Exploring drawdown options - take from cash vs equities first?

OldScientist

I agree very much and would go further.  IMHO (of course) all this work on SWRs and mechanistic withdrawal strategies is fine for keeping a group of academics in work but is of little relevence to the real world especially when it comes to how you actually manage your money in retirement from one year to the next.   I suspect that most people who start off with Guyton Klingor will abandon it in a few years as circumstances pan out quite unlike anything seen before in a 100 years of backtested history.

One of the dangers of these approaches is that the value of the SWR is controlled by only a very few major events.  If one of those events didnt happen or happened at some other time the SWR could be very pessimistic.  We could well find that most people most of the time will end up on their death bed with a very large sum of money, perhaps more than they started with.  

Just done a run with cfiresim: a 5% initial withdrawal rate with Guyton Klinger starting with $1M pot   It shows a 95% success rate, so not a 100% SafeWR.  However the median pension pot 25 years later is $1.26M.  Despite this the median annual withdrawal amount during the final quarter of the 25 years is less than the initial.

Much of the academic work appears to be aimed at financial advisors rather than DIY investors, so may be keeping two lots of people in work (some of whom wear two hats!). To be fair there are some useful takeaways from their work - there are some guidelines as to spending, e.g. for the UK spending an inflation adjusted 8% is highly likely to cause early portfolio exhaustion, 5% may be problematic for some retirements, 3% is probably going to last but may leave you with a lot of unspent money. Arguing over 0.1, 0.2% here and there (guilty!) is probably fruitless given the crudity of the testing. Secondly, the various dynamic withdrawal approaches at least attempt to answer the question a) what the signs that the portfolio is in trouble (or doing well), and how much shall should I aim to cut (or add).

I agree - for the USA 1965/66 sets the agenda for the worst case while 1937 (or in some cases the early 1900s) sets that of the UK. Ending up with large amounts of unused money is a problem using the fixed inflation adjusted approach.

The implementation of GK on cfiresim appears to have some problems - e.g. the prosperity rule (i.e. the one that increases withdrawals in the event of good conditions) is not triggering consistently (this can be checked by setting the cut % to zero - there's a few cases with raises, but nothing after 1930 or so) and the inflation rule is not included (which introduces other cuts - set both cut and raise to zero and the withdrawal will be fixed throughout). Personally, I think GK is too complex and, as you said, it still has a chance of failure if market conditions are unexpected.

Thrugelmir

Notepad_Phil said:

BritishInvestor said:
...
I was reading through Monevator recently where a ~70 year old admitted he was starting to struggle with some of the logic he'd used for a withdrawal strategy - this has to be taken into account, surely?

I've always thought that a key differentiator between DIY platforms could be the offering of withdrawal strategies that they would run for you - even something as simple as an automatic increase in the withdrawal amount by inflation; but as far as I can see there's nothing available from any of them.

DIY means minimal cost. If you want bells and whistles , along with monitoring of your personal account. Then be prepared to bare the cost. 

BritishInvestor

OldScientist said:

BritishInvestor said:

OldScientist said:

jamesd said:

Regrettably, ERN falsely claimed to use the Guyton-Klinger rules and have refused to correct their major mistakes after requests from many people. The result was a performance mess as they created their own poor rules instead of using the real set.  Wade Pfau produced graphs showing how something closer to real Guyton-Klinger works, though he also didn't use the Portfolio Management Rule, and also explains two of the changes from Guyton to Guyton-Klinger.

One potential drawback from G-K (like any dynamic withdrawal approach) is that the withdrawals can become small - e.g. in the example on the link to Pfau's work you provided, the income falls to about 2% in real-terms (i.e. half the initial withdrawal) so that degree of flexibility has to be built into a retirement plan. For a UK based portfolio, in the worst case, the withdrawal falls to about 1.5% (with the same conditions as for Pfau, i.e. 50/50 and initial withdrawal of 4%).

To remain on topic, the results presented in G-K's 2006 paper indicate that the portfolio management rule had a relatively small effect on the income (peaking at 6-7%) compared to the the other rules. It is interesting that McClung's backtesting (Figure 10 - see http://livingoffyourmoney.com/wp-content/uploads/2016/05/LivingOffYourOwnMoney_eBook_FirstThreeChapters.pdf), suggests that the PMR rule shows a 'modest improvement over traditional rebalancing' (I found his footnote dryly amusing).

I think it's worth digging down into the GK strategy. 2 things that concern me.

1. If intelligent people can't seem to agree on the implementation, how is the average 70 year old going to cope?
2. As you say, there is a chance that GK will give large cuts in real spending if we have unfavourable market outcomes. For example, starting at a 5.5% rate (~4.6k monthly income on £1m), I see this falling to around £1.3k p/m (1969 start). Unfortunately no free lunch.....

1) I've implemented about 10 dynamic withdrawal methods (including variations on a fixed withdrawal) and G-K was the hardest to get the details right and is not always summarised consistently elsewhere (e.g. even Pfau had a typo in one paper and cfiresim has omitted the inflation rule - correctly if you follow the last version in the 2006 G-K paper where there are two versions discussed (with and without the inflation rule). In my opinion, a paper crying out for a flow chart (Table 5 comes closest to this)!

