4% Drawdown If Preservation Of The Capital Is Not A Concern ?

MK62

OldScientist said:

As @Linton says, the 'safe' refers to historically safe and is different for different countries and different asset allocations (not only different ratios of stocks and bonds, but the maturity of bonds) - it is not a physical law!

The maximum inflation adjusted withdrawal rates that avoided running out of money before the end of the planning period for 60% UK stocks and 40% UK bonds (longish maturity) are given in the following graph (UK returns and UK inflation from macrohistory.net, my own calculations). SAFEMAX was about 2.9%.



There were two problem periods:
1) Early 20th century (high inflation in WWI)
2) 1937 (stock market crash, WWII, and post-war currency and debt problems)

The relatively low SWR that the US simulators give in the mid-1960s were not present in the UK.

It is probably unlikely that anyone would invest solely in UK stocks nowadays, so here is the same graph with an allocation of 30% UK stocks, 30% US stocks, and 40% UK bonds (at some stage I'll get around to calculating a cap-weighted international index from the mactohistory.net dataset). The SAFEMAX was 3.3% for this allocation



The worst cases remained centred around the same periods as before, but the SWR was now a little higher. However, not all retirements were improved by holding US stocks - e.g., the dip in the late 1960s is worse with US stocks than without.

Are those US equity return figures for a US based investor or a UK based investor.....just asking as the returns each might get, especially in drawdown, might be quite different.

jamesd

Linton said:

It would be interesting to see an SWR simulation that started just before 2001 but sadly it is too early for that.

Kitces took a look back in 2015 and they were roughly on a par with 1929 and better than the other bad US starts. Favourable conditions then the recent inflation blip probably produced recovery then back to 1929 but I haven't looked closely. In 2015 the just before 2008 group were already well ahead but since 2022-3 inflation was half way through a 30 year plan that may lead to more stress later.

jamesd

OldScientist said:

In real terms, the portfolio has fallen to just under 40% of the original value and must now last another 8 years or so (assuming a 30 year planning period) which may (or may not) be successful. Alternatively, assuming they were 65 in 2000, an RPI annuity could now be bought with a payout rate of 9.5% (single life, 10 year guarantee, quote from moneyhelper assuming no health problems), 17.1% (single life, no guarantee) or 12.9% (joint life, no guarantee). With the former an income of 3.8% (of the original portfolio) could be purchased (leaving nothing in the portfolio) while in the second case, an inflation protected income of 4% could be purchased with about 60% of the remaining portfolio leaving some legacy, while with the joint annuity 4% income could be purchased with just under 80% of the remaining portfolio.

This group reached state pension age before 6 April 2016 and still has the option to defer their state pension to get 10.4% increase per year of deferral with CPI increases.

It's pretty much standard for me to recommend at least five years of current 5.8% state pension deferring at the start, partly for longevity protection and partly because it pays more than SWRs do so each year increases both safety and income.

If you're willing it'd be interesting to see what deferral does to the numbers, though the recent inflation blip which really shows its greatest strength is too recent unless you fake that for a much earlier year.

Linton

MK62 said:

Linton said:

It would be interesting to see an SWR simulation that started just before 2001 but sadly it is too early for that.

You can model for a 24 year timespan though - but instead of running your pot down to (near) depletion, just model it to run your pot down to a minimum of 6/30ths of the starting amount (for a 30yr retirement), adjusted for inflation over the 24 yrs.

SUrely that does not work since after 24 years you are withdrawing 4% of the original 100% pot, not 4% of the then current pot.  In @OldScientist 's example after the 24 years or so you would be drawing down 10% of the pot each year.  The depletion rate increases with the amount of depletion as with a mortgage.

jamesd

OldScientist said:

It is probably unlikely that anyone would invest solely in UK stocks nowadays, so here is the same graph with an allocation of 30% UK stocks, 30% US stocks, and 40% UK bonds (at some stage I'll get around to calculating a cap-weighted international index from the mactohistory.net dataset). 

Perhaps worth mentioning that it was adding US small caps that led to the increase from 4% to 4.5% for constant inflation- adjusted income. That also led to 60:40 equities:bonds becoming optimal instead of the substantially wider previous equity range.

I use a global small cap fund as part of my undrawn mix and the inherently very small VCTs for the tax efficiency of pot withdrawing portion that's outside the pension.

MK62

Linton said:

MK62 said:

Linton said:

It would be interesting to see an SWR simulation that started just before 2001 but sadly it is too early for that.

