I find that firms meeting quarterly earnings forecasts earn positive abnormal returns of about 20 percent over the next 3 years. Partitioning these firms based on forecast errors in the prior quarter (t-1) reveals a surprising W-shaped relation between abnormal returns earned over the 3 years after quarter t (when forecast errors are exactly zero) and forecast errors in quarter t-1. For the subgroup of firms with negative forecast errors in t-1 (actual earnings less than forecasts), the relation is negatively sloped with firms with large negative forecast errors exhibiting very positive abnormal returns (30 percent) and firms with smaller negative forecast errors exhibiting negative abnormal returns (-10 percent). For the subgroup of firms with positive forecast errors in t-1 (actual earnings greater than forecasts), the relation is positively sloped with firms with large positive forecast errors exhibiting very positive abnormal returns (35 percent) and firms with smaller positive forecast errors exhibiting negative abnormal returns (-20 percent). I investigate whether these abnormal returns are due to revisions in expected future flows and/or revisions in perceived risk. I find that the former effect is far more important than the latter effect. This methodology has potential uses beyond the context of this study: it can be used to identify the extent to which risk shifts are responsible for abnormal returns associated with other stock price anomalies. Subsequent to completing this study I was informed of substantial mechanical effects when using IBES summary database that could negate the many of the results described above.