% with an aggressive 40/40/20 portfolio. Then let's use VPW with a bonds-heavy 15/15/70 portfolio. Looking at the 1955 cycle, this happens to lead to a fairly similar std-deviation over 40 years in both cases (around 13%). Here is the trajectory. Hm, which one do you prefer? Blue is G-K, orange is VPW.

Image
Image

Now, let's look at more extreme cycles. 1966 (uurg!) and 1975 (yahoo!). Seems to me that, if one can sleep at night with the portfolio's swings, G-K and the aggressive AA have something to say for themselves. Ok, granted, a more thorough analysis would be required than just picking a few cycles, but still, room for thoughts.

Image
Image
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

siamond wrote:
longinvest wrote:Let me add that, in a worse than worst-historical scenario, G-K still carries a risk of premature depletion. VPW doesn't carry such risk, but it will lower withdrawals more-or-less proportionally to the loss in value of the portfolio; it can seem brutal, but that's still better than dropping withdrawals to zero a few years later (e.g. premature depletion).
Yes, I would somewhat agree, which is why I stated in a previous post that G-K is 'quite safe', but not perfect in this respect. Although I don't know that we can actually make G-K fail in a realistic scenario (if failure means full portfolio depletion). At its core, this is a % of the current portfolio scheme (which cannot fail). So the main risk is that the 10% adjustments are not fast enough and well, this is a stretch. I just tried using an extremely aggressive drag on returns (-5% on equities, -1.5% on bonds), I ran all cycles from 1926, and I just couldn't make it fail. Sure enough, the withdrawal average became pretty bad and the heirs bequest wouldn't be exactly pretty, but the same would be true with VPW.
While I haven't investigated G-K a lot, I can believe that it would still be pretty resilient as it remains, essentially, a percent of portfolio withdrawal method. All such methods will look pretty much the same, given a similar percentage, if you ignore shorter-term differences.

G-K is a constant-percentage withdrawal derivative method; I'm sure one could build a similar derivative, but based on a variable-percentage method. Actually, I wouldn't be surprised if you had already investigated this.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

Thank you all for these excellent posts--they are very helpful. Does the following seem like a fair summary?

In regard to my original question, the consensus here is that Guyton-Klinger is a very acceptable set of decision rules that will tend to maximize the safe withdrawal rate, even permitting users to select an initial SWR in excess of 4.5 or 5%, and also to preserve their portfolio for heirs over a 40 year timeline. The G-K rules appear able to do this even under very unfavorable historical circumstances, such as those with a retirement beginning in 1973 (a period which was subject to both sequence of returns risk and to very high inflation). G-K offers initial safe withdrawal rates of over 4.5% with portfolio success rates of over 95%.

In order to take advantage of G-K, one must be able to adjust one's living arrangements without undue distress to occasional reductions in the amount that one can withdraw during market downturns. Thus, the G-K decision rules work especially well for individuals who can cover some or most of their essential expenses through sources such as social security, annuities, or pensions. Also, being willing and able to vary one's discretionary expenses in a way that does not cause undue distress is important--for example, by cutting back on travel during a year of reduced withdrawals.

The more one can comfortably tolerate reductions in annual withdrawals during market downturns, the higher one can set one's initial withdrawal rate. So, it gives you immediate access to more of your cash during the most active early years of retirement, as long as you can also be happy during the inevitable years when you can take less.

The G-K rules favor musti-asset classes (in order to provide non-correlated returns that can be selected from for withdrawals) and a fairly heavy allocation to equities (65% or greater is recommended in the G-K articles I have looked at).

For my purposes and situation, these rules appear ideal. We have two social securities and a joint life pension that will cover our essential living expenses. We have strong bequest motivations. We plan to travel and we enjoy both expensive international trips but also less expensive domestic excursions and weeks at the beach. We are very likely to spend more on discretionary expenses in our first 10 or 15 years than we are in later years. All of these factors fit very well with the G-K withdrawal rules and will let us take more as an initial SWR than I otherwise would have dared.

