if you are going to use the S&P500 plus the short bond yield then a new bell curve should probably be created. I would imagine it will be something closer to 18.5 to 21.5 when done over a 25 year span. Elaine G. back in 1987 said that it centered around "20" and didn't stray very far. Unfortunately I don't know exactly what index she used for the P/E or interest rate she added to it.
On further thought it shouldn't matter too much which index the PE is taken from or whether its forward or historical PE - providing the index is sufficiently wide (diverse) enough.
Barclay's UK Equity Gilt Study (2006) highlights the negative correlation of 15 year historical price performance to forward price performance and specifically identify that that the inverse correlation improves more towards the boundaries of extremes - and that PE serves as a time-independent relative measure of price.
So as you suggest Tom, constructing a Bell and identifying the 10% extremities should suffice.
Which in turn lowest common denominates down to the Dividend Yield being an indicator of PE such that in principle the 7/PE element only serves as a confirmation measure on the DY/6 (or US DY/5) measure.
A simplified version is therefore simply Div Yield / 6 generally or Div Yield / 5 for the US. Which on a 1.83% S&P current div yield implies 36.6% stock 63.4% cash.
Which taken yet further still implies that a diverse range of high yield stocks should outperform overall.
Market Data 
Markets 
AI
