Empirical researchers of civil war rarely collect data on violence themselves and instead rely on other sources of information. One frequently used source is media reports, which serve as the basis for many ongoing data projects in the discipline. However, news reports rarely cover a conflict comprehensively and objectively and may therefore be prone to various reporting issues. This article provides an analysis of the accuracy of information given in news reports. In particular, if focuses on two types of "hard facts" that event data sets require: the location of an event and its severity. By linking media reports to firsthand accounts from a military database, the article does two things: (1) it analyzes the determinants of inaccuracy and confirms the expectation that events with a low number of observers tend to have higher reporting inaccuracies and (2) it assesses the magnitude of these inaccuracies and the implications for conducting empirical analyses with media-based event data.