I find that analysts who forecast company earnings more accurately also issue more profitable stock recommendations within the earnings forecast period. I observe that this effect holds where the analyst is following a specific stock (analyst-firm level), but does not extend across that analyst's entire portfolio (analyst level). At the analyst-firm level, the average factor adjusted return achieved by the most accurate earnings forecasters exceeds the matching return of the least accurate earnings forecasters by 0.57 per cent per month, during the period of this study, which is April 1994 to March 2006. At the analyst level, the corresponding excess return differential is 0.44 per cent per month, but is not statistically significant. In addition, I document persistent differences in the ability of security analysts to forecast company earnings over time, above that predicted by chance. Previously accurate (inaccurate) forecasters tend to continue to issue accurate (inaccurate) forecasts in the future. Hence, investing according to the current stock recommendations of previously accurate earnings forecasters should deliver abnormal returns relative to those less accurate. However, I do not find statistically significant evidence of this. The strength of persistence regarding earnings forecast accuracy is not sufficient. Finally, I find that analysts are biased towards large, low book-to-market growth stocks, as well as those stocks with strong price momentum relative to the market.