Smart Portfolios - a 4th generation review

[Newroad]

Morning All.

I finally got around to buying Smart Portfolios by Robert Carver. I wish I had done so much closer to its 2017 publication date, but que sera. More on it further below, but if you fancy yourself as (or simply enjoy) stock picking and/or are less enamoured with diversification, it's probably less useful to you. Conversely, if your view is more top down (as opposed to bottom up) and/or you're relatively more skeptical about the ability of yourself or fund managers to beat the market, then it will likely suit.

Some background first. My/our investing has gone through some clear phases

1. Company pension mandatory contributions complemented by regular savings into single product investment trusts
2. The single product investment trusts morphed into simple broking accounts (with II) holding a 60% or so residual of the investment trust and a 40% holding in a bond investment trusts
3. The "discovery" of ETF's and splitting those equity and bonds holdings 50% passive (ETF) and 50% active (investment trust) within
4. Calibrating the above 60/40 ratio a little, moving towards around 80/20 (E/B) for the kids JISA's and around 70/30 for the our ISA's and SIPP's

So, I am now on the verge of a "5th generation" further calibration, where there are some key pages and concepts in this book that are relevant.

Concept A (p40): (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Standard Deviation)

If anyone can tell me a free online source when I can get at least two of the above three for a listed stock or ETF, I would be grateful (as it will help me tailor some upcoming calibrations even more accurately). Note, by Geometric/Arithmetic Mean above, the intent is Geometric/Arithmetic Mean Total Return.

Concept B (p201): There are various implied cash weightings (derived from suggested risk weightings) for various asset classes and degrees of portfolio complexity (e.g. do you include alternative investments as well as equities and bonds). However, in its simplest form, which mine approximates (i.e. global and diverse, but only equities and bonds), if you

Want to optimise Geometric Mean, your split should be 78%/22%
Want to optimise your Sharpe Ratio, your split should be 29%/71%
Want a suggested compromise suited for many, your split might be 48%/52%

This forum contains many members advocating 100% equities. I'll leave you to conduct your own analysis (or read the book) of why Carver asserts that any more than about 80% increases only risk but not (geometric mean total) return*, but to me the basis is fairly self-evident - start by considering the effect of periodic large drops in equities (and what a 20% bond holding going into those might allow) even if they later fully recover and more.

Concept C (p283): Internal bond weightings for UK investors. In short, he suggests 75%/25% split between developed emerging markets. Within developed, he suggests 90%/10% normal/inflation linked. With "normal", the most relevant to me, he suggests 40%/30%/30% government/corporate/high yield risk weighting, which critically translates to circa 44.4%/30.7%/25.0% cash weightings.

So, what does this all mean for me?

My instinctive equity/bond ratios appear to have been reasonable - the kids were very close to the optimal for return. I'm probably now going to calibrate an age based reduction from optimal geometric mean to then compromise ratio, assume 100 years, (i.e. at birth, investments would have been 78% equities and if and when we turn 100, investments would remain 48% equities). If I can get better risk (volatility) information as mentioned earlier, I might tweak the VWRL/VWRP vs MWY/ATST/FCIT ratios, most likely leading to a slightly higher ratio from the former group.

The current active vs passive bond ratios (risk) materially over-weights the actives, so I need to correct that. This is something that I knew in my heart, but it's good to have positive statistical confirmation.

There's a lot more in the book, but I hope the above is a useful summary. If you buy from Harriman House, you get the e-book free as well. However, you'll pay rack rate by default - so wait until one of the sales days (I bought it recently on World Book Day for 25% off and free delivery).

Regards, Newroad

* If you have another preference, that's fine and down to you - and out of the scope of this discussion.

[Hariseldon58]

The relationship between Geometric and Arithmetic means sparked some interest, long ago I read mathematics at Oxford and my gut reaction was it depends….

I have found this interesting paper https://img1.wsimg.com/blobby/go/4d4cf479-d9ad-4bcc-86c9-0ff104d9a3d2/downloads/1br845gs1_784971.pdf

Which provides some data that you requested by DMS, for different countries plus three more approximations and the calculations.

The approximation tends to underestimate the geometric returns, it’s in the right ball park…most of the time, but occasionally its out by quite a lot,( My gut instinct of it depends..)

[Hariseldon58]

Morning All.

Concept B (p201): There are various implied cash weightings (derived from suggested risk weightings) for various asset classes and degrees of portfolio complexity (e.g. do you include alternative investments as well as equities and bonds). However, in its simplest form, which mine approximates (i.e. global and diverse, but only equities and bonds), if you

Want to optimise Geometric Mean, your split should be 78%/22%
Want to optimise your Sharpe Ratio, your split should be 29%/71%
Want a suggested compromise suited for many, your split might be 48%/52%

.[/i]

Anything that gives you a mathematical answer such as 78%/22% based on previous data for what you should do going forward is BS. It may be interesting to speculate on what to do going forward and the past is illuminating, but the future has so many more variations and uncertainties , hidden relationships, that change and may be reflexive.

