REITS

[Dod101]

Following the annual results for SEGRO reported on another Board by tjh (thanks!) I have been looking at other REITS. I hold SEGRO and am quite happy with it although the yield is modest at around 2.9%. In the Annual Report of A & J Mucklow, I came across some figures for the total shareholder return over the last ten years of a selection of REITS. Makes for very interesting reading, especially for those still in the building phase.

TSR to 30 June 2017

Shaftesbury 174.3%
Primary Health Properties 106%
Big Yellow Group 99.7%
A & J Mucklow 99.5%
Derwent London 76.5%
Great Portland Estates 51.8%
McKay Securities 3.0%

All others are negative including SEGRO, British Land and Land Securities, all quite usual HYP candidates. As for yield, in the above list, only PHP (4.5%) and A & J Mucklow (4.3%) are over 4% and most are under 3%. B Land yields 4.6%, Land Securities 4.5% and SEGRO a mere 2.9%. This is over a 10 year period so there will be some years when TSR will be on the positive side but as we are mostly LTBH investors, I think a 10 year view is helpful. If I were a builder, I would not be holding B Land! I am actually considering switching my B Land holding for Mucklow. I like it; a small company where the founding family still has a near 20% stake.

Dod

[absolutezero]

It's one I hold and am happy to for now, but I like PHP. Decent yield and virtually guaranteed rent.
Slight political concerns as it might be seen as leeching money from "our NHS" but not in the league of the utilities.

[Arborbridge]

Following the annual results for SEGRO reported on another Board by tjh (thanks!) I have been looking at other REITS. I hold SEGRO and am quite happy with it although the yield is modest at around 2.9%. In the Annual Report of A & J Mucklow, I came across some figures for the total shareholder return over the last ten years of a selection of REITS. Makes for very interesting reading, especially for those still in the building phase.

TSR to 30 June 2017

Shaftesbury 174.3%
Primary Health Properties 106%
Big Yellow Group 99.7%
A & J Mucklow 99.5%
Derwent London 76.5%
Great Portland Estates 51.8%
McKay Securities 3.0%

All others are negative including SEGRO, British Land and Land Securities, all quite usual HYP candidates. As for yield, in the above list, only PHP (4.5%) and A & J Mucklow (4.3%) are over 4% and most are under 3%. B Land yields 4.6%, Land Securities 4.5% and SEGRO a mere 2.9%. This is over a 10 year period so there will be some years when TSR will be on the positive side but as we are mostly LTBH investors, I think a 10 year view is helpful. If I were a builder, I would not be holding B Land! I am actually considering switching my B Land holding for Mucklow. I like it; a small company where the founding family still has a near 20% stake.

Dod

Looking at the share price charts on HL, Bland does look much the laggard, I agree. Mucklow is good, but Shaftesbury is by far the winner. This isn't taking into account the dividends re-invested, though - nor that we are after income!

[Dod101]

It's one I hold and am happy to for now, but I like PHP. Decent yield and virtually guaranteed rent.
Slight political concerns as it might be seen as leeching money from "our NHS" but not in the league of the utilities.

I hold PHP and had not thought of a political angle there as well! Anyway I will unusually (for me!) invoke strategic ignorance and hang on to it.

I also hold B Land and SEGRO. SEGRO I like because it is not in the usual big flash office blocks of Central London, but rather workaday industrial units around Heathrow, inter alia, as befits their original name, Slough Estates. Just wondering about B Land. I like the income and Mucklow is not far from the same. Certainly for TSR it would be a no brainer and when I used to invest mainly for capital growth, I gave up on B Land and land Securities, on account of their cyclical nature for NAV.

Dod

[Gengulphus]

TSR to 30 June 2017

Shaftesbury 174.3%
Primary Health Properties 106%
Big Yellow Group 99.7%
A & J Mucklow 99.5%
Derwent London 76.5%
Great Portland Estates 51.8%
McKay Securities 3.0%

All others are negative including SEGRO, British Land and Land Securities, all quite usual HYP candidates. As for yield, in the above list, only PHP (4.5%) and A & J Mucklow (4.3%) are over 4% and most are under 3%. B Land yields 4.6%, Land Securities 4.5% and SEGRO a mere 2.9%. This is over a 10 year period so there will be some years when TSR will be on the positive side but as we are mostly LTBH investors, I think a 10 year view is helpful. ...

I would take care about that. A 10-year view will often be very reasonable, but there are exceptions. And some quick investigations using share price charts suggest that this is one of those exceptions - only suggest, not prove, since share price figures differ from Total Shareholder Return figures both by ignoring dividends received and by possibly ignoring share count changes caused by corporate actions such as share splits and consolidations and rights issues. But in this case, the suggestion is strong enough that I'd be surprised if it could be anywhere near adequately explained by dividend differences, even over 10 years, and I know the share price charts I used do adjust for share splits and consolidations. I don't know for certain about rights issues, and there are some of them involved, but the share count differences caused by rights issues aren't generally big enough to adequately explain the suggestion either... I could investigate further to get a more definite conclusion about whether the suggestion is correct, and might well do so if I get enough time to do so - but it would depend on putting together TSR data over a range of dates for multiple companies, which is not a quick job by the techniques I know about and trust. I'm definitely not going to get around to that in the near future, as my available number-crunching effort in at least the next week or two is fully needed for more important (to me!) purposes. So I'm reporting the suspicion now, in case anyone else wants to investigate it more deeply and to get around to those investigations more quickly than me.

