4% SWR rule….well, rules are there to be broken!

gm0

Michael McClung - Living off your money.  Excellent insomnia cure which explores this topic in a structured manner. 

My takeaways if TL;DR (which for many it undoubtedly is)

- There are marginal gains available from different methods and approaches to managing a pot in deaccumulation.

- More positively a fair range of commonly described methods do all basically work - adequately (comforting)

- A few do fall a bit short in particular conditions (less comforting)

Tolerating Variable income is particularly beneficial to efficiency. More efficient income harvesting.  (More income - fewer leftovers - still fairly safe along the journey vs running out in worst of cohort sequences)

You can use caps and collars which reduces the risk management yield of variable income to success but in turn provides greater certainty around an income floor. Which is nice.

The asset approach and what is held and what is sold in what order can also have 2nd order helpful affects.

Yet the whole thing is (in WR %) annual magnitude ~0.25% level not ~2.5%  wrt gains. South of 0.5%
The benefit of more thorough testing with "difficult" market conditions and MC simulation is not that - but the confidence that a plan continues to operate fairly linearly and without "surprises" across a broad range of to be encountered conditions  

Also a number of techniques are parameterised and have their parameters or thresholds set and tuned to optimise via the use of backtesting US market data.  Magic numbers are somewhat offensive here as in physics and yet are a thing.  Testing these with other market data series outside USA as McClung attempts - is important to validate whether a particular approximation is of any practical use or just data mining noise in the US market sequence.

I consider 0.25% when we are discussing safe WR rates at 3.5% for UK in GBP as "helpful" enough to be interesting.   Of course market returns and currency FX expecially will dominate.  And platform and advice costs (unless needed and valued) are worth addressing also.

- Envelope testing (MC Simulation of known StdDev or markets/adjusted markets sequence backtests) does tell us which plans behave worse or better. And which failed more often in the given range of data simulation of market conditions and cohorts (or MC runs).  This is (of course) not predictive at all of future markets.  Any stress test I create.  You can always +10% my worst of the worst.  But it is not "nothing".   An already demonstrated to be weaker plan across a range of stress test conditions - is non-preferred over one that isn't.  Probably.  Already failed inside the estimated envelope. Vs "could fail" in novel conditions.

If you reject the idea that this is helpful to planning in making method/algo comparisons.

Then all outputs of backtesting and MC simulation of distributions of returns and random sequences are likely fairly worthless to you.  Due to the unknowable future.

In which case you need an alternative way to choose allocations, methods and plans to suit this

A few starter suggestions.  Dice. Oracular divination.   Chatgpt. (Which to be fair does suggest the 4% rule and a range of 30% to 70% equities for conservative to growth orientated.  It was forthcoming on US funds but coy about recommending anything on LSE for UK investors.  Asked questions about the impact of being a sterling investor on the 4% rule - it is surprisingly effective at laying out some relevant considerations albeit few answers or details).  Just don't ask how many r's in strawberry.

michaels

Thank you.  Lots of common sense there

Bostonerimus1

gm0 said:

Michael McClung - Living off your money.  Excellent insomnia cure which explores this topic in a structured manner. 

My takeaways if TL;DR (which for many it undoubtedly is)

- There are marginal gains available from different methods and approaches to managing a pot in deaccumulation.

- More positively a fair range of commonly described methods do all basically work - adequately (comforting)

- A few do fall a bit short in particular conditions (less comforting)

Tolerating Variable income is particularly beneficial to efficiency. More efficient income harvesting.  (More income - fewer leftovers - still fairly safe along the journey vs running out in worst of cohort sequences)

You can use caps and collars which reduces the risk management yield of variable income to success but in turn provides greater certainty around an income floor. Which is nice.

The asset approach and what is held and what is sold in what order can also have 2nd order helpful affects.

Yet the whole thing is (in WR %) annual magnitude ~0.25% level not ~2.5%  wrt gains. South of 0.5%
The benefit of more thorough testing with "difficult" market conditions and MC simulation is not that - but the confidence that a plan continues to operate fairly linearly and without "surprises" across a broad range of to be encountered conditions  

Also a number of techniques are parameterised and have their parameters or thresholds set and tuned to optimise via the use of backtesting US market data.  Magic numbers are somewhat offensive here as in physics and yet are a thing.  Testing these with other market data series outside USA as McClung attempts - is important to validate whether a particular approximation is of any practical use or just data mining noise in the US market sequence.

I consider 0.25% when we are discussing safe WR rates at 3.5% for UK in GBP as "helpful" enough to be interesting.   Of course market returns and currency FX expecially will dominate.  And platform and advice costs (unless needed and valued) are worth addressing also.

- Envelope testing (MC Simulation of known StdDev or markets/adjusted markets sequence backtests) does tell us which plans behave worse or better. And which failed more often in the given range of data simulation of market conditions and cohorts (or MC runs).  This is (of course) not predictive at all of future markets.  Any stress test I create.  You can always +10% my worst of the worst.  But it is not "nothing".   An already demonstrated to be weaker plan across a range of stress test conditions - is non-preferred over one that isn't.  Probably.  Already failed inside the estimated envelope. Vs "could fail" in novel conditions.

