Replication Data for: Geography and the Certainty of Terrorism Event Coding

While event data provides researchers with insight into contemporary security threats, many are built upon secondary sources that may insert bias into empirical studies. Specifically, we argue that one form of bias—description bias—can be conditional on an event's characteristics or locale, thus influencing the certainty an observation is coded as an act of terrorism. We find that, using the Global Terrorism Database's own variables, attacks on civilians, particular types of tactics, and attacks that occur closer to a populated place are more likely to be coded as terrorism. These findings speak to broader conceptual issues in terrorism research and reiterate the need for researchers to evaluate the validity of their data before making claims.

尽管事件数据集为研究者提供了洞察当代安全威胁的视角，但多数此类数据集依托二手数据源构建，可能为实证研究引入偏倚。具体而言，我们认为存在一类偏倚——描述偏倚（description bias），其表现依赖于事件的特征或发生地点，进而影响某一观测事件被编码为恐怖主义行为的确定性。我们的研究发现，依托全球恐怖主义数据库（Global Terrorism Database）自身的变量，针对平民的袭击、特定类型的袭击战术，以及发生在人口聚居区附近的袭击，被编码为恐怖主义行为的概率更高。上述研究结论直指恐怖主义研究领域更广泛的概念性议题，并再次强调研究者在得出研究结论前，需先评估所使用数据的有效性。