The macroblog posts a chart showing that the S&P 500 P/E rises and falls with the growth rate of real money balance, measured using MZM deflated by the PCE deflator.
This is a good way to predict the S&P 500, provided you can also predict corporate earnings, money supply and inflation.
In the meantime, Barry Ritholtz warns us not to make oversimplified predictions based on few variables.
Any single variable will give you an easy prediction, a goiod bumper sticker, but have a low probability of a correlated predictive outcome. The more variables you introduce -- up to a point -- the more likely your predictive outcome will be. I freely admit to not knowing precisely where the line of introducing too much complexity is -- but its likely much higher than 4 or 5 variables.