I’ve written a lot recently about the FTSE 100, whether it’s expensive or cheap and different ways of visualising the range of values it could reasonably reach; so when a friend of a reader suggested that my valuation “bands” were too wide, I decided to review my approach.
If you look back at some of my recent articles on FTSE 100 valuations, you’ll find my CAPE Probability Fan Chart and Valuation Heat Map (CAPE is Robert Shiller’s valuation tool, the Cyclically Adjusted PE ratio).
In both of those charts, I use a range of bands covering CAPE valuations from 8 to 32. The general idea is that the FTSE 100’s CAPE valuation generally moves in a range from half to double its long-term average value (which is about 16).
In the latest UK Value Investor newsletter I labelled these valuation bands from “Very Expensive” to “Very Cheap” as you can see below:
And these are the bands that are accused of being too wide. So the question is, are they?
Too wide, or poorly defined?
Having looked at this for a while, I don’t think the bands are too wide. In fact, I think they may be too narrow, but even that isn’t the real problem. The real problem is that the boundaries between one band and another are arbitrary, or worse, they are based on an irrelevant idea; the idea that they should be evenly spread.
Given that the bottom half of the range runs from a CAPE of 8 to 16, it seemed reasonable to split that half into four groups, each with a range of 2 (8-10, 10-12, 12-14, 14-16). And the upper half of the full range goes from 16 to 32, so do the same there and you end up with another four groups (16-20, 20-24, 24-28, 28-32).
I think those groups look nice, they’re balanced but they’re also meaningless.
Cheap and expensive relate to expected future returns
When people talk about “cheap” and “expensive” investments they are usually implicitly talking about expected future returns. If something is cheap but expected returns are normal then it isn’t really cheap; it’s fairly valued.
So what happens if I organise these valuation bands based on expected future returns? A market with high expected future returns could then be sensibly labelled as “cheap”, and one with low expected future returns could be called “expensive”.
Future returns can be estimated by assuming that cyclically adjusted earnings (their 10-year inflation-adjusted average) will grow in line with inflation plus 2% (i.e. 4% a year in total), which is about the long-run average growth rate.
I’ll assume that the median CAPE value is 16, so that’s our “fair value”, and the market will have that valuation in 7 years (the typical time required for mean reversion, according to Jeremy Grantham’s GMO). So over 7 years (according to this model) the FTSE 100 will produce capital gains of 4% a year plus or minus whatever gain or loss is required to get its valuation back to a CAPE of 16.
From fair value, the long-run average real return from UK shares is 5% a year (2% real growth plus 3% from dividends). If the market is high enough so that the 7-year annual returns are half that amount (i.e. 2.5% a year) then I’ll call the market Slightly Expensive, and if 7-year real returns are expected to be negative then I’ll call that Very Expensive.
On the other hand, if the market is low enough so that 7-year real returns are 50% above normal (i.e. above 7.5% a year) then I’ll call that Slightly Cheap, and if those real returns are expected to be more than double the long-run average (i.e. above 10% a year) then I’ll call the market Very Cheap.
Of course, those are arbitrary and subjective definitions, but they seem reasonable to me.
The table below shows what sort of FTSE 100 level we would have today if future returns were as outlined above:
The current valuation band is highlighted.
With this approach, any extreme valuations like the dot-com bubble with a CAPE of 32 just get lumped into “Very Expensive”, which seems fair enough as I can then quantify the level of expensiveness by measuring the exact expected 7-year return (in case you’re interested, the expected 7-year real return from 1999 was minus 6% a year; not what I’d call an attractive proposition).
Here’s what those bands look like using my traffic light colour-coded “heat map”:
The top of the red band is the all-time high valuation of CAPE 32 from 1999, while the bottom of the green band is the lowest CAPE of 10, from 2009.
Currently, the FTSE 100 is at 6,680, so that puts it in the “Slightly Cheap” band with 7-year expected real returns of 8.2% a year (10.2% if you include inflation at 2%). So although this bull market has come a long way from the dark days of 2009, it may have plenty of wind left in it.
John,
The logic of this article is solid, but it seems to be based on a weak evidence of the FTSE long term CAPE, which you suggest is “about 16”. But a sensitivity analysis of just 10%, means it could be 14.4 or 17.6. This makes a material difference to your conclusions. Surely to be confident in the conclusion this analysis needs a much more robust evidence base to assess the long term FTSE CAPE?
Ben
Hi Ben, I don’t think it does. We can only be reasonably confident about forecasts at extreme valuations, and at extreme valuations it doesn’t really matter if the average is 14 or 16 or 18. If CAPE is 10 then the average is very likely to be much higher, and if CAPE is 30 then the average is very likely to be much lower.