So on to what the suspicion is: 10-year-view figures can be quite vulnerable to variability of the data used that specifically relates to either the starting date or the ending date of the period used, and much less so to variability of the data taken from a range of dates within the period. For example, the CAGR of dividends paid over the last 10 years will be very sensitive to whether dividend cuts fell just inside or just outside the 10-year period - i.e. dividend cuts that were close to the starting and ending dates of the period. That's because the CAGR is calculated only from the initial and final annual dividends. In contrast, the total dividends paid in the 10-year period are calculated from the annual dividends for all 10 years, and are a lot less sensitive to any one dividend cut (though unfortunately they quite simply aren't a measure of dividend growth, and so not any sort of substitute for the CAGR of the dividends).

And 10 years ago was early 2008, which was in the lead up to the financial crisis. The markets were getting very nervous about what might be coming, without any clear consensus that it was coming, and that resulted both in high volatility of share prices (i.e. high variability of the same share's price over short periods of time) and high variability of how different shares' prices changed on the same dates, caused by different shares being hit by specific news, rumours, analyst opinions, etc, on different days. So there's good reason to suspect that the 10-year-view comparison of different shares as seen now might vary considerably for quite small changes in exactly when "now" is, due to events that were recent 10 years ago but definitely aren't now!

Anyway, a good test IMHO of whether a suggested "10 year view" measure of how a company is doing is actually a good measure for a LTBH investor to pay attention to is to see what happens to it if one changes its starting and ending dates a bit. Especially its starting date - there can be good reasons to want such a measure to change quite rapidly in response to recent events (*) - but letting one's LTBH assessment of an investment change rapidly because events that happened about 10 years ago shift from being just under 10 years ago to just over doesn't seem all that sensible. (And doing that at the same time as only letting it change much less for events that happened about 3, 5 or 8 years ago (**) and not at all by events that happened about 15 or 20 years ago seems downright silly to me!)

So I tried looking at share price charts comparing British Land, Land Securities, Mucklow and SEGRO (***) over various periods of in the area of 10 years. First, over the 30/06/2007-30/06/2017 period that matches the claimed period for the figures you quote from Mucklow's 2017 annual report: over that period, they agree with the report that Mucklow came first of the four and British Land last. Between those two, the share prices have Segro ahead of Land Securities rather than the other way around, but those two are quite close in the annual report and that order difference is well within the explaining powers of not having accounted for dividends and possibly rights issues. The sizes of the gaps between them show some suspiciously large differences, especially the annual report showing Mucklow's TSR around twice that of its nearest rival of the other three, while the corresponding share price gap is only about 20% ahead. But this is enough to tell me that, as I've said above, this share price work is all suggestive of things requiring further investigation, not conclusive.

Next, I looked at the most recent 10-year period (which was 17/02/2008-17/02/2018 because I did the looking yesterday evening - I've only got around to completing this post today). It showed a significantly different picture: Mucklow and SEGRO were neck-and-neck at the top, both about 50% up over that period (and too close for looking at the chart to see which was actually top - not that it really matters when differences are that small, as the positions can very easily be swapped by such short periods of market trading...) and British Land slightly ahead of Land Securities at the bottom. That change to the Mucklow vs SEGRO comparison is rather a big difference for about 7.5 months to make in a 10-year measure that is intended to tell one about how an investment is performing long-term... A bit of further looking at the charts said that SEGRO's share price had underperformed Mucklow's by somewhat over 10% for the period 30/06/2007-17/02/2008, and outperformed it by somewhat under 10% for the period 30/06/2017-17/02/2018: the combination of removing the former from the comparison and adding the latter to it accounted for the change from Mucklow being about 20% ahead to them being neck-and-neck. (One might still expect roughly the same dividend/rights issues adjustments to be needed when shifting to TSR comparisons, so assuming the Mucklow report's TSR figures are correct, Mucklow would still be well ahead of SEGRO - but significantly less well.)

