Arborbridge wrote:For the moment, let's enjoy the artificially inflated yield for as long as it lasts. I would judged my HYP income thrown off is not easily going to catch up with that from my Arbit - by the time my dividends improve, the ITs' coffers will also be filling up enabling their dividends to carry on flowing. I will keep you posted about what happens, naturally!
Yes, it's highly likely that the ITs' dividends will carry on flowing. But the HYP income thrown off will be increasing by as much as company payouts are increasing, and the ITs' income thrown off will be increasing by less than that because a significant fraction of the companies' increased payouts will be being diverted into refilling their coffers. So I'm pretty certain the HYP income thrown off will do some significant catching up with the ITs' income thrown off - what I'm uncertain of and will look out for over the coming years (it seems likely to play out over a timescale of years rather than months) is how much it will catch up by, which could end up being the amount it's fallen behind by, less than that or more than that.
As a case that
has played out by now, the financial crisis caused HYP1's income to fall
from £5,040 to £3,187 between its years to November 2008 and November 2009, an income fall of 36.8%. Meanwhile,
CTY's dividends rose from 2.96p+2.96p+3.08p+3.08p = 12.08p for the four quarters to December 2008 to 3.08p+3.08p+3.08p+3.08p = 12.32p for the four quarters to December 2009, a rise of 2.0%. To catch up on the amount that its dividend had fallen relative to CTY, HYP1's income then needed to rise by a factor of (1+2.0%)/(1-36.8%) = 1.614 more than CTY's did, i.e. it needed to catch up by 61.4%. It successfully completed that catching up in 2017: HYP1's income in the year to November 2017 had risen from its 2009 low of £3,187 to £7,327, a 129.9% rise, and CTY's dividends in the four quarters to December 2017 had risen from 2009's 12.32p to 4.3p+4.3p+4.3p+4.3p = 17.2p, a 39.6% rise. So relative to CTY's, HYP1's income had risen by a factor of (1+129.9%)/(1+39.6%) = 1.647, i.e. a 64.7% rise relative to CTY. Or looked at another way, overall between 2008 and 2017, HYP1's income had risen 45.4% from £5,040 to £7,327, slightly more than CTY's dividend increase of 42.4% from 12.08p to 17.2p.
A couple of comments about the chart of yours that Itsallaguess reposted above:
* For the HYP income thrown off to catch up with the IT income thrown off, the increase needed is from about 5.4 to about 8.3, i.e. the HYP income-per-unit increases need to outstrip the ITs' income-per-unit increases by about 54% to completely catch up - a fairly daunting task. But the visual impression given by the chart is that outstripping by several hundred percent is required, making catching up completely a
far more daunting-looking task. That visual impression is deceptive and created by the vertical axis of the chart not starting at zero - something we discussed in the earlier thread that Itsallaguess took the chart from. No need to repeat that discussion here - I'm just alerting readers of this thread to the fact that while the amount of catching up required is large, a casual look at the chart makes it look vastly greater than it actually is. In particular, the above example of HYP1 in the financial crisis shows that although increasing by 54% relative to CTY is a fairly daunting task, it's within the bounds of possibility over the next several years. Several hundred percent would be a different story!
* Such charts comparing different portfolios do have a problem about the choice of starting date. The problem is most pronounced when the starting date is deliberately chosen to be a time when one of the portfolios is particularly high relative to the other - in this case, the comparison could be made to look even worse for HYP by deliberately choosing to start them at equal levels in June 2012 (when the HYP line on the above chart is quite a bit above the ITs line) or considerably better for HYP by similarly choosing a June 2018 starting date (when the opposite was the case). I'm pretty certain that hasn't been done here, but when the two portfolios tend to behave rather differently in the short term, creating a fair amount of volatility of the difference between them, there's a considerable chance that any rather arbitrary choice of starting date will happen to hit a starting date that significantly favours one portfolio over the other.
Basically, the only way I know to properly avoid that problem is only to compare two portfolios which have been given the same cash deposits on the same dates (each cash deposit date effectively being the starting date for part of the portfolio), and if applicable, have had the same cash withdrawals taken from them on the same dates (cash withdrawals effectively being negative cash deposits). That is of course most unlikely to happen unless the two portfolios are deliberately set up right from the start to be used for such comparisons... In particular, in the normal situation where someone wants to compare two normally-run portfolios after the fact, they're practically certain to run into this starting date problem of not really knowing whether the starting date used significantly favours one portfolio over the other, and if so, which portfolio is favoured and by how much.
This 'starting date' problem is least problematic for portfolios which quickly change their holdings in response to what shares are available on the market - basically, if what the portfolio holds on any given date is close to what it would hold if it were run by the same strategy but starting on that date rather than earlier, then any date can be treated as the starting date without distorting the comparison all that much. Of course, by the same token, it's particularly problematic for portfolios run according to LTBH strategies, since they don't change their holdings at all readily...
What one can do to avoid the problem is compare a real portfolio with a 'paper portfolio' one sets up now to run a different strategy which has been given the same cash deposits and had the same cash withdrawals taken from it as the real portfolio. The main potential problem with that is hindsight bias: if the different strategy uses human discretion in
any of its investment decisions, it becomes open to suspicions that those decisions are made differently now than they would have been at the time due to knowing more now. To avoid those suspicions, the only really effective method is for the different strategy to be entirely mechanical - and even then, if the different strategy is one of a large number of similar entirely-mechanical strategies, it's open to suspicions that hindsight is involved in the choice of exactly which one of them is used.
That makes it a good idea to make the entirely-mechanical strategy that one uses as simple as possible, to reduce the number of aspects of it that can be varied and hence the number of similar strategies as much as possible. By far the simplest entirely-mechanical strategies are single-investment, fully-invested ones, i.e. ones that use just a single investment, buying it with all the available cash whenever there is cash available and selling as much as needed of it whenever cash is needed. The only variable aspect of that is the choice of the single investment it uses...
I'd imagine that people will recognise that process of comparing a real-life portfolio against a paper portfolio run with the same cash deposits and withdrawals according to an entirely-mechanical, single-investment, fully-invested paper strategy: it's what is more commonly (and much more briefly) known as "benchmarking". For example, using the FTSE100 as a benchmark is basically such a comparison with the single investment being an idealised FTSE100 tracker. So if one wants to compare two real-life portfolios, especially if one (or both) of them is run according to an LTBH strategy, I would consider the best technique to be the somewhat indirect one of comparing each of them over its
entire history against a suitable benchmark (the same benchmark for each) and see how their outperformances or underperformances of that benchmark compare.
All of which is not to say that charts like the one of Arborbridge's that Itsallaguess reposted are useless - rather, they're useful provided one takes them with a pinch of salt. Specifically, keep their potential problems of the vertical axis not starting at zero and which starting date happens to have been chosen in mind, and think about whether any conclusion one draws from them is real or has simply been created as a result of one or both of those problems.
Gengulphus