Reliable forecasting is a key success factor of most organizations and companies. Where historical data is not available, the forecasts address questions in the far future, information is dispersed regarding location and form, or conflicting goals or values have to be considered, judgmental forecasting methods based on experts or the crowd are typically applied. However, several judgmental forecasting methods exist and each method has some individual weaknesses. Delphi-Markets are an integrated approach of prediction markets and Real-Time Delphi studies. Depending on their implementation, they allow to combine several properties of both approaches in order to overcome individual weaknesses. Three different ways to integrate the method are presented and discussed in this work. In order to better understand challenges and potentials of Delphi-Markets, the FAZ.NET-Orakel was instantiated and made publicly available for evaluation and improvement of an exemplary Delphi-Market under real-world conditions. In this context, four proposed improvements for the integrated approach were evaluated in four research projects. These projects corre ... mehrspond to the four sources of forecasting error according to the Judgmental Forecasting Improvement Model, introduced and derived in this dissertation as well. On the one hand, these improvements deal with common problems of prediction markets: Cognitive errors, such as partition dependence, and motivational errors, such as manipulation and fraud. On the other hand, these include common problems of Real-Time Delphi studies: The selection of experts for Delphi studies and retention during the surveys. As contributions to the overall IS research derived from the examinations of the Delphi-Markets and this dissertation, design principles for two extensions (social Real-Time Delphi and a crowd-based approach for manipulation and fraud detection) are formulated, implemented, tested, and suggested for application. Further, the role of complexity and expertise in the occurrence of the partition dependence bias is examined and a selection approach for experts for Delphi studies based on trading data is suggested and evaluated.