Civilian killings and disappearances during civil war in El Salvador (1980‒1992)

• Main Authors: Amelia Hoover Green, Patrick Ball
• Format: Article
• Language: English
• Published: Max Planck Institute for Demographic Research 2019-10-01
• Series: Demographic Research
• Online Access: https://www.demographic-research.org/volumes/vol41/27/

<b>Background</b>: Debate over the civilian toll of El Salvador's civil war (1980-1992) raged throughout the conflict and its aftermath. Apologists for the Salvadoran regime claimed no more than 20,000 had died, while some activists placed the toll at 100,000 or more. But no rigorous estimate was available to settle the question, which has returned to prominence because of newly opened war-crimes trials. <b>Objective</b>: Estimate the total number of civilian killings and disappearances in El Salvador during 1980-1992. <b>Methods</b>: We match and merge four lists of reported civilian killings and disappearances, gathered by multiple organizations - approximately 20,000 unique episodes in all. We use log- linear modeling to estimate the number of uncounted cases from patterns of list overlap. To account for uncertainty in model selection, we use Bayesian model averaging over the set of decomposable graphical models. <b>Results</b>: We estimate that there were about 71,629 (60,326, 83,775) civilian killings and disappearances during the conflict, or about 1-2Š of El Salvador's prewar population. Correlations between recorded deaths and estimated deaths are imperfect over both time and space. <b>Conclusions</b>: Unusually, our conservative global estimate is similar to nonstatistical 'expert guesses.' However, the estimated distribution of civilian killings and disappearances does not closely track the reported distribution over time or space. Our approach achieves high quality estimates, even in the presence of relatively sparse data. <b>Contribution</b>: This is the first rigorous estimate of civilian killings and disappearances in the Salvadoran civil war. Both the global estimates and estimates for specific spatiotemporal strata will inform ongoing accountability processes.