In addition to the problem of simply defining civil war, conflict researchers who are interested in studying the duration of civil war must address a number of measurement problems, which can lead to selection bias. Moreover, the characteristics of the data on civil war complicate duration analysis, including: repeated events, competing risks, non-proportionality, unmeasured heterogeneity (a general problem related to omitted-variable bias), and selection bias. This essay identifies controversies in the field and suggests ways to improve this research. We compare the effect of estimation issues with the effect from data collection biases. Our finding is that among the two, the lack of data seems to produce the larger problem.
Earlier versions of this paper have been presented to the 2003 and 2004 Annual Meeting of the Midwest Political Science Association, Chicago, IL, 3-6 April 2003; 15-17 April 2004, and at the the “Resources, Governance Structures, and Civil Wars Workshop”, part of the Joint Session of Workshops of the European Consortium of Political Research, Uppsala, Sweden, 13-18 April 2004.