Traditionally, research on civil conflict has maintained a cross-national focus, tracking macro-level trends across states, but overlooking variation within them. A growing movement in sub-national research has disaggregated actors and actions to a more granular local level, but mostly within a single country or conflict. With few attempts to generalize beyond the idiosyncrasies of an individual case, subnational conflict research has produced a series of contradictory findings, impeding the accumulation of knowledge. As a result, social scientists have been unable to answer the kinds of questions that activists, policymakers and ordinary citizens care about most, such as when and where violence is likely to occur, how it are likely to unfold, and – importantly – whether the lessons of previous conflicts apply to new ones. Our aim is to eliminate the disaggregation-generalizability tradeoff and enable a more theoretically comprehensive and rigorous approach to this field of research, by developing a unified platform for Cross-National Data on Sub-National Violence (xSub).
To a greater extent than macro-level work, disaggregated, sub-national research on civil conflict has faced steep problems of external validity, perpetuating disagreements over how idiosyncratic these findings really are. The discovery of a positive relationship between, say, repression and dissent in country A is not evidence of a similar relationship in country B, yet most sub-national analyses stop at the borders of country A, without attempting to validate results with data from other conflicts. Also unknown is how robust empirical results are across sources and measurement strategies – whether media-generated data agree with archival sources, or whether manually-classified event reports tell the same story as data collected with automated techniques, like natural language processing and machine learning.
The chief barrier to generalizability is not a lack of data on violence or covariates in country B – in many cases these data already exist and are in the public domain. The problem is that no one has yet undertaken the entrepreneurial effort to merge and combine these disparate sub-national conflict datasets into a unified, analysis-ready format, with consistent theoretical constructs, definitions, measurement and levels of analysis. Without such an effort, the field cannot move forward.
xSub addresses this problem explicitly. xSub is a web-based program that pulls together dozens of existing sub-national databases, and aggregates the relevant conflict events and covariates to consistent units of analysis across countries and conflicts. As a public good, xSub will significantly reduce the barriers to comparative sub-national research, empowering researchers to better situate their theoretical investigations and quickly construct custom, analysis-ready datasets. Similar initiatives have long existed for macro-level cross-national conflict data, most notably the EUGene software by Bennett and Stam. No resource of this kind currently exists for sub-national conflict data.