Rebalancing is more attractive as a means of controlling risk rather than enhancing returns, it may well do so but don’t rely on it.

If one asset has greater returns but more volatility, then it’s more likely that the more you have of it, the higher the returns. It’s likelihood rather certainty, ( I have done some modelling that did not dispel that) The big issue is achieving your investment returns, undershooting is bad news but exceeding your required returns is much less useful… I wouldn’t disagree with the conclusions you draw re % of holdings but I would accept them as rules of thumb rather than a mathematical justification.

We all strive for certainty, the idea of using mathematics is attractive given my background, but there is little likelihood of it being more than one factor in making a judgement call.

EG the mix between Developed and Emerging, a previous experience looks great until it doesn’t. The well cared for Turkey wakes up one morning in December not knowing that the day will end badly….

The current dreadful situation in Ukraine reminds us that events occur that may have ramifications beyond the parties concerned and affect the asset class more broadly.

[Newroad]

Hi HariSeldon58.

What I have described in this post is but a slightly simplified and limited summary, with a natural bias to considering the factors most pertinent to me. If you have an interest, whether it be practical or in the underlying mathematics, you may well enjoy it (even if only to be given an opportunity to refute the arguments presented). Were you to do so, I think you'd be OK with the (absence of) level of certainty provided behind the numbers - and that he explains their statistical underpinning - including the modelling using historical data but not assuming that history will repeat.

Rebalancing was and remains one of the things I do - there is a section of the book on this that I have not yet digested - so I may have more to add on that later.

On the mixes, it goes down to much greater depth (and what size of portfolio might justify holding that level of granularity). I haven't myself worried to much about that as I don't at this point need to - I'm happy with my very limited number of holdings. However, to give you a flavour for larger portfolios, for geometric mean total return optimisation, he calculates/suggests

Equities: 70.2%
Equity-like alternatives: 3.7%
Bonds: 20.7%
Bond-like alternatives: 1.1%
Genuine alternatives: 4.3%

The explanation of the types of alternative is not simple - one for reading the book if you're interest and/or have a large enough portfolio for it to be relevant.

Regards, Newroad

[Hariseldon58]

@newroad

It’s very easy to spend a lot of time on Portfolio Asset allocation and I have in the past spent too much time on the detail and excessive granularity of my choices.

I have returned to my simplified portfolio which is mostly passive, a bias to USA and around 18% in non equities ( mostly US TIPs and commercial property, (direct holdings and equity))

I am not against rebalancing but as risk management tool primarily and in bad times of course there is a chance to pick up bargains, however the bonds/cash that allow the purchase of bargains often comes with a cost of poor returns whilst you wait… At times of great volatility eg summer 2020 there are some short term trades available that can be profitable, at the risk that a short term trade can become a long term trade if markets go against you !

Thank you for your reply and you must follow the path that you are comfortable with and fulfils your need. I will keep an eye out for the book on my Kindle buy list.

[Newroad]

Thanks iwm.eu.com

I'll fully digest the book first, then further consider some of the potential issues you raise.

Big picture, however, it's best to be approximately right rather than specifically wrong - so I'm hoping this will further my personal attempt to achieve that aim.

Regards, Newroad

[globalarbtrader]

It's a very weird experience reading this thread, which I found through a google alert. I first got interested in finance in 1999, reading the whole fool.co.uk message boards under my original username 'trojantrader'. If you had told me then that I'd be reading a book review of one of my own books in the same place, or at least on it's spiritual successor, I'd have been more than a little gobsmacked. I even note that on another thread https://www.lemonfool.co.uk/viewtopic.php?f=15&t=33704&start=20 I'm being mentioned in the same sentence as the legendary 'pyad', who gets a footnote on page 347 of Smart Portfolios:

If you follow this process with the minimum amount of capital you’ll end up with one share in each sector, each the highest yielding. This kind of portfolio construction is by no means original. I first came across it on the Motley Fool’s UK website fool.co.uk where the idea was proposed by Stephen Bland, who named this approach the ‘High Yield Portfolio’ (HYP). An important difference between this and my approach is that Stephen advocates putting everything you have got into this HYP strategy, which leaves you exposed to a single country (the UK) and asset class (equities). I personally think it’s extremely important to enforce geographic and asset class diversification before considering stock selection.

Sadly I never quite made the move across to lemonfool (I used to post annually on the portfolio review board, but that has now fallen into abeyance), as I spend more time thinking about trading than investing.

