Andrew Smithers of the London based economic consulting group Smithers and Co is not impressed with these measures, and that is putting it mildly. As readers of the Herald and the London Financial Times well know he is not a fan of stockbrokers in any way, shape or form indeed he reckons that stockbrokers have manufactured their own brand of shonky economics which he has named "stockbroker economics".
Smithers research has two important features which differentiate it from much of the free continuing professional development presentations typically offered in NZ by high cost fund managers; it isn't free but it is useful.
The beauty of stockbroker economics is that all economic news is good news for the stockmarket including rising interest rates, falling interest rates, inflation, deflation etc etc.
Anyway, recently Andrew decided to write the definitive critique of stockbroker valuation methodologies and sent the report to his client's which is how it landed on my desk. Smithers research has two important features which differentiate it from much of the free continuing professional development presentations typically offered in NZ by high cost fund managers; it isn't free but it is useful.
In an email Andrew said we were his only NZ client, which is a bit sad and an indictment of the Code Committee and the industry as a whole. Incidentally Mr Smither's featured last week in the Financial Times "Lunch With The FT" column where he was interviewed by Martin Wolf. It is worth a read. Andrew is apparently in his 70s now and due to retire at the end of the year. Mr Wolf finished the article by saying "often criticised, but usually right, Smither's and Co has been a fount of original and penetrating ideas. When it is closed this year it will be greatly missed".
Not by many in NZ unfortunately.
Where Andrew displays his independence and economics background is when he says "forecasting ability cannot be judged over a single time period as these returns will depend as much on the level of the market at the end of the period as at the starting point.
Smithers' work on valuation tools are the subject of today's column so if you are a stockbroker, financial planner or fund manager turn away now. Smithers kicks off by saying that the worth of any valuation model obviously must be measured by its ability to forecast market returns. For example if the stockmarket's PE ratio is low or the dividend yield is high is this a reliable signal of relatively high stockmarket returns in the future?
Hint; it isn't. Where Andrew displays his independence and economics background is when he says "forecasting ability cannot be judged over a single time period as these returns will depend as much on the level of the market at the end of the period as at the starting point.
We avoid this with "hindsight" which measures the average return over a large number of future years. This allows us to put values on the market in earlier years provided enough time has passed".
So as a first step to test the models for each year Smithers compares the average return of the subsequent 30 years and those years with the best average return will be the years where the market was relatively cheap and will thus have a high "hindsight value". He then charts these hindsight values for the years 1899 - 1983 and on the same graph to test the ability of various valuation models he charts the relative PE ratio, relative dividend yield etc.
Confused? An example will make it clearer.
In 1916 the US stockmarket was cheap according to the price earnings ratio model as the PE then was just 6.4x which is barely half of the average valuation over the 1899 - 1983 period. If PE's were a good valuation model you would expect high returns over the next 30 years but returns over that period were less than 1 per cent real.
Similarly in 1921 the market PE was 25.2x ... about twice the average yet market returns were 11.5 per cent pa real for the next 30 years. Prospective PE's produce even worse errors leading Smithers to comment that "the comparison of prospective PE's with hindsight values shows that their use by investment bankers and, quite disgracefully by Janet Yellen, to make claims about stockmarket values is nonsensical and would presumably be illegal if rules on "truth in advertising" were applied to their comments".
In the same way the dividend yield in 1916 of 5.7 per cent was 20 per cent above the average dividend yield yet real returns for the next 30 years were less than 1 per cent real. So dividend yields don't meet the test either.
"Not every story has a happy ending but it should be said that both of these measures have been signaling that the stock market is overvalued for quite some time and anyone following these models would have missed out on much of the recent rally.
I know what you are thinking ...... these are just individual data points and there could be lots of other instances where, for example, low price earnings ratios forecast high returns. Mr Smithers addresses this concern by employing a statistical method whereby he calculates the R² for each valuation model.
These are summarised in this table:
R² is simply the extent to which one variable explains another variable so what the .25 for the trailing PE says is just 25 per cent of hindsight values are forecast by the PE.
So are there any worthwhile valuation models? Look down the table and you will see that Tobins Q explains 79 per cent of future returns and the cyclically adjusted PE multiple explains 55 per cent. It is not necessary to know how these latter two models work but the key point is that both signal that the US stock market is about 70 per cent overvalued.
Not every story has a happy ending but it should be said that both of these measures have been signaling that the stock market is overvalued for quite some time and anyone following these models would have missed out on much of the recent rally. Let's hope next time really is "different". Unfortunately it probably won't be.
Brent Sheather is an Authorised Financial Adviser. A disclosure statement is available upon request.
