An Earnings Yield Anomoly

[Colt McCoy]

A hedge fund I work with has been backtesting various investment criteria and found an oddity about earnings yield (E/P).

Basically we divided a group of small cap stocks into deciles based on earnings yield. As you'd expect, the stocks with the highest earnings yield have tended to do better with returns declining as the yields get lower.

The oddity, however, was the stocks in the very last decile with the lowest yield (heavily negative earnings). These stocks have also tended to substantially outperform. Any ideas why? I have a theory but would like to hear some other ideas first. Ideas?