In this paper, we analyse the final decisions for merger cases prepared by the European Commission (EC) since 1990 and build a unique subsample for all non-cleared cases. These incorporate all merger notifications which were either withdrawn by the notifying parties or have been prohibited by the European Commission.We find a sudden decline in prohibitions and withdrawals of cases since 2002 and explore three judicial defeats of the European Commission as determining factors behind these developments. We also find a higher likelihood of withdrawal or prohibition if cases are registered in sectors which incorporate firms in the business of information and communication or transportation and storage. When classifying the documents with a supervised machine learning algorithm, we are able to automatically identify the cleared versus the non-cleared cases with over 90% accuracy. Finally, we find that network effects, high market shares and the risk of collusion are the main competitive concerns which contribute to prohibition decisions in the information and communications sector.