This paper develops a technique to infer the accuracy of analysts' forecasts of different components of earnings when databases contain forecasts of only bottom line (or limited component) earnings forecasts. We focus our analysis on three components - sales, operating profit percentage, and effective tax rates - and find that larger changes in all three components are associated with larger contemporaneous forecast errors, but that the change in operating profit percentage appears to be the most difficult to forecast. This difficulty in forecasting changes in the operating profit percentage appears to be concentrated in those observations in which net income decreases from the prior year.