We like to think that past profits are a good predictor of future profits, and that past growth is a good predictor of future growth. As “The Level and Persistence of Growth Rates” by Chan, Karceski, & Lakonishok (2003) shows, that is not the case. Growth rates are not persistent. In fact, they appear to be random. In other words, companies with high earnings growth rates have no greater chance of continuing to grow earnings quickly than those with low growth rates. Sales growth rates do tend to persist, but it does no good for the investor if a company such as Pets.com can generate 200% sales growth if it fails to ever make a profit.
There is little indication that anything can really predict which companies will grow earnings at fast rates in the future. Nothing correlates with future earnings growth: not analyst forecasts, not past growth, and not P/E or other valuation metrics. That last thing can be good for us, though—companies we buy with low P/E ratios can turn out to be growth stocks!
Does anything predict future growth? A couple things might—companies with high dividends tend to grow more quickly (if we count the dividend yield as part of growth) than non-dividend paying companies. Also, companies that spend a large portion of revenues on research tend to grow more quickly. I will investigate both of these findings in more detail in the future.
Because of the unpredictability of earnings growth rates, those companies that are priced cheaply are the best investments, while those with high prices (high P/E ratios, high P/BV ratios) are poor investments. If you need more confirmation of this fact, see The Predictability of Stock Returns by Fluck, Malkiel, and Quandt (1997). Unfortunately I could not find a full-text version of their paper available free online.