The large influence of the 30/06/2007-17/02/2008 period's underperformance on the share price comparison seemed particularly unreasonable to me, given that other similar-length periods from roughly that long ago (that didn't actually overlap with 30/06/2007-17/02/2008) didn't affect the comparison at all. That's what gave me the idea of varying the starting date of the period but not the ending date, and so I tried out comparisons for the 12-year period 17/02/2007-17/02/2018, the 11-year period 17/02/2007-17/02/2018, the 9-year period 17/02/2009-17/02/2018 and the 8-year period 17/02/2010-17/02/2018. The results were:

* 12-year period: Mucklow and SEGRO neck-and-neck at the top, British Land slightly ahead of Land Securities at the bottom (very similar to 17/02/2008-17/02/2018);

* 11-year period: Mucklow about 15% ahead of SEGRO at the top, Land Securities slightly ahead of British Land at the bottom;

* 9-year period: SEGRO massively ahead of all the others at the top, around 650% up, Mucklow very nicely up, but only a bit short of 150% up (so SEGRO outperformed it about threefold, having roughly 7.5-bagged rather than roughly 2.5-bagged), Land Securities ahead of British Land at the bottom (about 75% up compared with about 50%);

* 8-year period: SEGRO top (about 85% up), then Mucklow (about 70% up), then Land Securities (about 45% up), British Land last on about 40% up.

The 9-year period result is the one that stands out in particular. The reason for it is quite clear when one looks at the specific charts involved: all four shares hit price lows in early 2009, as did many other shares, but SEGRO's was particularly deep and sharp, falling from around 200p at the start of Q1 2009 to a low in the region of 63p and then rising back above 200p before the end of the quarter. A purchase on 17/02/2009 wouldn't quite have managed to hit the absolute low - it would probably have been in the region of 70-80p - but it would have bought far more shares than a purchase even a month earlier or later would have done, and that would have led to far higher portfolio values (and hence far greater TSRs) ever since...

Without further investigation, I can't be certain whether that deep, sharp price spike downwards is real - it's possible that the price data the chart is based on are suspect. But I have had a quick look at RNSes around that period, and have found a "Statement re Market Speculation" from precisely 10 years ago today and within the following few weeks, a "Statement re Media Speculation", a "Rights Issue" and a few "Disclosure of Short Position"s. The rights issue gave 12 rights to subscribe to new shares at 10p each, which (now that I've been reminded of it) is the largest number of rights per share and the lowest amount of extra capital raised by the company per right compared with previous share prices I've ever encountered - both indications that the company could only just find a ratio and subscription price that raised the extra capital it needed and that they could get major shareholders to take up. In other words, the market and media speculation appears to have been thoroughly justified - the company looks to have been in real danger - but the announcement of the rights issue appears to have dealt with that danger quite effectively (note that it was fully underwritten, meaning that the underwriting insurers had basically guaranteed that the company would receive the required capital, from themselves if not from its shareholders). So all in all, those RNSes look entirely consistent with both a big share price before the rights issue was announced and a big share price recovery afterwards. In other words, while the deep, sharp price spike downwards looks suspect, it looks less so after that look at the RNSes.

Anyway, such share price spikes (either upwards or downwards) are rare - it's a bit fortuitous that my look at share prices over six roughly-decade-long periods happens to have included one period whose start hit such a spike. But the fact that it did indicates that there is a real risk of it happening, and it's quite clear that if it does happen, it will affect the TSR for the period involved (since the TSR basically measures what happens to the value of a shareholding bought at the start of the period and subsequently run on the basis of reinvesting any cash it throws off in more shares of the same type). Or more briefly, there is a real risk that what happens to TSR over an N-year period can be distorted in a major way by short-term events around the start of that period, no matter how big a value of N one chooses - a distinct flaw if one wants to regard N-year TSR performance as a measure of how well a company has performed long-term.

That isn't to say that 10-year TSR performance is a useless measure for a LTBH investor - but it is a measure to be treated with caution: look at what was happening 10 years ago and if major things were happening in the short term then, it's liable to be a poor measure for LTBH purposes. And right now, for the rest of this year and some time beyond that, the market turmoil in the financial crisis makes that particularly likely to happen, especially for the sectors near the centre of that crisis (REITs and other property companies aren't top of that list - that position has to belong to banks! - but they're pretty high on it).

In short, IMHO N-year TSR performance might be a good LTBH measure for some values of N, but those values do not currently include N=10 or values in its region.

Finally (apart from the footnotes below), I find it intriguing that the 10-year TSR data in the Mucklow annual report is in its directors remuneration section, and is introduced by:

"The essence of the problem is this: awards made under the [Performance Share Plans] are principally dependent upon a measure of total shareholder return (TSR) compared against a comparator group of companies (those in the FTSE EPRA NAREIT UK Index (which is written into our Remuneration Policy and so is not susceptible to amendment without a shareholder vote in favour)). Certain companies within the comparator group have a tendency to volatility in their performance. This disadvantages a company which performs very steadily over the performance period, as against a much more volatile company which, over the same performance period, may be recovering from a low position and so delivers a greater TSR. The anomalous result is that volatility in the comparator group can disadvantage the participating Executive Directors, but it is a lack of volatility in the Mucklow performance which we believe is prized by shareholders. To that extent, the PSPs seem not to align shareholder and Executive Director interests.

To illustrate this, I set out below a table showing the TSR performance of the companies that have been REITS over the whole of the past 10 years. ..."

Essentially, it's highlighting another case where shorter-term volatility damages TSR as a comparative measure of long-term company performance... I haven't yet fully considered the similarities to and differences from the problem I describe above, but it will be some food for thought!

