Guyton-Klinger withdrawal model

RogerPensionGuy

I feel all these different views and rules on a safe withdrawal % are great to read and food for thought. 

In my case I use them as a guide, but my personal SWR Safe Withdrawal Rate will have a big margin buffer built in to investments that tend to go up & down.

One important view I hold is investments should be held in low cost environments and simple to operate especially as we get get older. There appears a big press on low cost investments during the building years like plenty of headlines saying the difference in value of 30 or 40 years of investing, if people start investing at maybe 18, that 30 or 40 viewing window needs to be about 70 or 80 years in my opinion. 

I like having a few varieties of income, mine are DB, Annuity, ISAs, GIAs, State Pension, cash and SIPP.

I'm happy to leave value in the SIPP due potential IHT benefits for others, but I'm also happy to empty it if I like. 

Back on topic for me, low cost SIPP and venting at between 3 & 3.5% appears pretty ok, but if and when the big drop occurs and historically it will come, maybe turn off SIPP withdrawals for a few years and use other suitable cash flows.

This topic sorta reminds me about observations from well heeled people, they very often have various different cash vehicles to use as required. 

I think after years of people not using annuities, slowly they appear more fashionable. 

3% or whatever number people pick, it's well written that an unlucky sequencing tome event can be tuff to manage and very very especially if in the early years, lets remember, in the future at some point, we will probably see the longest downturn in markets, maybe this will occur very shortly, but it probably will occur and that could be tuff for people if they haven't mitigated well enough for it.
☆☆☆

https://www.investopedia.com/terms/s/sequence-risk.asp#:~:text=Protecting against sequence risk means,in your peak earning years.

OldScientist

MetaPhysical said:

What is the collective wisdom on this strategy?  It certainly seems sensible and common sense.  Obviously, no model can accommodate every possibility and all models will fail with a series of poor years early on, statistically unlikely though that would be.  But generally speaking, what do people think?  it seems "safer" than the 4% rule.

In the following plot, the real (i.e., inflation adjusted) withdrawal rate (WR, upper row) and current portfolio value (CPV, lower row) both expressed as a percentage of the initial portfolio value (%IPV) are plotted for a UK retirement case starting in 1937 (one of the worst years for UK retirements) for GK (with standard parameters, e.g., plus/minus 10%) in the left column and constant inflation adjusted withdrawals (i.e., the '4% rule') in the right each for 3 different initial withdrawal rates (3%, 4%, and 5%)*.



From the figure, the benefits of GK (in this example) compared to using constant inflation adjusted withdrawals are:
1) The portfolio did not run out of money in the first 30 years (it even increased over the last few years)
2) The outcome was relatively insensitive to the initial withdrawal rate (i.e., the risk of getting that choice wrong is reduced)

The problems with GK in this example is
1) The withdrawals got quite small (~1.5%) after the first 15 years or so and did not recover much even when the portfolio did.

Personally, I think GK is slightly more complicated than other hybrid withdrawal methods (e.g., Vanguard dynamic, Carlson's endowment method, and portfolio smoothing) and produces slightly worse results. The discontinuities in it also offend my mathematical sensibilities (but that is not a reason to not use it!).

Anyway, GK is 'safer' than the '4% rule' provided your retirement spending can be flexible. How flexible then depends on the level of your guaranteed income (e.g., state pension) compare to the income derived from your portfolio and your what it gets spent on - but that is a different discussion.

* I've used a portfolio with 60% UK stocks, 20% long gilts, and 20% UK 3-month bills with returns and inflation sourced from macrohistory.net - other portfolios would have produced different results, but the conclusions would be similar.

Bostonerimus1

OldScientist said:

MetaPhysical said:

What is the collective wisdom on this strategy?  It certainly seems sensible and common sense.  Obviously, no model can accommodate every possibility and all models will fail with a series of poor years early on, statistically unlikely though that would be.  But generally speaking, what do people think?  it seems "safer" than the 4% rule.

In the following plot, the real (i.e., inflation adjusted) withdrawal rate (WR, upper row) and current portfolio value (CPV, lower row) both expressed as a percentage of the initial portfolio value (%IPV) are plotted for a UK retirement case starting in 1937 (one of the worst years for UK retirements) for GK (with standard parameters, e.g., plus/minus 10%) in the left column and constant inflation adjusted withdrawals (i.e., the '4% rule') in the right each for 3 different initial withdrawal rates (3%, 4%, and 5%)*.



From the figure, the benefits of GK (in this example) compared to using constant inflation adjusted withdrawals are:
1) The portfolio did not run out of money in the first 30 years (it even increased over the last few years)
2) The outcome was relatively insensitive to the initial withdrawal rate (i.e., the risk of getting that choice wrong is reduced)

The problems with GK in this example is
1) The withdrawals got quite small (~1.5%) after the first 15 years or so and did not recover much even when the portfolio did.

