In lieu of an abstract, here is a brief excerpt of the content:

ONE OF THE FIRST QUESTIONS to arise in this inquiry concerned the character of civil war recurrence. Is the phenomenon of recurrence any different from that of onset? If not, then it would make more sense not to pursue an inquiry into recurrence at all but to draw on the more numerous and better-trodden ground of civil war onset in thinking about recurrence. If recurrence does differ from onset, then how does it differ? Although recurrence may differ from onset in its manifestation , its consequences, or its modes of termination, I am here concerned with the causes or risk factors associated with recurrence. One can imagine that societies that have experienced civil war once may be more prone to turn to violence again, because of the availability of arms, of former militarized networks, of prior training and capacity among rebel groups, and/or of the socialization toward war. One can also imagine that war weariness, scars of warfare, social and individual traumas, the depletion of resources and social capital brought about by prior warfare, and other aftereffects may make recurrence less likely. Microlevel violence (Kalyvas 2006) may aggregate or articulate in different ways as well, with unforeseen consequences. This chapter seeks to do two things. First, I endeavor to answer this question : Do the causes or risk factor of civil war recurrence differ from its onset? I explore this question through a quantitative exploration of the question drawing on well-known data sets. I draw on research conducted in collaboration with the economist John Schmitt (Call and Schmitt 2009), using linear regressions to determine the extent to which the recurrence of armed conflict differs in its causes from that of initial onset. If recurrence were not to differ appreciably from onset, then this inquiry would best be served by analyzing all civil war onsets. After explaining the pool of cases, I proceedto specification of variables and then present results. Second, the chapter demonstrates why one should be deeply skeptical of the findings of much of the quantitative literature in the absence of other methods A Quantitative Analysis and the LimitsThereof 2 Is Civil War Recurrence Distinct from Its Onset? Is CivilWar Recurrence Distinct from Its Onset? 51 that can confirm or provide greater causal sustenance to correlation-based theses . The quantitative exercise also demonstrates the severe limits of quantitative methods generally for explaining the phenomenon of civil war recurrence, not to mention possible responses to it. This chapter details several of the reasons readers should be extremely cautious, though not disregard entirely, findings about armed conflict that rely exclusively on quantitative methods. The second section of this chapter calls into question the weight one should give to poverty and to primary export dependency as factors in analyzing and responding to civil war recurrence. A Regression Analysis of Civil War Recurrence How does one determine whether civil war recurrence differs from its onset? I draw on a strong data set to seek to replicate the findings of one serious quantitative study and then to see how that analysis applies specifically to recurrence. This section presents the pool of cases, data sets, and variables, as well as the results of the analysis. The Pool of Cases Several databases of civil wars exist. One of the earliest and most frequently utilized is the Small and Singer data set, used by Walter (2004), for instance. However , this data set was created initially for interstate wars, and its criteria for classifying the initiation and termination of civil wars were not as nuanced as other data sets. The Small and Singer data also suffer from a large number of missing values and imputations for important variables over periods of many years. James Fearon, in his collaboration with David Laitin (Fearon and Laitin 2003), improved and updated the Small and Singer data set, drawing on others.1 In his collaborations with Michael Doyle, Nicholas Sambanis (Doyle and Sambanis 2000, 2006) prepared a strong data set with detailed analysis of specific cases, but this set has the disadvantage for many purposes of not containing data on countries (and years) without civil wars. The Fearon and Laitin (2003) data set includes data on countries and years with no civil war, which permits the introduction of more numerous variables. Therefore, I here draw on Fearon and Laitin’s data set, which includes country-years from 1945 to 1999, with a total of 6,610 observations, including both interstate wars and colonial wars. Their definition of civil war includes (1...

Share