(*) For instance, a 10-year-and-4-months view of Carillion's TSR now gives a totally different picture than a 10-years-minus-4-months view of its TSR would have done 8 months ago. The two periods have the same starting date and differ only by having different ending dates, so the huge difference in the pictures they give is due entirely to the events of the last 8 months. Those events are obviously highly relevant to one's judgement of what Carillion is like as an investment - 8 months ago, it was highly debateable whether it was or was not a good investment; now, there is absolutely no question that the answer was in fact that it was not! And at least in the case of Carillion, having the "10-year view" change rapidly in response to the last 8 months' events seems very reasonable. (Not that it's actually helpful now, since it's no longer possible to make an investment in Carillion even if one wanted to... One of the annoying features of "good investment or not?" questions is that definite answers to them only ever emerge after they're no longer of any use!)

(**) The 10-year TSR is influenced by dividends paid 3, 5 and 8 years ago, and by share price changes at those times to the extent that they affect how many extra shares those dividends would have bought when reinvested, but those are comparatively small effects compared with the direct effect of the starting date share price on the number of shares originally bought and the ending date share price on the final portfolio value.

(***) More than about five shares make such charts too visually confusing to my eye, and I chose British Land, Land Securities and SEGRO because as you say, they're quite usual HYP candidates. Mucklow is a less usual choice, but as it happens, it's the third REIT (besides British Land and SEGRO) in my own HYP as well as being the REIT whose annual report supplied the TSR data, so for me it was a pretty obvious choice for the 4th share. I didn't see a reasonably clear 5th choice, so I stopped short at comparing four shares.

Gengulphus

[Dod101]

I won't quote from any of G's post but I am not sure I understand if he has come to any conclusion and if so what it is. It seems to me that it generally produces much the same picture; that the big popular property companies are not really a very good investment on a Total Return basis. in particular British Land and Land Securities. I gave up on both a long time ago but then bought back into B Land again on account of its yield. I am still thinking I might sell and replace it with Mucklow quite soon as I like the family holding in it. That usually means that there is a higher than usual conservatism and long term survival bias to it that is sometimes lacking in other companies.

Dod

[tjh290633]

A point to remember about Segro is that they had a very big rights issue in 2009, when they bought Brixton Estates. I think it was 12 for 1 at a very low price, 10p per share by the look of it, followed by a 10 for 1 consolidation. I see that I bought on 04 Sep 2007 at 543p, not far different from today's price of 591p, yet had I not taken up the rights issue, that original price would have been 10 times as much, because of the consolidation. So for a single share held until today, the IRR becomes -17.6%.

I don't know how much attention had been given to this event. As it is, my IRR is 4.35%, taking into account the rights issue, the open offer which followed consolidation, at 285p, adding in 2011 at 223p, taking up the rights issue in 2017 and trimming the holding back on Friday.

TJH

[Gengulphus]

A point to remember about Segro is that they had a very big rights issue in 2009, when they bought Brixton Estates. I think it was 12 for 1 at a very low price, 10p per share by the look of it, followed by a 10 for 1 consolidation. I see that I bought on 04 Sep 2007 at 543p, not far different from today's price of 591p, yet had I not taken up the rights issue, that original price would have been 10 times as much, because of the consolidation. So for a single share held until today, the IRR becomes -17.6%.

Yes, I spotted that rights issue in my post, in connection with the especially glaring anomaly of how Segro currently shows up in a 9-year view of TSR compared with 8-year, 10-year, 11-year and 12-year view. It was indeed 12-for-1 at 10p, and I didn't spot the 10-to-1 consolidation - thanks, that's the sort of detail that will be helpful if and when I investigate the events more thoroughly.

But I will comment that had you not taken up the rights issue, you would have received quite a large amount of cash back, either because you sold the rights or because you let them lapse and received a lapsed-rights payment. You need to take both that cash payment and the 10-to-1 consolidation into account when calculation what the IRR becomes - have you done so? I somewhat suspect not (-17.6% feels rather too large a negative number to me if you have), but even if you have, I think it's worth pointing out for the benefit of anyone else who attempts a similar exercise! Also, they may find https://www.investegate.co.uk/segro-plc ... 1537P7844/ useful - it indicates that if the rights had been allowed to lapse, the lapsed-rights payment would have been slightly under 19.5p-10p = 9.5p per right, or equivalently slightly under 12*9.5p = 114p per original share = 1140p per consolidated share.

Gengulphus

[tjh290633]

But I will comment that had you not taken up the rights issue, you would have received quite a large amount of cash back, either because you sold the rights or because you let them lapse and received a lapsed-rights payment. You need to take both that cash payment and the 10-to-1 consolidation into account when calculation what the IRR becomes - have you done so? I somewhat suspect not (-17.6% feels rather too large a negative number to me if you have), but even if you have, I think it's worth pointing out for the benefit of anyone else who attempts a similar exercise! Also, they may find https://www.investegate.co.uk/segro-plc ... 1537P7844/ useful - it indicates that if the rights had been allowed to lapse, the lapsed-rights payment would have been slightly under 19.5p-10p = 9.5p per right, or equivalently slightly under 12*9.5p = 114p per original share = 1140p per consolidated share.

