How Jon Guyton's firm does drawdown (Guyton-Klinger rules JG)

jamesd

Thrugelmir said:

jamesd said:

No, because I understood the way things worked in the original Guyton and Guyton-Klinger research.

Then you must have access to data that as far as I am aware isn't in the public domain. Particularly as you've modified the model US portfolio considerably. Your inclusion of P2P as a substitute for Government Bonds and VCT's as a substitute high quality equities. Bears no correlation or accord with any research released by GK.  

What does my choice of investments have to do with that blog, the research papers or the example calculation, none of which use my investments?

Though there is one relevant aspect: the blog used a different investment mixture from the Guyton and Guyton and Klinger research and it's a mixture that can be expected to have lower returns. But that's not very significant compared to them getting the drawdown rules wrong.

michaels

I thought the VCTs were just a wrapper to avoid tax rather than a specific investment vehicle? 

Isn't P2P equivalent to consumer/business credit backed bonds?

jamesd

VCTs are a combination of tax wrapper and investment type.

Investment trusts are a form of fund that is bought and sold on the main UK stock market, like ordinary shares. VCTs are like investment trusts that specialise in smaller and younger businesses than even small cap or micro cap specialist funds, though there is a subset that also or mostly invests in shares on the UK AIM stock market. Because these young companies more often have problems the government offers various tax breaks: 30% refund of the purchase price (but not more than your income tax due in the year), tax exempt dividends and no CGT.

While all VCTs must invest in that type of company, the companies range from startups to those getting big enough for the main stock market so there are different levels of risk and potential reward that can be picked. You can see an illustration of that in this report about the Draper Esprit VCT https://www.draperespritvct.com/wp-content/uploads/2021/02/Tax_Efficient_Review_Issue_422_Draper_Esprit-VCT_2021.pdf . Tax Efficient Review is a well known analyser of VCTs.

P2P varies from unsecured consumer credit through small business loans secured on assets of the business to property development or bridging loans secured on the property. Personal guarantees using the assets of the individuals involved are often required in the business types. There's also pawn, done by Unbolted. Different characteristics and you can choose which you're comfortable with and in the business lending often which specific deals. So yes, often like bonds, but you'll never find anything that would be classed as investment grade by a bond ratings agency and held in most bond funds. Think more of sub-prime bonds.

gm0

I think that a broader issue is with applying intuitiion or indeed any "mix n match" approach.  Changing a parameter, a threshold, a rule or a specific asset in the mix that you happen not to like. 

Applies in many of the variable income methods. Unless you then go and rerun the backtesting and/or MC simulations again to observe the behaviour of your newly modified approach. You don't know if it now does what it said on the tin in the original version.   And it may well not - for a quite small change.  So the risk profile (based on known market risk) has subtly changed by an unknown amount.  Some literature and competent sets of tests run by others may give you confidence for a "range" of values - such as with setting a minimum income floor.

Of course some methods and decision parameters have themselves been "trained on the backtesting data" i.e they have "magic" numbers in - such as the thresholds to sell what asset type in Prime Harvesting as an example or the 10% here i.e. this is the "optimised" value for the test data sets explored.  Is it best for the future. Who knows.  Is it a local maximum in the data sets used. Yes.  Minor variances in testing and reproducibility arise all the time with bugs, maths libraries, rounding, interpretation of original articles, squeezing things into an existing model at acceptable effort.  It's all fine provided - you show working - and a level of caution about extrapolating critical conclusions where you test a variation and find a "delta" from the original "claims".  The delta/shortfall/failure of reproduction may be your own - not the original authors' so a little humility against fate is in order.

bostonerimus

jamesd said:

Yes, my own pension withdrawing is largely governed by tax planning: look to withdraw full basic rate band and buy enough VCTs to make income tax due for the year about nil. Pretty much totally decoupled from funding regular spending.

If using Guyton-Klinger you should do what you did and take all dividends and fund distributions as cash, where it'll be drawn from to pay income. This means using fund income units where available.

Even though I have had lower than my normal equity percentage the price rises have been nice. I'm back up to about 59% equities, 69% including VCTs. The VCTs are delivering some nice tax exempt income, about 6.6% of total share price (ignoring recent purchases).

Just before retirement I decided to follow the "rising equity glide path" and to do that I stopped rebalancing and when I invested money it went into equities so that from a pre-retirement 60/40 allocation I am now at 80/20. Even given the modeling of people like Pfau that's a pretty risky allocation and a year ago I watched as stocks sank and in fact did do some rebalancing, but stuck to my rule and sold bonds to buy equities and now that stocks are at all time highs again that's skewed the allocation even further towards equities. The value of the portfolio is up, but the volatility is also. I could not be so sanguine about that volatility if I didn't have the regular income from the rent and the DB pensions that are somewhat decoupled from the stock and bond markets. My tax planning is similar to yours and I'm using withdrawals to fill up my tax band, and putting the money into tax efficient investments and also gifting some to my grand nieces as part of a long term estate plan.

Bravepants

bostonerimus said:

jamesd said:

Yes, my own pension withdrawing is largely governed by tax planning: look to withdraw full basic rate band and buy enough VCTs to make income tax due for the year about nil. Pretty much totally decoupled from funding regular spending.

If using Guyton-Klinger you should do what you did and take all dividends and fund distributions as cash, where it'll be drawn from to pay income. This means using fund income units where available.

Even though I have had lower than my normal equity percentage the price rises have been nice. I'm back up to about 59% equities, 69% including VCTs. The VCTs are delivering some nice tax exempt income, about 6.6% of total share price (ignoring recent purchases).