In many ways, the G-K rules seem like a formalized version of what Taylor Larrimore has described he and his wife as doing following his early retirement: taking more when the market was good and less when it was bad in a way that felt safe to them and that permitted them to enjoy life as they went along. This system formalizes what you can safely take each year, without relying so much on one's feelings and intuitive sense.
Last edited by duffer on Thu Mar 05, 2015 8:02 am, edited 1 time in total.
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

siamond wrote:
longinvest wrote:It is my personal opinion that the most effective tool, for lowering withdrawal volatility, is the addition of bonds in a portfolio (VPW), instead of using smoothing methods (G-K). Most smoothing methods require markets to recover quickly enough, as they tend to hurt the portfolio in bad times by keeping withdrawals higher than the PMT function would indicate to allow the portfolio to survive. (I know that you worked on using valuation metrics to adapt the PMT's internal growth rate based on market conditions, to help smooth things a little more than plain VPW). Of course, more bonds usually leads to somewhat less returns (we're definitely not talking about a 100% bonds portfolio, here!).
Yes, this is THE interesting question. Ok, so let's test this assumption a little bit. Let's use G-K at 5% with an aggressive 40/40/20 portfolio. Then let's use VPW with a bonds-heavy 15/15/70 portfolio.
You're really going to compare an 80% stocks portfolio (G-K) to a 30% stocks portfolio (VPW)?! OK, I'll play your game and assume that we have two investors: a risk-seeking investor with bequest motives using 80/20 G-K (no real need for the money for himself) and a risk-averse investor with no bequest motives using 30/70 VPW.

Sigh.
siamond wrote: Looking at the 1955 cycle, this happens to lead to a fairly similar std-deviation over 40 years in both cases (around 13%). Here is the trajectory. Hm, which one do you prefer? Blue is G-K, orange is VPW.

Image
Image
It's not about my or your preference! It's about the preference of the risk-seeking and risk-averse investors.

The risk-seeking investor with bequest motives will obviously prefer the blue path (G-K); it preserves a nice bequest, and the investor shouldn't freak out when losing 50% of his portfolio in two years (1973-1975).

The risk-averse investor with no bequest motives will obviously prefer the red path (VPW); it yields very stable withdrawals (probably even more so in nominal terms) and expose the portfolio to much lower losses in the 1973-1975 period.
siamond wrote: Now, let's look at more extreme cycles. 1966 (uurg!) and 1975 (yahoo!). Seems to me that, if one can sleep at night with the portfolio's swings, G-K and the aggressive AA have something to say for themselves. Ok, granted, a more thorough analysis would be required than just picking a few cycles, but still, room for thoughts.

Image
Image
I see no real room for thought, here. Let me point out that you omitted the portfolio balance charts, which are important in such comparisons! I will assume that they do look as expected: much higher final value and volatility for the 80/20 portfolio (G-K), $0 final value and lower volatility for the 30/70 portfolio (VPW).

The risk-seeking investor with bequest motives will obviously prefer the blue path (G-K); it preserves a nice bequest, and the investor shouldn't freak out when losing 50% of his portfolio in two years (1973-1975).

The risk-averse investor with no bequest motives will obviously prefer the red path (VPW); it yields very stable withdrawals (probably even more so in nominal terms) and expose the portfolio to much lower losses in the 1973-1975 period.

Actually, I am completely amazed that the 30/70 portfolio (VPW) delivered, overall, as much money in withdrawals in 1966 as a 80/20 portfolio (G-K). Aren't you?

Now, it would be interesting to look at the portfolio balance path and withdrawal path in nominal terms. Inflation is very volatile, yet it is mostly felt on longer time periods (therefore the money illusion that fools so many people). In the short term, most people simply adapt their spending patterns.

I don't think that it is a good idea to try lowering inflation-adjusted volatility. Instead, I think that volatility-averse investors should seek to preserve long-term inflation-adjusted buying power and only worry about volatility in nominal terms.

Finally, I would have considered more reasonable a comparison of a 40% stocks (or maybe even 50% stocks) portfolio (G-K) to a 30% stocks portfolio (VPW), with volatility considered in nominal terms. Comparing a 80% stocks portfolio (G-K) to a 30% stocks portfolio (VPW) looks as an unfair attempt to badly portray VPW.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
BahamaMan
Posts: 896
Joined: Wed Oct 01, 2014 5:52 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by BahamaMan »

longinvest wrote:You're really going to compare an 80% stocks portfolio (G-K) to a 30% stocks portfolio (VPW)?!
Agree ! ... This type of Pomegranates to Kiwi Fruit comparison serves little purpose in analysis of Withdrawal methods. You might as well include a graph of VPW completely invested in Pork Bellies.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

longinvest wrote:Actually, I am completely amazed that the 30/70 portfolio (VPW) delivered, overall, as much money in withdrawals in 1966 as a 80/20 portfolio (G-K). Aren't you?
Yes, agreed, there is no discussion that a PMT-based methodology is really good at extracting maximum value (i.e. spend) from a portfolio. Now a 1966 starting point is actually tough for an aggressive portfolio, it's that sequence of returns thing...
longinvest wrote:Now, it would be interesting to look at the portfolio balance path and withdrawal path in nominal terms. Inflation is very volatile, yet it is mostly felt on longer time periods (therefore the money illusion that fools so many people). In the short term, most people simply adapt their spending patterns.

