The thesis will consist of three individual papers, which will be combined into a cumulative dissertation. The research focuses on the question of how the different granularity of information affects the economic value of stock return forecasts. The first part of the thesis investigates how reducing the level of information of stock return forecasts affects the allocation and performance of an optimal portfolio. The second part focuses on increasing the information level of return predictions by taking into account additional, model-inherent information, that is generated by the return prediction model, and how this additional information affects the portfolio allocation and performance. Finally, part three examines how the predictive power of stock returns and their economic value is changed by taking into account additional, external information.