Gengulphus

Thanks for pointing that out. I only have a figure of 5.2p for the price of SGRN, on 27th March 2009, although digging further the ex-rights price fell to 24.5p from 136.5p cum-rights on 23rd March. The nil-paid rights ranged from 9.6p to 4.2p, being 6.7p the day before becoming fully paid. The original shares rose from 17.75p XD to 20p when the shares became fully paid. As there were 10 rights per share, I should add perhaps 67p to the cash flow of the original share. I think that the 9.5p is a bit unrealistic, as the price of SGRN went 8.55p, 9.6p, 6.95p, 8p, 5.2p, 4.2p, 4.8p, 6.7p and 10p at the close on the 9 working days (23 March to 02 April) when SGRN was quoted.

Having done that the IRR remains at 17.59%, so it was of little overall effect. Even adding 95p the IRR only falls to -17.53%.

TJH

[Gengulphus]

I won't quote from any of G's post but I am not sure I understand if he has come to any conclusion and if so what it is. It seems to me that it generally produces much the same picture; that the big popular property companies are not really a very good investment on a Total Return basis. in particular British Land and Land Securities. ...

That gives me the impression that you've been looking for conclusions about the big popular property companies in my post. If so, I'm not surprised you failed to find them, because I wasn't trying to draw such conclusions! Rather, I was looking at and critiquing your statement that "as we are mostly LTBH investors, I think a 10 year view is helpful", specifically with regard to the context in which you used it, i.e. 10-year views of TSR. I found that they were suspect, because they're highly liable to be changed by short-term events - and that means not just short-term events that happen about now, but also those that happened about 10 years ago. And those short-term events can be company-specific, so that they don't cancel out when comparing different companies.

IMHO that casts a fair amount of doubt on 10-year-view TSR comparisons as a good technique for assessing potential LTBH investments, and thus on any conclusion either you or I try to draw from them about whether any particular companies are good LTBH investments now. They're fine for comparisons of how good the companies would have been if they'd been bought exactly 10 years ago, but that's a different question. Especially if either now or 10 years ago just happens to hit a price spike - a price spike upwards now or downwards 10 years ago will make a company look unusually good, one downwards now or upwards 10 years ago will make it look unusually bad. Of those, the price spikes now aren't a particular problem, as they mean that the shares will look like an unusually good or bad bargain now and that will be reflected in a suspiciously high yield or a low yield now.

But why on earth should a LTBH investor's choice of share now be affected by whether it happened to be an unusually good or bad bargain 10 years ago, much more so than any other date many years ago??? That only makes sense to me for those with a time machine capable of taking them back 10 years to buy it then - and any such investors have it easy!

Gengulphus

[Dod101]

Well I see what you are getting at G but the fact is that over a longish period -and 10 years seems to me to be that- many of the bumps and advantages will iron themselves out. Obviously we cannot wind the clock back but I think over 10 years we can get a fairly good idea of the sort of company we are dealing with because for instance a 'quick fix' on say year 3 will work itself out over the ensuing 7 years. I am looking for good positive culture and a TSR calculated over a 10 year period will give a good indication, nothing more, of the consistency of the returns. Every share as you know will have a poor year from time to time. The trick is to be able to recover and restore the record. No amount of analysis will get it right all of the time but I am not looking for that. I am looking for indicators that is all. I can then take a closer look and try to assess what I can reasonably expect over the coming decade or two.

Dod

[Gengulphus]

Thanks for pointing that out. I only have a figure of 5.2p for the price of SGRN, on 27th March 2009, although digging further the ex-rights price fell to 24.5p from 136.5p cum-rights on 23rd March. The nil-paid rights ranged from 9.6p to 4.2p, being 6.7p the day before becoming fully paid. The original shares rose from 17.75p XD to 20p when the shares became fully paid. As there were 10 rights per share, I should add perhaps 67p to the cash flow of the original share. I think that the 9.5p is a bit unrealistic, as the price of SGRN went 8.55p, 9.6p, 6.95p, 8p, 5.2p, 4.2p, 4.8p, 6.7p and 10p at the close on the 9 working days (23 March to 02 April) when SGRN was quoted.

Having done that the IRR remains at 17.59%, so it was of little overall effect. Even adding 95p the IRR only falls to -17.53%.

It will be a bit more than that, as there were 12 rights per (original) share, not 10. Also a bit less (for the company's trading costs) than 9.5p is not unrealistic: it is what an investor who let the rights lapse rather than selling them would have got in practice. Obviously only some investors who didn't take up their rights would have got 9.5p, but equally only some would have got any other price - not taking up rights is one of those situations where the investor is forced into a short-term trading decision, either about when to sell the rights during the few weeks that they can be sold or to let the rights lapse and get the share price just after those few weeks minus the subscription price and trading expenses, which is basically what the rights would have sold for just after those few weeks had they still been sellable. I.e. letting the rights lapse is basically equivalent to selling them just slightly later than would normally be possible, so that choice is effectively also a short-term trading decision.