If CAPE is 14 or 18 then yes, things are much less certain and perhaps I didn’t state that loudly enough.
One problem with quantifying any of these things (such as my nicely defined bands above) is that it looks like you can have precise knowledge about the future when in fact you cannot. Precision about the past does not lead to precision about the future. We’re all stuck in a fog of Knightian uncertainty, and so any research I do, or anyone else does, must be seen as existing in that fog. Sometimes the fog clears, such as when valuations are extreme, but usually it’s thick, opaque and fear-inducing.
That sounds pessimistic. What I’m really trying to say is that if you re-run the calculations with an average CAPE of 14, 16 or 18 it really doesn’t make a lot of difference, in theory and more importantly in the real use of thinking about valuations in terms of implied future returns.
If you really want I can produce charts with the average value set at 14 and 18 to see what difference it makes (although my wife might complain if I’m late home for tea).
Hi John,
Interesting stuff.
I too am a fan of CAPE but struggle to find the data on the UK – other than via Mebane Faber who charges for it here http://www.theideafarm.com/
One of his sample tables is here http://www.mebanefaber.com/2013/04/01/cape-updates which showed in April 2013 that (as now) the UK stock market is neither extremely hot nor cold, but that the US is clearly overheated on this measure.
But i think we have to bear in mind that any big correction of the US market will likely hit the UK and also other world markets.
We should also note that Russia was one of the cheapest markets in April 13 – and now, presumably cheaper still – but who is brave enough (or foolhardy) to buy some given Mr Putin’s disregard for asset ownership (including land!) ?
I’m keen to know the period over which you’re taking your averages here.
Is it from the late 1980s?
If so, then surely this has been a period of exceptionally high PEs – as shown on Shiller’s CAPE chart here http://wp.me/p45vXF-8T
I would have thought that the CAPE needs to be looked at over the VERY long term to pull in all the big peaks and troughs and thus determine a sensible average.
I also think that the Barclays latest equity gilt study – which introduces the idea of SCAPE (sectorially adjusted CAPE) is useful – as it adjusts the US down for it’s high exposure to high PE sectors like tech – whilst adjusting the UK valuation up due to higher exposure to banks and Utilities.
It seems to me that the US is overvalued and the UK is too.
Paul
Hi Paul
I only have data going back to the late 80s for the UK, which is barely enough to start thinking about long-term average CAPE values, but the US has 130 years or so (i.e. the very long-term) where the average is 16.5. My average for the UK cheats by combining Shiller’s data for the US with the UK data that I have, by simply replacing the US data from 1988 onwards.
I think 16 is likely to be a reasonable guestimate of the population mean for the UK (i.e. the mean CAPE value from when shares were first traded to whenever the stock market closes for the last time). If you look at CAPE values for other countries (so horizontally rather than vertically) it tends to be between 10 and 20 most of the time, see:
https://www.allianzglobalinvestors.de/MDBWS/doc/Market-Insights-Kapitalmarktperspektiven-2014-EN.pdf?67aba2557ea09bd725df8fea6d22eef3ab555090webweb (page 3)
https://www.fidelity.com/bin-public/060_www_fidelity_com/documents/Developed-Market%20Equities-Q1-Conversation-Starter-Fidelity.pdf (also page 3)
Plus various other bits of research into global CAPE ratios.
All this talk of the mean CAPE excludes the other half of the story which is the dispersion, whether range, inter-quartile range, standard deviation and so on. If I get time I might do a post on that as everybody (including me) just seems to focus on the central tendency number instead.
There are so many caveats that if you wrote them all out you’d never get round to the meat of an article!
So as far as my “forecast” goes, the idea is to show what returns you would get, from today, assuming 2% inflation, 2% real growth and a CAPE value of 16, all of which are entirely reasonable assumptions as inflation, real growth and CAPE have all spent a lot of time somewhere near those values.
Of course in reality none of those assumptions will be correct, but they are still reasonable.
As for adjusting the UK CAPE, I personally wouldn’t bother. Once you get beyond looking at the index you might as well start valuing sectors or individual stocks, which is my other area of interest.
But the UK overvalued? I don’t think so. Not with a dividend yield of 3.5% for the FTSE 100.
John
Thanks for the substantial reply John.
I accept that 3.5% sounds juicy today but look at market performance against yield long term – and that’s available very long term from equity gilt study – and you’ll find that below that number the risks of poor performance are high.
I guess we won’t know whether today was good or bad value till such time as central banks allow interest rates to get back to something close to normal (inflation plus a bit)
My concern is that markets are yet to build that move into their pricing.
Best