While the actions of economists and politicians can be influenced by facts, statistics or empirical predictions, narratives are becoming an increasingly important factor for the decision making in the field of economics and politics. Evaluating such narratives not at selective points in time but rather as a timeline can give us an insight on the effects of changing narratives on economic processes. We propose a model to detect two distinct types of temporal narratives by evaluating the relevance of entities in a timeline of newspaper articles. This methodology is based on the fundamental concept that all narratives are driven by and centered around certain entities. We provide a model to describe entity-based time dynamic media attention and detect both temporary (events) and permanent (structural break) changes of narratives by analyzing the number of appearances of an entity and the change in word frequency surrounding it. Our model detects several meaningful events and structural breaks, such as Mario Draghi’s well known “Whatever it takes” speech in 2012 or the change of narrative surrounding Wladimir Putin due to start of the Russian-Ukrainian war in 2022. For instance, this enables us to detect the narrative shift contained in newspaper articles about the Russian Federation from being a German business partner and gas trader to being called a war mongering regime.