Nevertheless, if 67p and 95p make that small a difference to the IRR, 114p won't make all that much greater a difference - I'd guess it takes it down to -17.49% or thereabouts. And even if you've also made the mistake of counting it per current share rather than per original share (which I'm not saying you have, just that it strikes me as a possibility), the 1140p it would be per current share would probably still only make a fairly small difference to the IRR. Which still left me somewhat surprised about how bad the IRR becomes...

But on further thought, I shouldn't have been. The major weakness of IRR as a performance measure for any specific investment is that it is affected by choices the investor makes about how much is invested, weighting periods when they choose to have more invested more highly than those when they have less. E.g. suppose you and I both buy £10k of shares in ABC plc, which halve in the first year we hold them and double in the second year. The difference between us is that you choose to have more invested after the first year, buying enough to quintuple your holding at that time, and I instead choose to have less invested after the first year, selling 80% of my holding and so dividing its value by 5 rather than multiplying it by 5.

So you put £10k in to start with, the holding drops to £5k in value, you put another £20k in to raise it to being worth £25k, and it doubles in the second year to end up at a value of £50k. And I put £10k in to start with, the holding drops to £5k in value, I sell £4k worth to leave a holding worth £1k, and the holding doubles in the second year to a final value of £2k. In IRR terms, that means you calculate the IRR of cash flows of (-£10k,-£20k,+£50k) at yearly intervals and I do the same with (-£10k,+£4k,+£2k). Sticking those into a spreadsheet, I find that you get an IRR of +44.95% and I get one of -31.01% - very different from each other. That difference is completely justified as a measure of how we've each managed to do investing in ABC plc - after all, you've managed to make an overall gain of £20k from capital invested totalling £30k, while I've managed to make an overall loss of £4k from capital invested totalling £10k.

But for assessing ABC plc as an investment, independent of any particular investor's strategy about how much is invested, it's clear that +44.95% and -31.01% cannot both be reasonable assessments! And in fact, neither is: an investment in ABC plc had a -50% performance in the first year and a +100% performance in the second, and you've weighted your overall IRR figure heavily towards +100% by choosing to increase your capital invested by a large factor for the second year, while I've weighted mine heavily towards -50% by instead choosing to decrease my capital invested by a large factor. A fair assessment of ABC plc as an investment, independent of investor choices about amounts of capital invested, must instead work on the basis that the amount invested is not meddled with mid-period - i.e. put in a specific sum at the period start, do whatever is needed to neither add nor remove capital during the period, and see what the result is worth at the period end. For ABC plc with £10k put in initially, that's simple: it halves to a value of £5k in the first year, then doubles back to £10k in the second, for an overall rate of return of 0.00%.

So to sum up the conclusions about that example: as an investment considered on its own, ABC plc had neutral performance of 0.00% over the two years; by choosing to increase your capital invested greatly at what turned out to be a good time to do so during that period, you improved your own IRR to a strongly positive +44.95%; by instead choosing to decrease my capital invested greatly at that time, I worsened mine to a strongly negative -31.01%.

How is all this relevant to the Segro rights issue? The answer is that basically the same sort of thing went on there. At the time the rights were sellable, the Segro share price would have been pretty close to the rights price plus 10p, and original holders of Segro shares would have owned a Segro share plus 12 rights per Segro share they originally owned. When the rights price was R, those holdings would therefore have been priced at (R+10p)+12R = 13R+10p. Selling all the rights at that point would have sold off 12R of that, leaving just the single Segro share worth R+10p, so would have reduced the capital value by a factor of (13R+10p)/(R+10p). Letting them lapse would have done the same with R = the lapsed-rights payment per right and at a time just after the rights were actually sellable. Taking up all the rights at that point would instead have added 120p of new capital, raising the holding to 13 shares per original share, priced at 13R+130p, and so would have increased the capital value by a factor of (13R+130p)(13R+10p).

Your range of rights prices from 4.2p to 9.6p and the lapsed-rights payment of a bit under 9.5p per right therefore imply that an investor who took up all their rights added enough capital to increase their holdings' value by a factor between (13*9.6p+130p)/(13*9.6p+10p) = 1.89 and (13*4.2p+130p)/(13*4.2p+10p) = 2.86, and that an investor who didn't take up any of them, instead selling them and/or letting them lapse, removed enough capital to decrease their holdings' value by a factor between (13*4.2p+10p)/(4.2p+10p) = 4.55 and (13*9.6p+10p)/(9.6p+10p) = 6.88. Also, the Segro share price did generally drop heavily for a fairly long period before the rights issue and has generally risen heavily since (even after adjusting for the consolidation). So the Segro rights issue shares the features with my ABC plc hypothetical example of generally poor investment performance before the event concerned, generally good investment performance afterwards, and one choice of action at the event increasing the amount invested by a large factor while the other reduces it by a large factor. The exact numbers differ, of course, but one can expect a similar result: the good IRR from taking up all the rights will have been increased considerably from that due to Segro itself by the choice to increase capital invested greatly at that point, while the poor IRR from not taking any of them up will have been decreased considerably from that due to Segro itself by the choice to reduce capital invested greatly at that point.

