Arborbridge wrote:As regards investigating the tinkering matter in general, looking at capital or income changes for known sales is easy enough. But there's the other category which is much harder to analyse: the companies which you should have sold but didn't. Dod would say, for instance, that I should have sold Tesco at some point. The problem with this more complete approach to studying "whether it is nobler to tinker or not" is that it is difficult to decide what the rules will be for the experiment, so as to satisfy potential critics.
At least with known tinkerings, I have the decisions I actually made, but this second group is not only intangible but - one suspects - quite significant. e.g. Tesco, Carillion, Interserve, Pearson and several others are important cases in which many would have sold at the first of trouble, but I didn't.
Basically, I'm pretty certain that there is no even remotely easy way of investigating the tinkering matter
in general.
One can compare the results in practice of any two tinkering methods. Each of the methods is basically a set of criteria that cause one to tinker and the tinkering action it triggers one to do, and such a set can be empty, i.e. nothing triggers one to tinker and so the method is non-tinkering. Or it can just contain "Tinker if I, after due consideration, think it a good idea", i.e. it's one's own tinkering method, however informally one has defined it. Or it can contain one or more criteria that are more mechanical, such as "tinker away the entire holding if its yield drops below half the yield of the FTSE 100" or "Sell 25% of, either instead of or in addition to that sort of informal criterion.
In practice, you can compare two tinkering methods if you can get a good historical record of each. You can do that for a method if
either you have actually run it in practice and kept records
or it is completely mechanical and all the historical data needed is available and it only makes decisions on data available at the time. E.g. it's comparatively easy to obtain records of what actual dividend payments were (at least for companies that haven't vanished through being taken over or going bust and if you're not trying to go back further than about the turn of the century), somewhat harder to obtain records of what share prices were but still possible (with the same proviso), much harder (to the point where I don't know how to do it) to obtain records of what forecast dividends were. So you're much more likely to be able to get a good historical record of a method that hasn't actually been run in practice if it relies on historical yields than if it relies on forecast yields.
And of the two methods I describe above, non-tinkering is completely mechanical and doesn't rely on any historical data, while "Tinker if I, after due consideration, think it a good idea" will generally have been run in practice and it's only a matter of whether one has kept the records. And that makes it pretty easy to compare them... But any comparison of two methods won't notice any difference between them with regard to decisions that happen to be shared by the two methods. So as you say, it doesn't spot the effects of decisions not to tinker, and that's because the non-tinkering method would also have decided not to tinker.
All of that analysis may seem to be making rather a meal of it, and if it only led to that agreement with you, it would be! But what it also indicates is how one might go about studying the cases where one didn't tinker. One needs a tinkering method that made decisions to tinker where one didn't and that one can get a good historical record for. Unless one has already been running that tinkering method for a good number of years and kept records of it (not necessarily with real money - a 'paper portfolio' will do as long as it's been run honestly, not making backdated decisions, and recorded reasonably accurately), one only really has two choices. The first is to start running and recording it and wait until that good number of years has passed, and the second is to make it a completely mechanical method, based only on data that one can obtain and that was available at the time. That second one is trickier than it might sound - a common problem with it is failing to say what is done with the tinkering proceeds.
Another common problem is making it depend on data that wasn't available at the time - any such method is hindsight-based to at least some extent, and there's a "the grass is always greener on the other side of the hindsight fence" effect. And it is indeed greener - but it's illusory! And while it's fairly easy to avoid falling foul of that effect in gross cases of hindsight bias - anyone who seriously presented a study of a tinkering method that involved tinkering those shares for which tinkering turned out to be a good idea subsequently (and only those shares) would be laughed out of court - but it's easy to fall into doing it. In particular, every time one thinks "if only I'd tinkered Carillion, I'd be better off today" is very close to doing it in a minor way - they don't quite do so if all they conclude is "therefore tinkering Carillion would have been a good idea". But if they go further and conclude "therefore tinkering is a good idea" then they've stepped over the hindsight line, the hindsight involved being picking out Carillion as the share to study. And the same goes if they also study Tesco, Interserve and Pearson and find that similar (though not so severe) bad things have happened - they're still exhibiting hindsight in their choice of shares to study. One needs to study
all one's holdings to avoid that - which unfortunately does make the study considerably more time/effort-consuming...