You can model for a 24 year timespan though - but instead of running your pot down to (near) depletion, just model it to run your pot down to a minimum of 6/30ths of the starting amount (for a 30yr retirement), adjusted for inflation over the 24 yrs.

SUrely that does not work since after 24 years you are withdrawing 4% of the original 100% pot, not 4% of the then current pot.  In @OldScientist 's example after the 24 years or so you would be drawing down 10% of the pot each year.  The depletion rate increases with the amount of depletion as with a mortgage.

Not sure I follow....after 24 years, you'd be drawing down 4% of the original portfolio (or whatever % you deemed appropriate....4% is actually too high for the 2000-2023 data set***), adjusted for inflation over the 24 years........after 24 years of a 30 year plan, you'd only have 6 years to go (assuming you stuck to the 30yr plan).....the percentage of the portfolio in year 24 would probably be iro 15% of the remaining pot, at that point.

***If you model a 4% starting withdrawal in 2000, with subsequent withdrawals increased by RPI each year, then you'd probably run out of money around 23 years later (ie around now - it's impossible to be precise here as it'll come down to exactly what you were invested in over the period, whether that stayed the same thoughout the period, and in the same relative quantities, and when exactly you did the drawdown each year.

If you started with £400k, you'd want 6/30ths of that remaining after 24 years = £80k, but then adjusted for 24 years of inflation.....using RPI, that would be iro £182k. That would entail a starting rate iro 3.27% (£13080), which after 24 years of RPI inflation would have grown to a withdrawal of c£29860 for year 25.......about 16.5% of the remaining £182k pot.

jamesd

Linton said:

The situation of a Sequence Of Returns before the start of modelling is a general problem with using SWRs...

One can as @jamesd suggests repeat the SWR calcuolation every few years. But as pointed out..

SWR takes no recognition of the state of the market at the time.  So if you determine the SWR after 5 years of rising prices you will get the same answer as soon after a large crash.  In the first case a recalculated SWR could well be be dangerously optimistic and in the second case it will be unnecessarily pessimistic even if the SWR itself was accurate.

SWRs that take account of estimated market conditions can be used and the use of P/E in those handles much of the boom or crash issue.

The main application I'd suggest for recalculation is if using the 4% rule and markets have done well for a sustained period, such that there's a greatly increased risk of not drawing as much income as has become possible.

The assertion that the SWR could be dangerously optimistic is untrue if you accept the underlying premise, that SWRs are safe for anything likely to happen based on historic performance. They already are calculated to work just before a major crash.

Bostonerimus1

MetaPhysical said:

Hi QB, thanks for your thoughts.  No, not quite.  Looking for around 50k gross to stay below the 50270 upper earnings limit.  So I have 24k index linked  at 58.   The 26k from the DC will only need to be at that level until 67 when it can then drop to 16k or so because the SP will then take up £10k.  Of course, all this is in today's numbers.  and as you allude to, this is all very subject to growth and inflation and that some years the withdrawals will need to be less if the market is down.  I might make up some of that 26k from ISA's as well.  Like someone else said, pensions were static for almost two years with no real growth - and even loss in many cases.  Since October they have come shooting back,.

Many thanks for the links to the tool.  I shall investigate that.

Many excellent posts have pointed out the uncertainties and probabilistic nature of safe withdrawal calculations. One aspect of the Bengen 4% historical type approaches is there is no feedback from your current circumstances and so I think the Guyton Klinger type approaches might be good for you to research. I would also urge you to look at the other side of the retirement income equation, namely how much you spend. The income we generate from investments has uncertainties, but you have control over what you spend and I would see what going through your budget with a frugal eye can do for your plans.

Also just to emphasize something I noticed in your illustrative spreadsheets; growth, inflation and spending are not constant. If you do projections using a Gaussian distribution for these numbers around some historical mean is better, although hardly fantastic. It's the variability in the sequence of returns and inflation that complicates the calculations and increases risk.

michaels

jamesd said:

Linton said:

The situation of a Sequence Of Returns before the start of modelling is a general problem with using SWRs...

One can as @jamesd suggests repeat the SWR calcuolation every few years. But as pointed out..

SWR takes no recognition of the state of the market at the time.  So if you determine the SWR after 5 years of rising prices you will get the same answer as soon after a large crash.  In the first case a recalculated SWR could well be be dangerously optimistic and in the second case it will be unnecessarily pessimistic even if the SWR itself was accurate.

SWRs that take account of estimated market conditions can be used and the use of P/E in those handles much of the boom or crash issue.