Personally, I think GK is slightly more complicated than other hybrid withdrawal methods (e.g., Vanguard dynamic, Carlson's endowment method, and portfolio smoothing) and produces slightly worse results. The discontinuities in it also offend my mathematical sensibilities (but that is not a reason to not use it!).

Anyway, GK is 'safer' than the '4% rule' provided your retirement spending can be flexible. How flexible then depends on the level of your guaranteed income (e.g., state pension) compare to the income derived from your portfolio and your what it gets spent on - but that is a different discussion.

* I've used a portfolio with 60% UK stocks, 20% long gilts, and 20% UK 3-month bills with returns and inflation sourced from macrohistory.net - other portfolios would have produced different results, but the conclusions would be similar.

Your analysis just confirms my belief that with any strategy the foundation is understanding your spending and having a plan to reduce spending in tough times. My love of feedback makes me despise a blind SWR %age plus inflation approach whatever the portfolio balance is doing...and whatever the Monte Carlo analysis shows.

OldScientist

Bostonerimus1 said:

OldScientist said:

MetaPhysical said:

What is the collective wisdom on this strategy?  It certainly seems sensible and common sense.  Obviously, no model can accommodate every possibility and all models will fail with a series of poor years early on, statistically unlikely though that would be.  But generally speaking, what do people think?  it seems "safer" than the 4% rule.

In the following plot, the real (i.e., inflation adjusted) withdrawal rate (WR, upper row) and current portfolio value (CPV, lower row) both expressed as a percentage of the initial portfolio value (%IPV) are plotted for a UK retirement case starting in 1937 (one of the worst years for UK retirements) for GK (with standard parameters, e.g., plus/minus 10%) in the left column and constant inflation adjusted withdrawals (i.e., the '4% rule') in the right each for 3 different initial withdrawal rates (3%, 4%, and 5%)*.

   [snip]

Your analysis just confirms my belief that with any strategy the foundation is understanding your spending and having a plan to reduce spending in tough times. My love of feedback makes me despise a blind SWR %age plus inflation approach whatever the portfolio balance is doing...and whatever the Monte Carlo analysis shows.

I'd agree, my own view is that retirement planning (from, IMO, most important to least) involves

1) Budget (including estimates of essential/discretionary or core/adaptive spending however you want to label/define these)
2) Sources of guaranteed income (state pension, DB pensions, annuities - preferably RPI linked) - ideally enough to cover essential/core spending
3) Portfolio withdrawals - if guaranteed income covers most of the basics (which for the vast majority of people, just the state pension will*), then the exact method becomes rather immaterial.

* The median pension pot is about £80k on first withdrawal (FCA data), at 3-4% withdrawals this will provide  £2.4k-£3.2k - small compared to the state pension.

OldScientist

[Deleted User] said:

OldScientist said:

* The median pension pot is about £80k on first withdrawal (FCA data), at 3-4% withdrawals this will provide  £2.4k-£3.2k - small compared to the state pension.

I do wonder how meaningful statistics like this are. I've had nine pension crystallisation events in the last ten months.  Some were chosen by me but the rest were admin issues.  The smallest was 19p and two others were less than £200.  

You're right - I missed off two provisos - the £80k or so is the median pot size on first access of those pots that entered drawdown and were not fully exhausted (see Table 3 in downloadable Excel tables ).

However, Figure 2 at https://www.fca.org.uk/data/retirement-income-market-data-2021-22 indicates that the median pot size, of all pots on first access, was in the range £10-30k (which strengthens my point that for most people drawdown is likely to provide a relatively small component of their retirement income). According to the same graph, the modal pot size was less than £10k and most of these were fully withdrawn (hence skewing the results in Table 3 linked above). Of course, many people, as in your case, will have more than one pot. If I remember correctly, there are some data at the ONS on household savings (which I think includes pension pots) that might be more meaningful.

westv

OldScientist said:

I've used a portfolio with 60% UK stocks, 20% long gilts, and 20% UK 3-month bills with returns and inflation sourced from macrohistory.net - other portfolios would have produced different results, but the conclusions would be similar.

When you say other portfolios, I assume you mean portfolios with a wider geographical spread. I doubt anybody would use 100% UK investments for drawdown.

OldScientist

Since my previous couple of posts were perhaps heading off the direct topic and in my earlier post I mentioned other withdrawal approaches that could be used instead of GK, here are the outcomes for two of the ones I mentioned for the same retirement and portfolio as previously.