I don't think that it is a good idea to try lowering inflation-adjusted volatility. Instead, I think that volatility-averse investors should seek to preserve long-term inflation-adjusted buying power and only worry about volatility in nominal terms.
Well, I REALLY don't know how to think in nominal terms. Most metrics become completely counter-intuitive and actually pretty much meaningless, charts or aggregate indicators. I don't see why you're thinking that way though. Sure, there is some discomfort in spending significant more (or less) year over year just due to inflation, but at the end this discomfort is nothing compared to your standard of living. And that is something to be assessed in a REAL terms, don't you think?
longinvest wrote:Finally, I would have considered more reasonable a comparison of a 40% stocks (or maybe even 50% stocks) portfolio (G-K) to a 30% stocks portfolio (VPW), with volatility considered in nominal terms. Comparing a 80% stocks portfolio (G-K) to a 30% stocks portfolio (VPW) looks as an unfair attempt to badly portray VPW.
All right, stay zen, we're exploring here... I wasn't trying to badly portray anything. My very first comparison (with the same expected-rate-of-return) was actually unfair to G-K to a large extent... Here, we were pondering about *spending* volatility, so I took an example where the std-deviation ends up being similar. I didn't start from 40/40/20, but ended there as G-K is actually quite good at controlling volatility. I do agree that from other perspectives (e.g. sleep at night when watching your portfolio), this doesn't compare. But then from a bequest perspective, its doesn't compare. From a year-to-year perspective, it doesn't compare. Etc.

In general, G-K and VPW simply do NOT compare. Both are aimed at a very different set of goals. I do believe VPW (and PMT derivatives) has a GREAT place in the continuum of withdrawal methods. But it's not the ONLY one, and there are pros & cons, and other ways to control volatility than an overly conservative AA. That's all I am saying.
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

As to the missing portfolio charts, I didn't include them for the sake of brevity and because they are quite predictable. Here they are. No surprise here. 1966 first, then 1975.

Actually, looking at those charts (knowing that the spending side displays the same low volatility), I am not sure which one makes you better sleep at night? Or how one would define risk-averse? Again, it depends on your perspective and goals. Personally, my REAL sleep-at-night factor is much more driven by higher spend with lower spend volatility and a perpetual portfolio, but again, it's a personal thing.

Image
Image
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

duffer wrote:Thank you all for these excellent posts--they are very helpful. Does the following seem like a fair summary?

In regard to my original question, the consensus here is that Guyton-Klinger is a very acceptable set of decision rules that will tend to maximize the safe withdrawal rate, even permitting users to select an initial SWR in excess of 4.5 or 5%, and also to preserve their portfolio for heirs over a 40 year timeline. The G-K rules appear able to do this even under very unfavorable historical circumstances, such as those with a retirement beginning in 1973 (a period which was subject to both sequence of returns risk and to very high inflation). G-K offers initial safe withdrawal rates of over 4.5% with portfolio success rates of over 95%.
If the objective is to preserve the portfolio for heirs, why not use the much simple constant-percentage withdrawal method, instead? G-K seems overly complex for most people.

Do you really expect grandma or grandpa to go looking for CPI-U values to compute her/his next withdrawal, based on a complex set of rules? But, if you tell her/him that s/he can withdraw 3.5% of her portfolio balance, each year, s/he'll be able to do it.

VPW only requires a tiny bit more from grandma/pa; it requires that s/he be able to use an age-specific percentage:

(30% stocks portfolio, last withdrawal at 99 years old)
Age Percentage
65 4.4%
66 4.5%
67 4.6%
68 4.7%
...
98 50.7%
99 100%

No need to loop up CPI-U values. No need for complex calculations. Just a simple multiplication.