So basically, both the "take up rights" IRR and the "don't take up rights" figure (and especially the difference between them), are very much assessments of the investor and the choices they made, rather than of Segro, the investment. Taking up the Segro rights has turned out to be a very good investment choice, and so unsurprisingly "take up rights" has turned out to produce a much better IRR in the particular case of that Segro rights issue. To get a fairer assessment of Segro as an investment from IRR, you need instead to assume a neither-add-nor-remove-funds method of getting through the rights issue, such as selling enough rights to raise the funds needed to take up the rest (aka 'tail swallowing').

Finally, you'll probably recognise this as exactly the problem with IRR that unitisation fixes - specifically, unit values are unaffected by deposits of capital into a portfolio and withdrawals of capital from it, and so take investor choices about such deposits and withdrawals out of the performance they measure. So you might think that unitisation would be an alternative method to fix this problem with IRR - and you would be sort-of-right. It would indeed fix it, but unitisation when applied to a single share (i.e. the "portfolio" looked at consists of a holding of that share and nothing else, not even cash), IRR calculated for that share using neither-add-nor-remove-funds methods of dealing with corporate actions, and a price history of the share with suitable adjustments for corporate actions such as share splits/consolidations and rights issues all turn out to be mathematically equivalent. So it would be somewhat better described as an alternative way of expressing the same method of fixing the problem...

Gengulphus

[tjh290633]

It will be a bit more than that, as there were 12 rights per (original) share, not 10. Also a bit less (for the company's trading costs) than 9.5p is not unrealistic: it is what an investor who let the rights lapse rather than selling them would have got in practice. Obviously only some investors who didn't take up their rights would have got 9.5p, but equally only some would have got any other price - not taking up rights is one of those situations where the investor is forced into a short-term trading decision, either about when to sell the rights during the few weeks that they can be sold or to let the rights lapse and get the share price just after those few weeks minus the subscription price and trading expenses, which is basically what the rights would have sold for just after those few weeks had they still been sellable. I.e. letting the rights lapse is basically equivalent to selling them just slightly later than would normally be possible, so that choice is effectively also a short-term trading decision.

Nevertheless, if 67p and 95p make that small a difference to the IRR, 114p won't make all that much greater a difference - I'd guess it takes it down to -17.49% or thereabouts. And even if you've also made the mistake of counting it per current share rather than per original share (which I'm not saying you have, just that it strikes me as a possibility), the 1140p it would be per current share would probably still only make a fairly small difference to the IRR. Which still left me somewhat surprised about how bad the IRR becomes...

Sorry, I mixed up the consolidation and the rights. And you are also right, in that I worked on the post consolidation share rather than the original share. Making that correction changes the IRR to -14.9%.

It may be of interest to note that, from the date of consolidation and the open offer, 31 July 2009, my IRR has been +11.1%. That includes the effect of buying some more (30%) in 2011, taking up the rights (1 for 5) in 2017 and the trimming back by 25% last Friday.

TJH

[Gengulphus]

Well I see what you are getting at G but the fact is that over a longish period -and 10 years seems to me to be that- many of the bumps and advantages will iron themselves out. ...

Indeed, but the question is how quickly they will iron themselves out and whether 10 years is long enough. My original impression from the price chart I used was that the price spike 10 years ago was down to about 1/3rd of what the price had been just a month before and again rose to within about a month afterwards, meaning that that short-term price spike would have resulted in about three times as many shares being bought on an initial purchase than would have been the case if it had occurred a month either side. That would have resulted in the portfolio being about three times what it would have been ever since - it would have generated three times the dividend income, reinvesting the dividend income would have bought three times as many shares and so maintained it at three times what it would have been, etc.

What could cause that 3x advantage to be ironed out? In general, there are three things that can:

* Having multiple similar events that cause a variety of advantages and disadvantages, so that you get some sort of statistical average of their effects and aren't at the mercy of just one of them being an unusually good or bad outlier. But a 10-year view of TSR basically assumes just one initial purchase, with any subsequent purchases being funded by dividends or corporate action proceeds, and those will all also be three times as big as they would otherwise - i.e. they are affected in the same way as the initial purchase, not a variety of different ways. The result is that this won't iron out an advantage or disadvantage from the initial purchase assumed by an N-year view of TSR at all, no matter how big N is. If you get an unusually good or bad share price when that purchase happens, you're stuck with its effects.

* Adding or removing funds that don't increase with the number of shares you've got - e.g. topping the portfolio up with regular savings, or withdrawing a constant income from it. Again, a 10-year view of TSR assumes that this doesn't happen (apart from funding the initial purchase itself) and so this doesn't iron out the effects of an unusually good or bad share price at the starting date of an N-year view of TSR at all, no matter how big N is.