A very quick review of some of the points raised in this thread :

- Approximate AMS vs GMS, "The approximation tends to underestimate the geometric returns" actually the approximation holds for Gaussian normal returns, and since most financial assets are negative skew with fat tails (high kurtosis) the approximation will normally over state the geometric return (here is a good explanation, notice the sign in the Taylor expansion on the third moment (skew) is positive and the kurtosis is negative https://math.stackexchange.com/questions/2787355/approximation-of-geometric-mean)

- "Anything that gives you a mathematical answer such as 78%/22% based on previous data for what you should do going forward is BS" True, which is why I spend a lot of time considering the impact of secular trends that are unlikely to be repeated and the effect of uncertainty. And related to this is:

- "It’s very easy to spend a lot of time on Portfolio Asset allocation" Indeed. Since I wrote Smart Portfolios I only spend about 10 minutes a year on asset allocation. I hope I get across the message that you shouldn't panic about whether you should be at 78/22 or 80/22 or 82/18... more that you should stick to an allocation once you've decided on it with a full understanding of the implications.

- "it's best to be approximately right rather than specifically wrong" This is very much in the spirit of the book.

- "as you've indicated above there is a mathematical reason why 100% equity allocation is not optimal, and he never goes into this" Actually it's not a mathematical truth, it will depend on the relationship between correlation, standard deviation, and relative return. If bonds and equities are fairly highly correlated, and the sharpe ratio of bonds is a lot lower than stocks, then the maximum GM would be achieved by 100% in stocks. I guess the point is I think it's very heroic to assume that you'd get conditions that would produce this result (This article written by someone who use to work for me, and you might find it interesting https://riskyfinance.com/wp-content/uploads/2019/07/The_Sharpe_Ratio_Ratio_ThomasSmith-FINAL.pdf)

- "he makes little of sequence risk and how to changes the normal distribution and therefore the probability of failure which is good feedback for those needing income from a portfolio" In fact drawing income was on the original outline I sent to the publisher, but I dropped it due to lack of space. To deal with this properly in the same framework considering uncertainty, with a full explanation, would have probably required another couple of hundred pages. However I would say that thinking about things in a geometric return / uncertainty framework means you are less exposed to sequence risk. A higher AM 100% stocks portfolio would be highly exposed to sequence risk compared to say 80/20.

- "he is too soft on Exchange Traded funds especially the smart ETFs in terms of undisclosed costs and platform fees, and therefore the message of when to drop funds and all their fees altogether kind of fades from sight - see trackingdifferences.com" Perhaps I could have been harder, but this is prety unequivocal:

To summarise, I’m fairly sceptical of more complicated smart beta weighting approaches, and as they normally cost significantly more than both market cap and equal weighted I wouldn’t recommend them

- "he does not quite make enough of the efficient frontier and how to calculate annually the best return the markets offered for the risk you selected and therefore give you the critical feedback on whether you have been efficient." I spend quite a lot of time exploring the consequences of being in different places on the efficient frontier, but to be honest I'm not sure I think the sort of annual exercise you advocate would produce much value.

Globalarbtrader AKA Robert Carver

[Newroad]

Hi Robert (aka "GlobalArbTrader" ).

The coincidences continue! I started this thread and, independent of and unaware that you had found it, wrote the following tonight

https://www.lemonfool.co.uk/viewtopic.php?f=31&t=34057

Thank you for commenting on this one - I hope here and elsewhere I haven't misrepresented what you wrote in your book - my apology if and where I haven't. I've tried to capture uncertainty and your underlying thinking, in so far as I understand it, where appropriate.

Regards, Newroad

[hiriskpaul]

Concept A (p40): (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Standard Deviation)

Should be (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Variance)

[Newroad]

Quite right, HiRiskPaul.

I had mistaken an indice for a footnote - it was the standard deviation squared, i.e. the variance, as you suggest

Regards, Newroad

[Kantwebefriends]

Concept A (p40): (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Standard Deviation)

Should be (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Variance)

How can that be? The variance has units of item^2 while the means have units of item^1.

[Kantwebefriends]

I spend quite a lot of time exploring the consequences of being in different places on the efficient frontier

Harry Markowitz invented modern portfolio theory, with its efficient frontier and so on. One day somebody asked him how he allocated his own portfolio. His answer was 50% stocks, 50% Treasuries.

Or so the internet claims. Apparently that's a bit inaccurate.
https://jasonzweig.com/what-harry-markowitz-meant/

[hiriskpaul]

Concept A (p40): (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Standard Deviation)

Should be (approximately) Geometric Mean = Arithmetic Mean - (0.5 * Variance)

How can that be? The variance has units of item^2 while the means have units of item^1.

Rates of return are dimensionless. Interesting observation though!

A better approximation, especially for higher volatility, is

G = (1+A)/sqrt(1+V/(1+A)^2) - 1

which looks a little better from a units point of view. If you do Taylor expansions on the above you get to G = A - V/2 + higher order terms.

[hiriskpaul]

ps, just noticed the approximation G = (1+A)/sqrt(1+V/(1+A)^2) - 1 is discussed in the link to the paper provided by hariseldon (Equation A4). Well worth a read if this sort of thing interests you.

In real markets of course V is not constant and market returns can sometimes drift a long way from being lognormally distributed. Other than those minor quibbles, all good stuff!