Why do high earning multiples imply a likelihood of low or even negative returns?

I quote Burton Malkiel in his May 4 2020 dialogue with Robert Shiller.

Some historical perspective is useful. Even during one of the most spectacular “bubbles” in stock market history (the dot.com bubble of the late 1990s, considered damning evidence against EMH), most investors who tried to time the market made egregious mistakes. In 1996, Bob (with John Campbell) presented an excellent paper to the Federal Reserve showing that earnings multiples possessed substantial ability to predict future rates of return. Since these multiples were then at all-time highs, the work implied a likelihood of low or even negative returns. This work influenced Federal Reserve Chair Alan Greenspan to make a famous speech questioning whether the stock market was at bubble levels and whether investors were exhibiting “irrational exuberance” in view of the economic outlook. The stock market rallied strongly for four years thereafter, and long-term investors who bought stocks after the speech earned generous rates of return.