If you reject the idea that this is helpful to planning in making method/algo comparisons.

Then all outputs of backtesting and MC simulation of distributions of returns and random sequences are likely fairly worthless to you.  Due to the unknowable future.

In which case you need an alternative way to choose allocations, methods and plans to suit this

A few starter suggestions.  Dice. Oracular divination.   Chatgpt. (Which to be fair does suggest the 4% rule and a range of 30% to 70% equities for conservative to growth orientated.  It was forthcoming on US funds but coy about recommending anything on LSE for UK investors.  Asked questions about the impact of being a sterling investor on the 4% rule - it is surprisingly effective at laying out some relevant considerations albeit few answers or details).  Just don't ask how many r's in strawberry.

With many people spending 1% in financial fees it would seem sensible to reduce those. That 1% should become a smaller part of your budget as your withdrawal amount increases with inflation, but at the beginning of retirement it might be a quarter of your spending and your largest single expense. Worry about the fees you are paying before you worry about a particular drawdown strategy.

TheGreenFrog

As others have said or implied, an SWR is at best a guide and I would favour being as conservative as possible in the initial phases of any drawdown or similar strategy.  Even taking as much risk as you can out of the equation leaves some risk:  if you put your retirement pot into IL gilts then a 4% SWR will mean you have mortality risk (unless you are >75 when you start!) as well as some residual inflation risk (i.e. your personal inflation rate being higher than the gilt rate).  And if you buy an annuity you are only taking away the mortality risk.

Mick70

the 4% rule is something I/We are going to use , although half of our pension will come from my DB pension , which does help in this regard.  when get time I'm going to post about my own situation, to see if what I am planning makes sense.  But as say, that rule is the one I intend to use

Linton

As someone who has been retired for nearly 20 years (early retirement!) my belief is that SWR should only be used as a rule-of-thumb sanity check of your retirement plans.  If the drawdown income you need to meet your expenses is less than 3-3.5% of your pot you should be OK.

If there is a major and lengthy crash early in retirement taking drawdown of much more than the SWR greatly adds to the risk that you will eat into the core investments needed to generate future income.  If the feared crash does not occur, as it probably wont, keeping to the SWR is likely to ensure you that you die rich, possibly richer than you were when you retired.  It is one extreme or the other, there isn't a predictable comfortable mid point.

If you are significantly dependent on drawdown income you will need to develop your own way of managing your finances in this environment.   I have mine but it wont necessarily suit everyone.

OldScientist

Systematic methods of drawdown from a portfolio can possibly be split into three broad categories

1) Constant inflation adjusted withdrawals ('SWR'): The amount withdrawn is known (the infamous '4%'), but the length of time over which it can be sustained is unknown and unknowable.
2) Percentage of portfolio withdrawals (constant percentage, VPW, ABW): The amount withdrawn is unknown in advance, but it will never go to zero.
3) Hybrid approaches (Guyton-Klinger, Vanguard dynamic, and many of the ones tested in McClung's book - referred to upthread). These have properties somewhere between the two - the variability of withdrawals is less and there is a reduced chance of portfolio exhaustion.

I've posted the following figure before. In the top panel the real (i.e, inflation adjusted) portfolio value* and in the lower panel the real withdrawal rate are plotted as a function of time for 5 different withdrawal strategies (CIAW=constant inflation adjusted withdrawals, GK=Guyton-Klinger, VG=vanguard dynamic, CP=constant percentage of portfolio, and MX is a 50/50 mix of CIAW and CP) for a single UK historical retirement case starting in 1937 (one of the worst for UK retirees).



The constant inflation adjusted strategy provided a constant inflation adjusted income until the portfolio ran out of money just under 20 years into retirement. The constant percentage of portfolio strategy delivered a highly variable income (ranging from 4% at the start to a minimum of 1.4% after about 20 years) but never ran out of money (in real terms, after 35 years, the portfolio was still worth about 50% of the initial value), while the hybrid methods (GK, VG, and MX) fell somewhere between CIAW and CP both in terms of the income delivered and the portfolio remaining after 35 years.

From a psychological point of view, the retiree following the SWR (CIAW) approach would have needed strong nerves to continue to take a constant real amount as the value of the portfolio declined (in real terms) to 50% after four years (resulting in a withdrawal of 8% of the remaining portfolio) and to 20% after 13 years (resulting in a withdrawal of 20% of the remaining portfolio). The retirement wouldn't have been nice for the other strategies, but at least withdrawals would have been cut in response to the drops in portfolio value over the first 20 years or so.

Which of these strategies is 'best' rather depends on the consequences of either variable income or complete portfolio exhaustion. My own preference has been to secure enough income floor in guaranteed income (DB pension and, eventually, SP - I haven't needed to add a RPI annuity) not to have to worry about variability in portfolio income, so I have ended up using ABW (which is a percentage of portfolio approach where the percentage increases with time, see https://www.bogleheads.org/wiki/Amortization_based_withdrawal of which, VPW is a better known case, see https://www.bogleheads.org/wiki/Variable_percentage_withdrawal ).



