Herding is extensively discussed in the literature, yet the results are so far inconclusive. The question of how additional information is processed remains unclear. This article presents the outcome of a novel two‐stage forecasting experiment that examines how capital market experts process ‘additional’ information. In the first stage, we test whether experts are engaging in herding or anti‐herding. In the second stage, we analyze how additional information is reflected in the experts' forecasting behavior in terms of hit rate, miscalibration, width of forecasting interval and confidence level. The paper delivers evidence on the presence of anti‐herding, which also tends to impact forecasting accuracy negatively. Moreover, additional information decreases the width of the forecast interval, leading to a lower hit rate. As a consequence, the level of overconfidence increases, while the level of forecasting confidence remains stable.