That's way simpler than G-K.
In order to take advantage of G-K, one must be able to adjust one's living arrangements without undue distress to occasional reductions in the amount that one can withdraw during market downturns. Thus, the G-K decision rules work especially well for individuals who can cover some or most of their essential expenses through sources such as social security, annuities, or pensions. Also, being willing and able to vary one's discretionary expenses in a way that does not cause undue distress is important--for example, by cutting back on travel during a year of reduced withdrawals.
I would say that this has nothing to do with G-K; it has to do with withdrawal from a portfolio. If one is completely inflexible, and money is tight, then the most reasonable solution is a bridging CD ladder with as-delayed-as-possible Social Security, and if a little money is left over, an small inflation-adjusted single premium immediate annuity (SPIA) with the remaining. (If money is too tight, a nominal SPIA is probably the best option for left-over money). Bye Bye, any bequest.
For my purposes and situation, these rules appear ideal. We have two social securities and a joint life pension that will cover our essential living expenses. We have strong bequest motivations. We plan to travel and we enjoy both expensive international trips but also less expensive domestic excursions and weeks at the beach. We are very likely to spend more on discretionary expenses in our first 10 or 15 years than we are in later years. All of these factors fit very well with the G-K withdrawal rules and will let us take more as an initial SWR than I otherwise would have dared.
They might seem ideal for you, they don't seem to me. For myself, I would much prefer using:
  • Delayed Social Security to 70.
  • CD ladder for bridging retirement to Social Security.
  • My future pension (or an inflation-adjusted SPIA*, if necessary).
  • Constant percentage on the part of portfolio I wish to leave to heirs.
  • VPW on the part of portfolio I wish to spend.
  • An asset allocation that is adapted to me (and my wife).
(Note that the constant-percentage and VPW could easily be combined in a single PMT-based formula with a non-zero end value, but that would be more difficult to explain to people.)

I have had no trouble explaining VPW (and how to rebalance our three fund portfolio) to my wife, but I would have no hope whatsoever to try getting her to understand G-K rules.

* If one is expecting a reduced need for income, as one ages, it could be argued that a nominal SPIA or a 1% indexed SPIA would be a good fit.
In many ways, the G-K rules seem like a formalized version of what Taylor Larrimore has described he and his wife as doing following his early retirement: taking more when the market was good and less when it was bad in a way that felt safe to them and that permitted them to enjoy life as they went along. This system formalizes what you can safely take each year, without relying so much on one's feelings and intuitive sense.
I will let Taylor decide for himself. I happen to think that the overall VPW-based plan exposed above is a very good approximation of his overall method.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
User avatar
siamond
Posts: 5641
Joined: Mon May 28, 2012 5:50 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by siamond »

duffer wrote:Thank you all for these excellent posts--they are very helpful. Does the following seem like a fair summary?

In regard to my original question, the consensus here is that Guyton-Klinger is a very acceptable set of decision rules that will tend to maximize the safe withdrawal rate, even permitting users to select an initial SWR in excess of 4.5 or 5%, and also to preserve their portfolio for heirs over a 40 year timeline. The G-K rules appear able to do this even under very unfavorable historical circumstances, such as those with a retirement beginning in 1973 (a period which was subject to both sequence of returns risk and to very high inflation). G-K offers initial safe withdrawal rates of over 4.5% with portfolio success rates of over 95%. [...]
Yes, I think your summary is very fair. Again, not suitable for everybody, but it seems that you found something that matches well your own goals, good for you. The only small thing I would change in your summary (besides emphasizing the low volatility on the spending side) is that the portfolio success rate is typically 100%, not 95%. This being said, such metric becomes pretty much irrelevant when looking at (sensible) variable spending methods, and one has to look at other aspects (e.g. volatility, spending average & related growth, how adaptive it is to changes in historical rates-in-return, etc). And the latter aspect is where I have some qualms about G-K. But hey, no withdrawal method is ideal, that is for sure.
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

Longinvest--

Thank you very much for your detailed comments. But I think I clearly prefer the characteristics of the G-K rules and lifestyle and results to the pattern that you prefer. Different strokes...

G-K appears ideal for my purposes, and not very difficult to implement, especially once one has done it for a few years. (Your multi-handed plan seems more complex to me, e.g., constant withdrawal for some money, VPW for other money, CD ladders, etc.) But, different strokes is how it SHOULD be.

G-K also has a very strong provenance in terms of the expertise that was deployed in developing it and testing it. It is an industrial-strength plan, so to speak, and that provenance and testing by experts is very meaningful to me in thinking about living and having the money to enjoy life without working.
Topic Author
duffer
Posts: 318
Joined: Tue Sep 09, 2014 8:18 pm

Re: Guyton-Klinger Withdrawal Decison Rules

Post by duffer »

siamond wrote:
duffer wrote:Thank you all for these excellent posts--they are very helpful. Does the following seem like a fair summary?