2) G-K is not alone in making substantial cuts in poor market conditions - however, this feature does help offset the risk of complete failure (i.e. running out of money). JamesD is right that floors and ceilings can help in this regard, but then add back in the risk of portfolio exhaustion. For example, McClung's extended mortality updating failure percentage approach has a spending floor (2.25%) that was successfully tested for USA, UK, and Japan (post-1950) - but this would have failed in markets with lower safemax values than this (e.g. Italy, Germany, and Japan - including WWII). Balancing these risks (running out of money or reducing income) is a personal decision - there isn't one size fits all here.

Personally, I've ended up with a variant of an actuarial approach (VPW with planned reduced income so as to leave a legacy) since it is simple and completely pre-calculable (which means I can leave a simple spreadsheet for my OH who has no interest in any of this). The downside is that in backtesting for the UK the lowest income (for 50/50 UK stocks/bills, UK inflation, JST dataset, with no legacy plans) is about 1.6% (1937 start - see below).

"since it is simple"

 That's got to be top of the strategy requirements, IMO

BritishInvestor

Thrugelmir said:

BritishInvestor said:

OldScientist said:

jamesd said:

The falls are limited in each year to either skipping inflation increases (if inflation is above 10% and is skipped in a down year) or the amount after the annual increase minus 10% if the capital preservation rule is tripped.

That can be repeated for many years. Since GK starts higher than 4% rule it follows that if you get a sequence that is close to the worst historic one you're going to have your income cut to below 4% rule to make up for the earlier higher paying earlier years. It's unlikely but unlikely sequences can happen.

Ways to use GK and handle this differently include tools which let you set an income floor and you recognising a bad sequence and making an adjustment faster than the rules do.

State pension, its deferral, possibly combining rules and possibly buying annuities are non-GK adjuncts that can be used to manage the downside.

For a 50/50 portfolio (UK stocks/bills and UK inflation - JST dataset at https://www.macrohistory.net, 1872-2015) and 30 year retirement, all G-K rules in operation (except portfolio management) and an initial withdrawal of 5.5% the spend that was achieved as a function of retirement start year is as follows (the black line is the mean spend over 30 years, the blue lines are the maximum and minimum withdrawal within each 30 year period (all in inflation-adjusted terms).

The worst case was in 1937 and the portfolio value (in real terms) and withdrawal rate for that start year are given below

Combining the UK stocks with international ones (as would be commonly done now, if not back in 1937) and/or increasing the stock percentage in the portfolio will improve the performance a little (e.g. a 50/50 split of UK/US stocks will increase the minimum withdrawal to 1.6% as will increasing the stocks held to 80%). Importantly, the money didn't run out in any cases (the lowest ending portfolio value was about 9% of the initial value in real terms - enough for 3-4 years more). I also note that the JST dataset only includes the largest UK companies up to about 1960 which does decrease the return by about 0.4% per year compared to datasets that include small or micro companies.

As jamesd said, dynamic withdrawals are ideally combined with sources of non-portfolio income (i.e.state pension, DB pension, and annuities - probably in that order of reference). In this case, provided essential spend (however that is defined by the individual) is less than non-portfolio income+1.5% of the initial portfolio value then everything is fine. As Jamesd also mentions, retired people tend to spend less as they grow older and the spend profile in the 1937 case might not have been as unpleasant as it looks. Of course, the upside of such dynamic withdrawal approaches is that in good years you can end up withdrawing more and they respond to conditions better making it harder to run out of money (retiring in the late 70s/early 80s would have been pleasant).

I struggle to see how an approach like this is going to work in the real world (unless there are big red flashing warning signs next to it), where I would wager the typical portfolio is nothing like that used in the analysis - how did Fundsmith, Lindsell Train, anything by Baillie Gifford fare during the 70s for example?

Very different era that's not comparable. Global investing per se didn't exist. 

I'm not sure how the existence or not of global investing is relevant unless it has put an end to market cycles (both the overall markets and styles that move in and out of fashion).

If you look a the portfolios that are constructed by people who have crunched the numbers to death, they tend to have a tilt towards small-cap value & emerging markets with (maybe) a low-cost property fund thrown in. This is balanced by a high quality, short-duration bond fund. Contrast this with the large cap-growth funds mentioned above and you will see how very different this approach is, yet people flock to what is working now rather than what the long term evidence shows.

Now you can argue all you want about whether factor premiums that have historically shown excess returns will reappear after a terrible decade (which is why you've seen the funds mention above "outperform*"), but I'd be very surprised if holding funds with PE ratios >30 is going to give you an enhanced SWR compared to the studies, partly for behavioural reasons. 

*Realistically, there hasn't been evidence of genuine outperformance in the retail space for years, despite what the marketing literature might tell you.

OldScientist

BritishInvestor said:

"since it is simple"

 That's got to be top of the strategy requirements, IMO

Definitely on top of my list too - after the start of the retirement, VPW only requires a percentage calculation each year (if withdrawing monthly then a division by 12 as well) - no calculation of inflation, growth or losses, and no list of rules. If the spreadsheet I've made stops working, then it is easy enough to do with paper and a calculator (and I've left a sheet of instructions for that too). If doing that becomes beyond my OH and I, then I don't think we'll be actively spending money. Since the withdrawals could get low in inflation terms (but non-zero), there is a need for a non-portfolio source of income (e.g. state pension) that covers most of Maslow's first level of needs, which means early retirees are going to have to cover that gap by other means.

In at least an oblique nod to the OP's topic, in terms of simplicity I'm also beginning to wonder whether rebalancing (beyond taking say, 60% of withdrawals from stocks and 40% from fixed income) is actually a complexity too far.