* A portfolio of 60% UK stocks, 20% UK long bonds, and 20% UK cash was used as an illustration. Asset returns and UK inflation from macrohistory.net. While the details would be different for a different portfolio, the broad conclusions would not.

michaels

OldScientist said:

Systematic methods of drawdown from a portfolio can possibly be split into three broad categories

1) Constant inflation adjusted withdrawals ('SWR'): The amount withdrawn is known (the infamous '4%'), but the length of time over which it can be sustained is unknown and unknowable.
2) Percentage of portfolio withdrawals (constant percentage, VPW, ABW): The amount withdrawn is unknown in advance, but it will never go to zero.
3) Hybrid approaches (Guyton-Klinger, Vanguard dynamic, and many of the ones tested in McClung's book - referred to upthread). These have properties somewhere between the two - the variability of withdrawals is less and there is a reduced chance of portfolio exhaustion.

I've posted the following figure before. In the top panel the real (i.e, inflation adjusted) portfolio value* and in the lower panel the real withdrawal rate are plotted as a function of time for 5 different withdrawal strategies (CIAW=constant inflation adjusted withdrawals, GK=Guyton-Klinger, VG=vanguard dynamic, CP=constant percentage of portfolio, and MX is a 50/50 mix of CIAW and CP) for a single UK historical retirement case starting in 1937 (one of the worst for UK retirees).



The constant inflation adjusted strategy provided a constant inflation adjusted income until the portfolio ran out of money just under 20 years into retirement. The constant percentage of portfolio strategy delivered a highly variable income (ranging from 4% at the start to a minimum of 1.4% after about 20 years) but never ran out of money (in real terms, after 35 years, the portfolio was still worth about 50% of the initial value), while the hybrid methods (GK, VG, and MX) fell somewhere between CIAW and CP both in terms of the income delivered and the portfolio remaining after 35 years.

From a psychological point of view, the retiree following the SWR (CIAW) approach would have needed strong nerves to continue to take a constant real amount as the value of the portfolio declined (in real terms) to 50% after four years (resulting in a withdrawal of 8% of the remaining portfolio) and to 20% after 13 years (resulting in a withdrawal of 20% of the remaining portfolio). The retirement wouldn't have been nice for the other strategies, but at least withdrawals would have been cut in response to the drops in portfolio value over the first 20 years or so.

Which of these strategies is 'best' rather depends on the consequences of either variable income or complete portfolio exhaustion. My own preference has been to secure enough income floor in guaranteed income (DB pension and, eventually, SP - I haven't needed to add a RPI annuity) not to have to worry about variability in portfolio income, so I have ended up using ABW (which is a percentage of portfolio approach where the percentage increases with time, see https://www.bogleheads.org/wiki/Amortization_based_withdrawal of which, VPW is a better known case, see https://www.bogleheads.org/wiki/Variable_percentage_withdrawal ).



* A portfolio of 60% UK stocks, 20% UK long bonds, and 20% UK cash was used as an illustration. Asset returns and UK inflation from macrohistory.net. While the details would be different for a different portfolio, the broad conclusions would not.

Do you know what the SWR was for this dataset?  It is worth fully considering that even the least variable withdrawal method required a halving of real income (that £1m pot and retirement at 60 quickly becomes a real income of 20k pa rather than the expected 40k) for large parts of the retirement journey - and this was for a 4% swr which many are saying is too conservative!

OldScientist

michaels said:

OldScientist said:

Systematic methods of drawdown from a portfolio can possibly be split into three broad categories



* A portfolio of 60% UK stocks, 20% UK long bonds, and 20% UK cash was used as an illustration. Asset returns and UK inflation from macrohistory.net. While the details would be different for a different portfolio, the broad conclusions would not.

Do you know what the SWR was for this dataset?  It is worth fully considering that even the least variable withdrawal method required a halving of real income (that £1m pot and retirement at 60 quickly becomes a real income of 20k pa rather than the expected 40k) for large parts of the retirement journey - and this was for a 4% swr which many are saying is too conservative!

The 30 year MSWR (i.e., zero failures) was about 3.0% (about 3.3% for a 10% failure rate). It certainly fits with the idea of a ballpark 3.0 to 3.5% for the UK.

Either the complete failure (if the actual SWR turns out to be lower than the initial guess) or a 50% drop in real income with variable strategies is why I think flooring is so important. For example, taking your example of a £1m portfolio. With two state pensions (a total of about £20k for simplicity), the income from the portfolio dropping from £40k to £20k with a variable strategy means the overall income drops from £60k to £40k which would unpleasant for those expecting a lifestyle needing £60k per year, but not completely disastrous.

An RPI annuity (currently about 4% payout for joint life, 100% benefits, 65yo - edit: about 3.4% at 60yo) bought with half the portfolio would, together with the SP, give a floor of about £40k, with the remaining half of the portfolio providing an initial £20k dropping to £10k, i.e., the overall income of £60k dropping to £50k with is far more manageable. Of course in a good retirement, the purchase of the annuity will reduce the upside potential, but (again in my view), good retirements can look after themselves!

Ibrahim5

Must make a big difference if you have an IFA living off your investments too. Porsches and golf clubs don't come cheap.