In regard to my original question, the consensus here is that Guyton-Klinger is a very acceptable set of decision rules that will tend to maximize the safe withdrawal rate, even permitting users to select an initial SWR in excess of 4.5 or 5%, and also to preserve their portfolio for heirs over a 40 year timeline. The G-K rules appear able to do this even under very unfavorable historical circumstances, such as those with a retirement beginning in 1973 (a period which was subject to both sequence of returns risk and to very high inflation). G-K offers initial safe withdrawal rates of over 4.5% with portfolio success rates of over 95%. [...]
Yes, I think your summary is very fair. Again, not suitable for everybody, but it seems that you found something that matches well your own goals, good for you. The only small thing I would change in your summary (besides emphasizing the low volatility on the spending side) is that the portfolio success rate is typically 100%, not 95%. This being said, such metric becomes pretty much irrelevant when looking at (sensible) variable spending methods, and one has to look at other aspects (e.g. volatility, spending average & related growth, how adaptive it is to changes in historical rates-in-return, etc). And the latter aspect is where I have some qualms about G-K. But hey, no withdrawal method is ideal, that is for sure.
Siamond--

Thanks very much--I am glad you think it is a fair summary and that the rules work for the personal situation I describe. The changes you would make in my summary are both more favorable to the rules and do fit what I have read--I didn't quite have the confidence in my understanding of the details to push the lower spending volatility advantages (in relation to other VWR methods, I assume you mean) and the 100% success rate. But I think you are right on these points.

I have a bunch of additional questions that I would be interested to hear you on. Understanding that everyone's different situation requires a different answer to these questions:

Where would you personally place a typical initial SWR with G-K, and with what guardrails?

What would be the reasonable range that you would place on the initial rate and the guardrails?

How would you estimate the implications of he variations within these choices?

What would you consider a reasonable or generic multi-asset allocation for G-K, with what reasonable variations?

I am asking a lot of questions here, but I think it would be valuable to hear your views on as many of these as you want to take the time to address.
longinvest
Posts: 4449
Joined: Sat Aug 11, 2012 8:44 am

Re: Guyton-Klinger Withdrawal Decison Rules

Post by longinvest »

siamond wrote:
longinvest wrote:Now, it would be interesting to look at the portfolio balance path and withdrawal path in nominal terms. Inflation is very volatile, yet it is mostly felt on longer time periods (therefore the money illusion that fools so many people). In the short term, most people simply adapt their spending patterns.

I don't think that it is a good idea to try lowering inflation-adjusted volatility. Instead, I think that volatility-averse investors should seek to preserve long-term inflation-adjusted buying power and only worry about volatility in nominal terms.
Well, I REALLY don't know how to think in nominal terms. Most metrics become completely counter-intuitive and actually pretty much meaningless, charts or aggregate indicators. I don't see why you're thinking that way though. Sure, there is some discomfort in spending significant more (or less) year over year just due to inflation, but at the end this discomfort is nothing compared to your standard of living. And that is something to be assessed in a REAL terms, don't you think?
Based on our prior discussions, you don't seem to be a risk-averse person, so I can easily understand why you would never want to put a lot of bonds in your portfolio. :)

Many people hate to see big losses in their portfolio. Even if you try to explain that they didn't really lose 50%, they only lost 48% as there was 2% deflation over the last year, they won't be happy and they'll want to liquidate their stocks at the worst time to do so.

Even in high-inflation periods, the rent of tenants didn't go up every month. Municipal taxes didn't go up every month. When the price of gas goes up, a retiree can decide to stay at home more (lower car usage). etc.

Here are charts showing inflation-adjusted and nominal amounts for withdrawals and portfolio balance when using VPW with a 30% stocks portfolio over a 35-year period:

Image

Image

Notice how in the 1973-1982 period, this investor had mostly increasing withdrawals (or very slim reductions), every year, and an increasing portfolio in nominal terms. No blood bath in the horrible 1973-1975 period, unlike with a stock-heavy portfolio.

Over this period, his withdrawals were losing 30% below the $40K planning number of a 4% SWR. I am sure that a part of this 30% could have been averted by selective spending choices of the retiree.

Over the 9-year period, 1973-1982, the retiree did certainly feel a loss in buying power, but he most certainly didn't feel as bad as one is lead to imagine by only looking at an inflation-adjusted chart. On the contrary, I'm actually sure that he felt pretty good, in contrast to some of his stock-heavy neighbors who sold their stocks in 1975, locking in huge losses. (Maybe they were leveraged and got a margin call. What greed gets some people to do!)

If one is highly risk tolerant, one could go with a 75% stocks portfolio. I personally think Graham's recommendation to stay within the 25% to 75% range is a good choice for most investors. But, there's a caveat: if one or his/her significant other panics during a downturn, irreparable damage could be done to a portfolio.
Last edited by longinvest on Thu Mar 05, 2015 10:27 am, edited 2 times in total.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
Post Reply