#208865
Postby Gengulphus » March 20th, 2019, 2:03 pm
To give my own views on the number-of-holdings question: it's really a bit of a red herring in my view. To see why, imagine a 101-holding portfolio - which I think many people around here would classify as over-diversified. Now imagine that 90% of the portfolio is in one of the holdings and 0.1% in each of the other hundred holdings - still think it's over-diversified?
For me, the crucial tests are what the biggest holding is, what the biggest combined holding in a sector is, and what the biggest combined holding in a group of strong-related sectors is (*). That hypothetical portfolio is hopelessly under-diversified as far as I'm concerned because its biggest holding makes up 90% of it: no matter how the other 10% of the portfolio is split among holdings, the portfolio's performance is going to be dominated by the 90% holding, rendering it seriously vulnerable to any trouble that company gets itself into. I.e. basically, it will behave in a way much more akin to a single-holding portfolio than to a normal 100-holding portfolio.
So I would judge the company diversification of Arb's portfolio by its 3.91% maximum weighting, not by the fact that it has 37 holdings. 3.91% is about the weighting of a share in a 26-holding equally-weighted portfolio, so I'd think of it as having something more akin to 26-holding diversification than 37-share diversification. Thinking of things that way does mean that portfolios are somewhat more diversified than one calculates (the 90%-in-one-holding example above is a bit more diversified than a single-holding portfolio, but only a bit). But under-estimating one's portfolio's effective diversification in that way is a safer mistake to make than over-estimating it in the way that just counting the number of holdings does.
Though to be precise, I'd prefer to judge it primarily by income weightings rather than capital weightings - but Arb's table doesn't give either those weightings or the share yields: I could look up the yields and deduce the income weightings from them and the capital weightings, but that's a bit more work than I want to put into this post, so I'll pass on it!
As regards whether 15 or so shares are enough, I have no real reason to doubt the validity of the academic studies that say it is - but I do have reasons to doubt their applicability as far as most HYPers are concerned. I've three main reasons for doubting that:
* HYPs are LTB&H portfolios - just how long-term varies from never 'tinkering' (= selling voluntarily) to a fairly low level of doing so. The academic studies generally study "select a portfolio; buy it with equal weights; hold for a year; repeat" strategies - and while the new portfolio selected each year could be similar enough to the previous year's for the strategy to end up qualifying as a HYP strategy (i.e. not to require much buying and selling), I suspect it generally doesn't. (A subsidiary reason associated with this is the trading costs associated with that buying and selling: the more buying and selling a strategy does, the more performance it loses to them - especially to percentage-based ones like stamp duty.) The longer a strategy tends to hold on to shares, the more weighting imbalances tend to build up - so to achieve the same maximum weighting, one tends to require more holdings.
* HYP strategies are income-oriented strategies, and while not all HYPers regard actually receiving it as dividend income among their objectives, some do (that's a fact: one can reasonably hold opinions about whether it should be one of their objectives, but whether it is is a fact determined by the HYPer, not by anyone else's opinion), while the academic studies generally just look at total returns. The dividend income risks of HYP companies have clearly different statistics to their total-return risks - for instance, the dividend income produced by a HYP company drops to zero far more often (though fortunately usually much less permanently!) than its capital value. And the degree of diversification needed to make a portfolio reasonably safe does depend on those statistics - e.g. a 1-holding portfolio of gilts is very safe, a 15-holding portfolio of start-up company shares is very risky. So at least for those HYPers who care about their HYPs' dividend incomes, it's not clear that the academic studies' figures of around 15 holdings being a level of diversification that one is unlikely to significantly improve upon are fully applicable.
* HYP strategies are a general class of strategies, whose boundaries are rather 'fuzzy', and any particular HYPer's HYP strategy will typically use a variety of dividend safety checks. The entire collection of dividend safety checks used by HYPers is quite large and almost certainly, whatever particular dividend safety check one choose, one will find that quite a few HYPers don't use it. (And in some cases, don't even understand it well enough to be able to use it - for instance, no matter how much someone insists that 'culture' is an essential check, I simply don't know what it is and so cannot apply it.) The academic studies generally choose a few precisely-defined strategies to study and compare, and it's fairly unlikely that one's own HYP strategy will be a very good match to any of them. One hopes of course that the differences between one's own strategy and the strategy studied academically are small and/or in one's favour, but one has to face the risk that one is inadvertently making one's strategy riskier and in need of a greater degree of diversification.
Together, those suggest to me that it is prudent to go for a somewhat higher degree of diversification than the 15 holdings indicated by the academic studies. Not hugely more, but effective diversification of 20-25 holdings seems to me like a reasonable allowance to make for the possibility that a HYP strategy will want somewhat higher diversification - with (as indicated above) "effective" meaning that that actually means a maximum holding weighting of about 4-5%.
Similar considerations apply to sectors and to groups of related sectors for me, though with larger maximum percentages to reflect the fact that fewer of them are available and so higher percentage weightings basically have to be accepted. I've generally used 10% and 20% maximum weightings respectively in the past, as easy-to-remember round figures, though I have been shading the 20% maximum for groups of related companies downwards recently, mainly (but not only) because of becoming a bit uncomfortable about the political risk of my "utilities" group. It's now at 18%.
I haven't done a full analysis of Arb's HYP from the sector or group perspective. But a quick look at IanTHughes's table earlier in the thread says that I would be very uncomfortable about the 28.01% weighting of my "financials" group (which includes all the companies he has listed as being in the Financials industry), and within it, mildly uncomfortable about the 10.71% weighting of the Life Insurance sector. I wouldn't be even mildly uncomfortable about the 20.47% weighting of the Consumer Goods industry, because I don't regard Persimmon as having anything much in common with the other companies Arb owns that the ICB classification includes in that industry. Without its 2.54% weighting, the weighting of the industry is 17.93%, so within my limit even if I regarded it as one of my groups (which I don't currently).
(*) Note I do not use the ICB classification as anything more than a starting point - I'll happily modify aspects of it when they don't make sense to me. And my "groups of strongly-related sectors" are a long way from being ICB "supersectors" or "industries": in particular, many sectors aren't in any such group for me, and it's possible at least in principle for a sector to be in more than one such group, though that doesn't happen for any of the groups I currently see good reason to track.
Gengulphus
Last edited by
Gengulphus on March 20th, 2019, 2:15 pm, edited 1 time in total.