Carlson's endowment method mixes inflation adjusted withdrawals (IAW) with percentage of portfolio withdrawals. In the above example, for each initial withdrawal, the IAW component is 2% (i.e., 2% of the initial portfolio then the amount is adjusted by inflation in subsequent years), while the percentage of portfolio component was 1%, 2%, and 3% (to make initial withdrawals of 3%, 4%, and 5%, respectively). Comparing with GK, the advantage is that the withdrawals did not fall below 2% (except for the 5% case where the portfolio was exhausted by 1963 - there is still some requirement to carefully select the IAW value). I note that the portfolio was spent down which is a benefit for those not fussed about leaving a legacy and a disadvantage for those that are.

Portfolio smoothing uses a percentage of portfolio approach except that the portfolio value is averaged, in this example, over the previous 3 years - this means the effect of temporary changes in portfolio value ('blips') on withdrawals are reduced but longer term changes do, eventually, affect the withdrawals. In this example, the withdrawals are quite similar to those from GK, except that the ones earlier in retirement are a bit smaller and the ones in later retirement a bit bigger and the withdrawals recover after the minimum that occurred in the 1950s. The CPV values are similar to GK. In my view, the approach is much simpler to implement than GK and it can also be used with other percentage of portfolio approaches instead of the simple constant percentage used here, such as Bogleheads ABW (see https://www.bogleheads.org/wiki/Amortization_based_withdrawal ), VPW (see https://www.bogleheads.org/wiki/Variable_percentage_withdrawal ), 1/N, etc..

Albermarle

A bit surprising that @jamesd has not joined in to what was his usual favourite subject. Seems to have gone quiet again after a burst of rather intense activity in January.

OldScientist

westv said:

OldScientist said:

I've used a portfolio with 60% UK stocks, 20% long gilts, and 20% UK 3-month bills with returns and inflation sourced from macrohistory.net - other portfolios would have produced different results, but the conclusions would be similar.

When you say other portfolios, I assume you mean portfolios with a wider geographical spread. I doubt anybody would use 100% UK investments for drawdown.

True - and a world portfolio clearly offers more diversification. However, the variation of outcomes between different retirement starting years (e.g. SWR varies from about 3% to 12%) vastly exceeds that of using different portfolios (e.g., SWR varies from about 3.0% to 3.5%). In other words, the data suggest that for retirement planning purposes, the difference between using a UK portfolio in backtesting and a more diverse one are marginal. I also note that how you construct a world portfolio across certain historical events (e.g., German hyperinflation and the effect of WWII and Japan and WWII) has a distinct effect. In this sense, historical backtesting then becomes useful for comparing methods but rather poor for selecting precise initial withdrawals (one of the reasons why dynamic methods such as GK are better than inflation adjusted withdrawals is that they are relatively tolerant the initial conditions chosen).

OldScientist

Just to demonstrate what O mean about the difference between portfolios... in the following figure I have plotted the 0th (worst case), 10th, 25th, 50th (median), and 75th percentiles of real withdrawal rate as a function of time since retirement for all historical retirements (rather than just the 1937 case) for GK with an initial withdrawal of 4% and two portfolios held in the UK, 60% UK stocks, 20% UK gilts, 20% UK cash (top panel), and 30% UK stocks, 30% US stocks, 20% UK gilts, and 20% UK cash (lower panel)*



While there are obvious differences between the two cases, for example, the median WR for the UK portfolio falls below that for the UK/US portfolio towards the end of the 30 year retirement (3% compared to 4% after 29 years), it is also clear that the difference between the worst historical case (WR of about 1.5% after 29 years) and the 75th percentile (between 5% and nearly 6%) is much larger. It is not possible to determine what path a future retirement will follow, but the information in the above graph tells us that in the median historical case, GK supported withdrawals above the SAFEMAX values of 2.9% (UK) and 3.5% (UK/US), but in the worst cases, the income was lower (but the portfolio didn't run out of money after 30 years).

So, GK is a reasonable approach provide this amount of flexibility is acceptable, which in turns rather depends on expenditure, other sources of income, and the preferences of the individual retirees. For example, for a couple with two state pensions and a £100k portfolio, total real income after 30 years would have ranged from £22.5k (£21k+£1.5k from SP and portfolio, respectively) in the worst historical cases to £27k (£21k+£6k) at the 75th percentile. Of course, the variation in total income would have been larger for a couple with a larger portfolio (assuming no other sources of guaranteed income) and, again, whether this is acceptable depends on individual circumstances and requirements.

Sorry, drifting a bit of topic towards the end there!

* I'm in the middle of trying to construct a more international return series from the macrohistory.net database and actual indices (e.g., MSCI returns in USD go back as far as 1978 - this is easy enough to convert to returns in GBP), but this example will do for an illustration.