Just before retirement I decided to follow the "rising equity glide path" and to do that I stopped rebalancing and when I invested money it went into equities so that from a pre-retirement 60/40 allocation I am now at 80/20. Even given the modeling of people like Pfau that's a pretty risky allocation and a year ago I watched as stocks sank and in fact did do some rebalancing, but stuck to my rule and sold bonds to buy equities and now that stocks are at all time highs again that's skewed the allocation even further towards equities. The value of the portfolio is up, but the volatility is also. I could not be so sanguine about that volatility if I didn't have the regular income from the rent and the DB pensions that are somewhat decoupled from the stock and bond markets. My tax planning is similar to yours and I'm using withdrawals to fill up my tax band, and putting the money into tax efficient investments and also gifting some to my grand nieces as part of a long term estate plan.

This is what I'm doing with my ISA for the exact same reasons. When I eventually retire I will likely continue investing, but probably in a 100% global equity fund, shifting the allocation now that I'm more certain that I won't actually need to use the money.

Thrugelmir

jamesd said:

Thrugelmir said:

jamesd said:

No, because I understood the way things worked in the original Guyton and Guyton-Klinger research.

Then you must have access to data that as far as I am aware isn't in the public domain. Particularly as you've modified the model US portfolio considerably. Your inclusion of P2P as a substitute for Government Bonds and VCT's as a substitute high quality equities. Bears no correlation or accord with any research released by GK.  

What does my choice of investments have to do with that blog, the research papers or the example calculation, none of which use my investments?

Why write copious amounts on the rules without any solid foundations to back up your assertions?  Akin to John Bogle reading a research paper on passively investing in the S&P 500 and saying let's apply the same principle to emerging markets, and market a fund to US investors. 

jamesd

Thrugelmir said:

Why write copious amounts on the rules without any solid foundations to back up your assertions?

The solid research required is reading both that paper and the blog and noticing the differences. I even provided you with a crib sheet so finding the differences shouldn't be hard.

For example, when the real rules say cut by ten percent once a year and the blog says do it once a month it's not hard to notice why a major G-K objective, smoothing of income, doesn't work properly. When the real rules cut by ten percent after a fall but the blog cuts by the fall percentage and ten percent it's even easier to understand dramatic income cuts in what the blog does.

Those things apply whatever market is involved because it's not about the investments but getting the rules wrong.

michaels

gm0 said:

I think that a broader issue is with applying intuitiion or indeed any "mix n match" approach.  Changing a parameter, a threshold, a rule or a specific asset in the mix that you happen not to like. 

Applies in many of the variable income methods. Unless you then go and rerun the backtesting and/or MC simulations again to observe the behaviour of your newly modified approach. You don't know if it now does what it said on the tin in the original version.   And it may well not - for a quite small change.  So the risk profile (based on known market risk) has subtly changed by an unknown amount.  Some literature and competent sets of tests run by others may give you confidence for a "range" of values - such as with setting a minimum income floor.

Of course some methods and decision parameters have themselves been "trained on the backtesting data" i.e they have "magic" numbers in - such as the thresholds to sell what asset type in Prime Harvesting as an example or the 10% here i.e. this is the "optimised" value for the test data sets explored.  Is it best for the future. Who knows.  Is it a local maximum in the data sets used. Yes.  Minor variances in testing and reproducibility arise all the time with bugs, maths libraries, rounding, interpretation of original articles, squeezing things into an existing model at acceptable effort.  It's all fine provided - you show working - and a level of caution about extrapolating critical conclusions where you test a variation and find a "delta" from the original "claims".  The delta/shortfall/failure of reproduction may be your own - not the original authors' so a little humility against fate is in order.

Generally the tools do at least support back testing different mixes of stocks, bonds and cash, but then I am probably back testing a global stocks index tracker against the S&P 500 and who knows how my bond portfolio compares to the model one or whether my return on cash will be better or worse than the model data series.  Given the huge level of uncertainty in the modelling anyway, I don't think stretching the definition of bonds to include P2P will make too much odds.  If you don't feel like your asset classes have a historical analog perhaps using known volatilities and montecarlo testing is more appropriate although this approach generally ignores observed historical correlations (or lack of) between assets and the tendency to 'revert to mean' that has historically been seen in assets prices (although of course there is no guarantee this is a 'law')

gm0

The fact that any approach or method checks out and has undergone backtesting (with a variety of international data sets) and some MC random return testing for a range of spreads (volatility) gives some confidence about it's behaviour - essentially around lack of surprises for the "normal" or stressed markets simulated.  Historic market cohort backtesting covers the sinusoidal over/under performance + mean reversion to long term averages nicely.  Cohort analysis shows up the impact of SORR on the "unlucky" cohorts/start dates as these start to appear as depletion failures at the edge case of high volatility + sequence vs long term return (MC) in deaccumulation or at MSWR (100%) in backtests.

All the spurious accuracy and detail is helpful but not predictive.  It provides very little about the future other than a level of confidence about the mechanism being considered remaining basically sound for a range of conditions already encountered and for "normal" distributions/random returns (or whatever shape was simulated mathematically) to some chosen level of volatility.

This is still an advance on an "untested method".

The problem is really one of definition.  "Better" as defined by those risk managing certain issues which lead to a Permanent Portfolio getting selected is a different "better" to an income maximisation "better" in "a future like known markets" or to an ethical holdings slant or indeed any other set of investment statement principles which alter the priorities.