The main application I'd suggest for recalculation is if using the 4% rule and markets have done well for a sustained period, such that there's a greatly increased risk of not drawing as much income as has become possible.

The assertion that the SWR could be dangerously optimistic is untrue if you accept the underlying premise, that SWRs are safe for anything likely to happen based on historic performance. They already are calculated to work just before a major crash.

The SWR toolbox spreadsheet also shows alternative 'SWRs' based on the current level of the markets compared to the all time high and the cape ratio.

OldScientist

GazzaBloom said:

OldScientist said:

Linton said:

One warning…

Do not take the word “Safe” in SWR too seriously. The calculations are based on a single set of data. Current calculated values such as a 95% success rate are based on what happened during 2 relatively short historic periods, a few years during the Great Crash and the other in the early 1970s when low market returns coincided with high inflation.  If 1 or 3 such events had taken place in the past 100 years or the two had taken place much less than 40 years apart the SWRs would be rather different.

It would be interesting to see an SWR simulation that started just before 2001 but sadly it is too early for that.

You're right that a 30 year retirement cannot, yet, be simulated, but the following graph shows a (so far) 22 year retirement for a UK retiree holding 60% UK stocks and 40% UK gilts ('all stocks') and retiring in 2000 (I've yet to update this dataset to end of 2023).

[snip for space saving]

In real terms, the portfolio has fallen to just under 40% of the original value and must now last another 8 years or so (assuming a 30 year planning period) which may (or may not) be successful. Alternatively, assuming they were 65 in 2000, an RPI annuity could now be bought with a payout rate of 9.5% (single life, 10 year guarantee, quote from moneyhelper assuming no health problems), 17.1% (single life, no guarantee) or 12.9% (joint life, no guarantee). With the former an income of 3.8% (of the original portfolio) could be purchased (leaving nothing in the portfolio) while in the second case, an inflation protected income of 4% could be purchased with about 60% of the remaining portfolio leaving some legacy, while with the joint annuity 4% income could be purchased with just under 80% of the remaining portfolio.

I assume these scenarios don't factor in a couples state pensions? As per my post above, SPs can make a big difference as a fixed income source covering a big chunk of annual living expenses when they start paying out so, how do you model a SWR on that basis if retiring before they start paying out? The SWR can surely be higher but is more exposed to sequence of returns risk in the years before the SPs start.

You are correct that the graphs above only deal with portfolio withdrawals (which is what is relevant to 'SWR', but not to overall income). There are a number of ways to approach this.

1) Use a fixed withdrawal approach from the portfolio and accept that total income will increase as each income stream (e.g., DB and SP) comes on line and the graphs given above may be of use. I note that this may not be desirable for many people.

2) As a number of people here have suggested, use a short term inflation linked gilt ladder to fill in the missing income until the sources come on stream. In other words, if there are 10 years to go until state pension, then a 10 year index linked gilt ladder is require to fill in the gap. For example, with real gilt yields of 0.8%, the payout rate is, using the pmt function in a spreadsheet, pmt(0.8%, 10,-100,0,1)=10.3%. In other words to obtain an income of (say) £12k, £115k would need to be invested in the ladder (and spent from the starting portfolio). The portfolio withdrawals can then be the same across the entire retirement planning period so the above graphs apply.

3) The portfolio withdrawals can be changed as each income source comes online. So, initial portoflio withdrawals would be to be high enough to cover the missing income and ca be reduced later. As an example, consider the 30% UK stocks, 30% US stocks, 40% UK bonds portfolio from before (note: this is held in the UK so affected by UK inflation) for which the constant 'safe' withdrawal was 3.3%. A naive starting approach might assume that the average withdrawal over 30 years should be equal to or less than 3.3%. So if 4% was required for the first 10 years, then 3.0% might be OK for the next 20 since (4*10+3*20)/30=3.33. In the following graph the WR as a function of time since retirement is shown at the 1st (near worst case), 10th, 50th (median), 90th, and 99th (near best case) percentiles.



In this case, somewhere between 1% and 10% (actually about 2%) of historical retirements failed so the naive starting approach is a bit optimistic, but not too far off (dropping both initial and subsequent withdrawals to 3.8% and 2.8% reduced the failure rate to 0) .

As another example, assume 5% for the first 10 years and 2.5% for the next 20 years.


Again, from the graph between 1% and 10% of historical retirements failed (actually about 5%) with the worst case failing after about 20 years. Again the naive approach was a bit optimistic, but not outrageously so either the initial or later withdrawals, or both, would need to be trimmed a bit.