Double Counting, by John Hussman

...
Let's look at this argument closely. First, why does Siegel say that the earnings yield is an estimate of the real return on stocks? Think of it this way. Reported earnings subtract out depreciation, which is another way of saying that earnings are reported as if the company reinvests only enough to replace depreciation and keep its stock of productive assets constant over time. If the company were not to invest anything for growth, it would theoretically be able to pay out all of its net earnings as dividends. If earnings on the fixed stock of capital could grow at the inflation rate by virtue of monetary factors alone, you would get zero real earnings growth. Then holding valuations constant over time, the earnings yield would be a measure of the real return on stocks. Fine if you believe the assumptions. Now let's look at the data.

The first problem is that in order to produce 2% real annual growth in earnings per share, companies have historically devoted about 50% or more of their earnings to reinvestment and repurchases - over 300 basis points of that earnings yield to get 200 basis points of real growth. That's a tip-off that historically, competitive pressures have prevented earnings from simply growing at the rate of inflation without new investment. You had to invest new money to get your earnings growth up to the inflation rate. In general, once your return on invested capital falls to the point where you're willing to buy your own stock instead of making new investments, the simple earnings yield overstates probable long-term real returns (especially if the yield is based on record earnings). In fact, about 1% of the long-term real return on stocks has come from an increase in the overall level of valuation in recent decades. Absent that increase in valuations, the real return on stocks would actually have been at least 1% less than the average long-term earnings yield.
...

For the complete article see http://www.hussmanfunds.com/wmc/wmc070430.htm