* The fact that annualising a given size of advantage over more years produces a smaller increase in the annualised rate of return it comes up with. This does iron out a 3x initial advantage over enough time, but 10 years isn't anything like enough to iron it down to insignificance. For example, a 3x initial advantage followed by 10 years of a 5% normal rate of return increases the portfolio value by a factor of 3 * 1.05^10 = about 4.89, which annualises over 10 years to a rate of return of about 17.19% - I would certainly not describe the difference between rates of return of 5% and 17.19% as having been ironed out! And the same calculation done for 100 years rather than 10 produces a rate of return of about 6.16% and 5%, a lot less significant but IMHO still not ironed down to total insignificance. And of course, working out a 100-year view of a company's TSR will almost always be totally impossible due to the company simply not having 100 years of history, and totally impractical due to the amount of research work required on that history even in the cases where it is possible in principle.

I will modify the details of that view somewhat as a result of further digging I've just done. I said when I brought the subject up that my conclusions did depend on the share price chart I used adjusting its prices for share splits/consolidations and rights issues. I've now looked at whether it does, in particular for the rights issue in early 2009 and the 1-for-10 consolidation later in the same year. It does, but in the case of the rights issue, on the basis of it being equivalent to a 13-for-1 share split. That's unreasonable: increasing one's holding by 12 shares for a payment of 120p is obviously not equivalent to increasing it by 12 shares for no payment at all! Based on TJH's rights price figures, a more reasonable treatment is to treat it as roughly equivalent to a 6-for-1 share split. A chart prepared on that basis still shows a deep, sharp price spike downwards, but somewhat less deep and sharp - the price roughly halved in the month before the spike's low point, and took about 5 months to recover from that halving. It's still short-term in a HYP context, and a 2x initial advantage will still take far more than 10 years to iron out at all properly, so nothing fundamentally changes about my conclusions above - but the numerical details do change.

I should also point out that I said when I first brought this up to take care about using 10-year views of TSR, which is not the same thing as saying not to use them. Specifically, the care I would suggest taking is a check of the share price volatility around 10 years ago: if it was low (i.e. the share price was fairly stable), the 10-year view of TSR is probably a good indicator of the company's long-term performance; if high (i.e. the share price was changing quite rapidly), it may well be distorted seriously by short-term events that happened around 10 years ago.

And when comparing two or more companies, do the same check for each of them: comparisons involving any company showing high share price volatility about 10 years ago are quite liable to be unreliable.

That does have the problem that it fails to say what to use instead if the volatility checks say that the 10-year view of TSR is likely to be distorted. One could try e.g. 12-year, then 8-year, then 15-year views of TSR instead until one finds one for which the corresponding volatility check says "low" - or if they all say "high", maybe giving up on the company on the grounds that it seems to be too badly affected by short-term issues for its long-term performance to be discernable at all.

Another possibility that occurs to me is that one might do something I might call an "extended investment period TSR": do a TSR-like calculation on a holding of the share, the basic feature of which is that you assume all dividends and corporate action proceeds are immediately reinvested to add further shares to the holding. But rather than starting it off by effectively assuming a single lump sum invested at whatever the share price happened to be 10 years ago, instead do so by assuming e.g. 20 purchases, each with 5% of what would have been the lump sum, at evenly-spaced times spread over the period from 15 to 5 years ago. That makes the effect of short-term share price volatility at the time of each purchase 5% of what it might have been with the single-purchase assumption, and while 20 such effects might add up to just as big an overall effect, they're very unlikely to - in effect, the changed assumption is highly likely to create the "Having multiple similar events that cause a variety of advantages and disadvantages, ..." cause described above of ironing out the bumps.

... Obviously we cannot wind the clock back but I think over 10 years we can get a fairly good idea of the sort of company we are dealing with because for instance a 'quick fix' on say year 3 will work itself out over the ensuing 7 years. ...

I'm not disputing that - I agree that over 10 years, we can get a fairly good idea of the sort of company we are dealing with. What I am disputing is whether we can reliably get such an idea from the 10-year view of TSR: I think there is a very noticeable risk that it will give us distorted versions of such an idea. Especially over the next 2-3 years, as "10 years ago" passes through the main years of the financial crisis.

Gengulphus

[daveh]

Just had a look at how SEGRO has performed for me. First bought in 2008, I tail swallowed the 2009 rights issue and bough more shares in 2011 and 12 and took up the 2017 rights issue. For me its been one of my better performing shares. A quick XIRR calc shows a 9.4% return on capital only and if I add in dividends ( I did the calculation by adding in each years total dividend on the 1st Dec so it will be an approximation) the XIRR rises to 12.5%.

SEGRO won't be getting any more of my money at the moment as it makes up too large a percentage of my HYP at 7.4% and I also own British Land at 2.6% of my HYP so the property/REIT sector is also unlikely to get a top up in the near future.