P/E vs S&P 500 (50 Years)

Take a look at this great chart of P/E versus the S&P 500.  It does a wonderful job of setting up the valuation question.  You can see what is wrong with it, even without looking for a better model.  The trailing P/E ratio method is poor both descriptively and prescriptivey — the two reasons we look for relationships.  It is exciting in that it explains why the parade of talking heads have been saying for years that the market is overvalued, and lets you judge the wisdom of their statement.

Take a look at the chart, courtesy of Mike Panzner via Barry Ritholtz, and then we’ll analyze it more carefully.

Link: P/E vs S&P 500 (50 Years).

As promised, today brings us to the 4th in our series of charts: P/E vs SP500click for larger chart courtesy of Mike Panzner, Rabo Securities I’ll get into the significance of what this means to the markets later, but for now, note where the P/E is over …

When you put variables together in a chart you are implying some kind of relationship or else why bother?  This can be explanatory, showing how the actual market prices follow the independent variable, in this case P/E.  Just from a glance, it is pretty obvious that the explanatory value is poor.  Just look at whether the S&P line tracks the shaded P/E area.  The only good correspondence is back in the Eisenhower, Kennedy, and Johnson years.  So as a descriptive model, this is not very helpful.  My research team would be looking hard to improve the model specification, looking for more relevant variables and improving measurements.

A relationship can have valuable prescriptive value, even if there are big gaps in explanations.  This would be true if by following a simple rule you could outperform the market.  So how does the simple trailing P/E method do on this score?  I don’t want to put words in Barry’s mouth, since he is going to explain the market implications further.  He will probably have a very interesting take on this.  Having said this, the obvious inference is that one should own the market when the P/E is low, the white periods, and sell it (or short it) when it is high, the shaded periods.

How would that have worked?  It looks like you would have made money in the Nixon-Ford-Carter and early Reagan years, and also during a small piece of the 90’s.  You would basically have missed most of the double(+) in market value from 1991 to date.  If you had shorted during that time, you would have gotten buried, particularly during the bubble era.

We should aspire to more — a better model on both descriptive and prescriptive criteria.  Fortunately, that can be readily accomplished in the following two ways:

  1. Adding consideration of interest rates.  There are various means of doing this, and one can argue about which method is best.
  2. Looking at forward earnings instead of trailing earnings.  This works both ways, helping when earnings estimates are going down, and when they are going up, as they have been for the last three years.

I strongly agree with Barry’s comments about markets overshooting from extended positions in both directions.  I look forward to his analysis.  I will also post a chart showing the results of a model using these two suggested improvements and looking for the overshoot factor.

You may also like

One comment

  • Barry Ritholtz June 15, 2006  

    Actually, the parade of talking heads have been claiming that P/E shows the market as cheap — based on forward earnings.