Article Preview
TopIntroduction
One of the most widely studied questions in contemporary political science is why terrorism occurs in certain places and times but not others. Since the year 2000, an article on this topic has appeared in, on average, every fourth issue of the American Political Science Review, every fifth issue of International Organization, and every third issue of The Journal of Conflict Resolution. There are currently at least nine peer-reviewed journals dedicated exclusively to the study of this phenomenon and dozens of terrorism databases and datasets have been constructed for associated analysis (SCImago, 2022; Bowie, 2021; Chenoweth, 2019). Despite the scale of this combined research effort, scholars have reached little consensus on the empirical determinants of terrorism. In the analysis of political, economic, and social factors, studies have recorded contradictory findings in signs, size, and significance in their findings. Why do studies on the same topic report such divergent results? Are these differences driven by some underlying heterogeneities and causal complexities or by differences in scope conditions, the usage of disparate datasets, or other some other elements of research design?
This article outlines a systematic approach for researchers to conduct cross-dataset comparisons, isolate sources of variation in empirical findings, and determine the robustness or uniqueness of determinants in the context of terrorism and other forms of political violence. To demonstrate the utility of this process, the author assembles three of the largest and most widely used terrorism datasets and applies the Cross-National Data on Sub-National Violence (xSub) data integration protocols to process event data into a common event typology with consistent categories and units by space (country, province, district, grid cell), time (year, month, week, day), and target (Zhukov et al., 2019). After processing the events onto a common typology, these standardized measures are combined with data for economic conditions, regime type, demographics, and weather and fit empirical models across a common set of spatial-temporal coverage and scales. Once these data are harmonized onto a standardized event typology with consistent categories and the analysis is confined to shared spatio-temporal dimensions, the findings exhibit greater consistency. Past divergences observed in cross-national studies may, in part, reflect that relationships between common correlates of terrorism are context-specific, and vary across different time periods and geographical locations examined in individual studies.
Terrorism research would also benefit from a process by which scholars can more confidently compare findings between terrorism datasets and general conflict event datasets that capture other forms of political violence, often perpetrated by the same actors. Our understanding of terrorism remains incomplete when studied in isolation as, “most uses of terror actually occur as complements or as byproducts of struggles in which participants…are engaging simultaneously or successively in other more routine varieties of political claim making” (Tilly, 2004). In fact, many hypothesized determinants of terrorism are shared with other forms of violent contention. Therefore, a framework, such as that presented in this article, which facilitates such a comparison, has the potential to provide additional insight into whether certain correlates explain the occurrence of terrorism or simply rebellion in general (Beuno De Mesquita, 2005).
Consider recent events in Afghanistan. The Taliban, an entity that has oscillated between status of an incumbent government and an insurgent organization, has employed a spectrum of political violence. Some of these actions, particularly indiscriminate attacks on civilians, align with most scholarly definitions of terrorism. Consequently, these events are likely to be reflected in terrorism event databases. However, what about other forms of violence in which the Taliban is engaged, such as skirmishes with Pakistani border guards or the former Afghan Army? What about the violence they pursue now that they have regained control in Afghanistan? Most terrorism databases exclude acts carried out by state forces. Therefore, if we exclusively consider events involving the Taliban using only terrorism event databases, we are likely to obtain an incomplete picture regarding the determinants of when they use terrorism or other forms of violence.