But there's another more subtle reason why "average earnings" is far more useful than current (or forward) earnings: in a free market, companies with undifferentiated products will tend to see their earnings naturally revert to the mean. This occurs because when profits are high, new competitors enter the market, old competitors ramp up production and customers seek out substitutes, until profits are driven to normal levels. When profits are low or negative, the weakest competitors are forced to close their doors and no new competitors enter the market, improving conditions for the companies that stick around.
Depending on the industry, this process will vary in the amount of time that passes between peaks (we discussed this process for the oil industry here). But Wall Street's obsession with current earnings (as discussed here by Ben Graham and David Dodd) provides opportunities to value investors when earnings are simply at the lower end of the cycle.
An important caveat is that one must be sure that earnings are low due to a cyclical effect, not a secular effect (the newspaper industry is often cited as an example of an industry in secular decline, thanks to the world wide web and other forms of media). Furthermore, one must only purchase the companies with little to no debt and low cost levels, as these are the companies that will survive the downturn. Adjustments must also be made for companies that have increased their book values substantially (either by retaining earnings, or issuing new shares); the earning power of such companies should be higher than they have been in the past, though this is not always the case.
The bottom line is, average earnings are a much more useful gauge of a company's earnings power than current or oft-cited expected future earnings, which are based on only one period and could be affected by infrequent items or the industry's current position in its business cycle.