A CAPE-based asset allocation strategy

I’ve mentioned my tactical asset allocation efforts in a couple of previous posts.  Both of those have been somewhat vague about how I actually decide on the stock/bond split, although not deliberately so. 

If Ben Graham can shout out the Net Net method to the world over 50 years ago and not have its effectiveness affected, then surely my tiny whispers on the web can do my approach no harm. 

In fact, it may to some miniscule degree make the markets more efficient.  Like the proverbial fly stopping an oncoming supertanker.  Perhaps I may even win a Nobel Prize, but I doubt it.

Both Shiller and Smithers and others have shown that it is possible to value markets and that market valuations are bound by an invisible elastic thread to both earnings and assets.  More importantly, these valuations allow you to say something about the expected future returns.  Higher valuations mean lower expected returns and lower valuations mean higher expected returns.  Also, average valuations mean average expected returns.

Shiller’s CAPE (Cyclically Adjusted Price Earnings) is my starting point when looking at earnings-related valuations.  This is the ratio of the current market price to the market’s average real earnings over the past decade.  For the S&P500 the long-term average CAPE is currently 16.35 and is shown below as the horizontal line.  The data for this can be found here.

The chart shows that 1920 was a good time to invest in the US, as was 1932, but 1929 was not so good, nor was 2000. 

If you had calculated CAPE at the time and compared it with the long-term average then you would have been able to see that this was true.  Perhaps you could have even adjusted your portfolio to take advantage of this insight. 

This was entirely possible since the long-term average CAPE was about 15.5 in 1920, 14.8 in 1929 and 1932, and 15.5 in 2000. 

None of these values are much different than today’s average of 16.35.  In fact, it does look like 16.35 is skewed by the recent liquidity bubble since the 1980s, so nearer 15 may be nearer the true average.  I’ll check again in 20 years.

A nice mental image is to see the current CAPE as being attached to its long-term average by a piece of elastic.  The further it gets from the average the more force there is pulling it back to the average, and it becomes more likely that the price will move CAPE back toward the average rather than continue away from it.  Eventually probability and mean reversion win the battle against fear and greed.

For the FTSE 100, I have data going back to 1993 (if someone has better then please let me know).  This gives me CAPE values from 2002 to 2010, the average of which is 16.47.  If I include the earlier years, averaging for what years I have (i.e. in 2001 I only have 9 years of earnings to average and 1993 only 1 year) then I get a somewhat fudged long-term CAPE average from 1993 to 2010 of 18.7. 

The higher value isn’t a surprise as this period covers the dot com bubble.  Overall, I don’t think these values are too different from the US long-term average, so my working assumption in the absence of better data is that the FTSE 100 long-term average CAPE is the same as the S&P 500’s.  Of course, this isn’t true but fiddling with the numbers doesn’t in the end produce meaningful differences; saying the average was 15 or 20 produces similar results.

The next thing to do is dream up some sort of function to decide on the stock/bond split.  By looking at the chart above you can see that the CAPE value never really gets below about 5, and only twice goes above 30.  My assumption is that it would probably be a good idea to be 100% in stocks near the low points and 100% out of stocks at the high point, and to have an average amount of stocks, say 30 or 40%, when CAPE is at an average value. 

Again, whether the limit values are 100% or 90% doesn’t make much difference.  5 is pretty close to 8, which is half the long-term average.  30 is pretty close to 32, which is twice the long-term average. 

This gave me the nice simple and dare I say it elegant idea of adjusting my allocation to stocks as CAPE moves between half and double its long-term average.  A simple transform function can take the current CAPE and adjust it to give a percentage between 0 and 100.

Stock allocation = 1 – (CAPE – AvgCAPE/2)/(AvgCAPE*2 – AvgCAPE/2)

As the market gets gradually more expensive the stock allocation reduces.  This is in line with the reduced expected future returns of stocks, and the increasing risk.  Here I’m measuring risk as the increased likelihood of the market falling from current levels, rather than as volatility.  Hidden in that sentence is, I think, the key to value investing.

So now I have a function which tells me how much to hold in stocks depending on the value of CAPE.  If CAPE is 8.18 then I’ll have 100% in stocks (and we saw about that level in March 2009).  If CAPE is 32.7 then I’ll have 0% in stocks.  If CAPE equals the long-term average of 16.35 then I’ll have 66% in stocks, which is pretty close to the 60/40 split recommended in many places.

The results of this function can be seen below, using my slightly fudged CAPE values for pre-2002 years.

In this chart, the split is between stocks and high-interest cash accounts as defined in swanlopark.  The FTSE 100 plot is with dividends reinvested; 60/40 is 60% FTSE 100 and 40% cash; and the cash plot is the high-interest cash.  TAA is the asset allocation portfolio.  It is rebalanced annually to bring the allocations back into line according to CAPE.  Trading costs should be negligible.

And there we have it.  An approach which gives roughly the same results as the market with much less risk.  More importantly, it does it without the terrifying drawdowns that the FTSE 100 has.  It’s much easier on the mind, which makes it easier to stay invested for the long term.  I know this first hand having seen relatives selling out last March after a 50% decline.  Now they are still in cash and regretting ever getting out.