One also needs to avoid hindsight in one's choice of tinkering method. One whose only criterion was "sell completely if the company's name is Carillion" would obviously have improved a lot of HYPs' performances, and not spoilt any of them - and is not a plausible rule to have come up with without hindsight. Somewhat more plausibly, one that picks out a particular stringent set of financial conditions that all applied to Carillion, and that seems very unlikely to have all applied to any other company in one's HYP, and said to tinker if a company exhibited all of those conditions, would just be a disguised way of saying the same thing. So beware of tinkering criteria that are complex combinations of conditions or are simple conditions that are so stringent that they only turn out to have applied to a very few holdings in one's portfolio - one is in danger of inadvertent hindsight bias simply caused by looking for criteria that would have helped one avoid one particular share that is known to have gone very wrong.
That bias is very difficult to avoid when looking for lessons to learn after events like Carillion's collapse, and probably impossible to avoid completely. But do treat any such 'lessons' with great caution: they're nothing more than ideas until they've had a good amount of confirmation from other examples. And in that context, beware of the data-mining effect: if you test a lot of hypotheses, some of them will receive a good deal of confirmation by pure chance. As a somewhat silly example of that, my HYP has contained five shares whose company names started with 'C' in the last decade: Carillion, Centrica, Carpetright, Capital Shopping Centres and Capital & Counties (of which Centrica is the only one I still own). Of those, Carillion produced the awful result we all know about, Centrica has done pretty poorly as again I think we all know, and Carpetright produced rather mediocre results for me - its dividends weren't bad for the first five years I owned it (2003-2008), but collapsed in the next three before I tinkered it away completely in 2011 (*) and I lost somewhat over a year's dividends to the 6%ish capital loss I made on it overall. The remaining two are part of a complex combination of an original holding of Liberty International demerging into them and Capital Shopping Centres subsequently renaming itself Intu, and I can't easily access summary figures about them, but I'm pretty certain that the overall picture up to the time I tinkered them away is pretty mediocre.
So basically, a criterion to tinker away (or not buy in the first place) companies whose names start with 'C' would pretty clearly have improved my HYP's performance. But despite having a fair amount of confirmation - five examples of when a particular tinkering rule would apply in 10 years of a HYP's history isn't bad - that conclusion is to be mistrusted, because it's an arbitrary criterion that lacks any likely investment logic behind it and has lots of possible variations. It's almost certainly due to the simple data-mining effect that if the records of all shares in my HYP were checked against starting with each letter of the alphabet, it's very likely that a few letters will show decidedly above-average performance and a few decidedly below-average performance by sheer chance.
Finally, the above means that while comparing one's own tinkering method with non-tinkering is not all that difficult, more generally comparing one's own tinkering method with other tinkering methods is probably at least an order of magnitude more work, and it's trickier to avoid all the traps. That needs at least another order of magnitude and various extra trickiness added for a study of tinkering in general! I.e. as conclusions to aim for, "my tinkering method is better than not tinkering", "my tinkering method is good" and "tinkering in general is good"
very rapidly increase in practical difficulty. And personally, I've done a few studies of the first of them and only reached the "my tinkering method is better than not tinkering" conclusion on a qualitative personal-preference basis, not a quantitative financial-outcomes basis. Basically, I can't see any significant difference in the financial outcomes, but I'm personally more comfortable with the portfolio not containing the hugely-above-average-size and more-admin-than-they're-worth holdings that my tinkering method is designed to fix (by top-slicing and tinkering away entirely respectively. That'll do for me as far as running my HYP in practice is concerned, so I'm not about to put a lot more work into studying such questions.
(*) Which I see with hindsight has turned out to be a fairly good tinkering decision, though a year earlier
or a year later would have been quite a bit better! And don't take that as a claim to be good at making tinkering decisions - counterbalance it is your mind with the fact that I
didn't tinker Carillion away!
Gengulphus
