What is a civil war? When are episodes of internal armed conflict classified as civil wars as opposed to coups, genocides, terrorism, or other adjacent phenomena?  Empirical analyses of the causes or consequences of civil war are premised on our ability to clearly define and measure the concept we are trying to explain, yet scholars classify instances of violent conflict into different categories based on fairly arbitrary criteria and the term ‘civil war’ is now used interchangeably to refer to conflicts as large as Syria’s conflict from 2011 to 2019 as well as any minor armed conflict causing more than 25 battle deaths in a given year.

In a new article, Nicholas Sambanis (Penn) and Jonah Schulhofer-Wohl (University of Virginia) argue that civil war is an instance of “sovereignty rupture” and is inherently a polity-level phenomenon.  Sovereignty rupture means that civil wars reflect a polity-level challenge to sovereignty and should be considered as events that happen to a society in the aggregate.  Civil wars should not be conceptualized as dyadic conflicts between the state and an armed group, which is the current trend in the literature.  These conceptual issues become even more pronounced due to ambiguities regarding the coding of when civil wars start and end and how civil wars are distinguished from other forms of violence within the same country.  This new article addresses these issues and applies the concept of sovereignty rupture to code a new list of civil wars from 1945-2016. The new data are then used in comparisons to other commonly used measures in the literature. Sambanis and Schulhofer-Wohl use their new data (SSW dataset) to replicate several studies that have used civil war data from the Armed Conflicts Dataset (ACD) and find that the results of those studies are often overturned or weakened when the analysis is based on the SSW dataset as opposed to ACD.  The findings from this article suggest the need for greater deliberateness in data selection in civil war studies, focusing on the fit between the question of interest and the concept of civil war that is underlying a given dataset.

As an example of the differences that emerge when we vary coding rules in collecting data on civil war, see the Figure below from the article by Sambanis and Schulhofer-Wohl.  The figure plots civil wars during the period from 1990-2016 according to the SSW coding criteria vs the ACD dataset (Uppsala/Peach Research Institute Oslo) using either the ACD annual 25-fatality threshold (first panel) or the ACD annual 1,000 death threshold (second panel).



The differences are apparent.  As might be expected, results from many prominent studies that use the ACD dataset are not robust to changing the list of civil wars used in the analysis.  Seven studies are replicated, on the effectiveness of power-sharing after civil war; on the impact of civil wars on post-war educational expenditures; on whether petro-states are more prone to civil war; and whether ethnic economic and political inequality increase the risk of civil war.  For a full discussion of these studies and a presentation of the results of these replications, read the full paper here.

This new list of civil war comes with extensive documentation and code to reproduce the replication analyses, which will be made available at the journal’s replication materials website.  Finally, this analysis builds on an earlier article published in 2004 in the Journal of Conflict Resolution, which you may access here.