Violent mortality during the internal armed conflict in Guatemala and its impacts on population health

Collecting data during armed conflict is difficult and can be dangerous—violence might limit physical access to communities or disrupt communication channels, organizational budgets might be strained, and threats of retaliation from perpetrators might prevent reporting. Furthermore, violence may occur in areas too remote to allow comprehensive data collection, and the local populations may mistrust officials and therefore choose not to report violence. These disruptions impact data collection by state agencies charged with generating vital statistics and administrative health data, as well as that of civil society and international organizations documenting human rights abuses. As a result, some victims’ stories are never recorded, and those whose stories are documented may still be missing critical demographic information about the victim, details about the perpetrator, or other contextual information about the death that allows for its accurate classification. The resulting databases are seldom a complete enumeration of all violent deaths that occurred, and generally are not a statistically representative sample of the entire victim population. Using these data alone to draw population-level inferences about conflict mortality or to examine downstream population health measures without first addressing the violence that was not recorded by any of the sources can result in incorrect conclusions about the true impacts of the violence.

This study focuses on violent mortality—civilian killings and disappearances—during the internal armed conflict in Guatemala between 1960–1996. We build on the original estimates published by the Comisión para el Esclarecimiento Histórico, incorporating a new dataset that became available after the end of the commission’s mandate and implementing current state of the art methods for record linkage, missing data imputation (MI), and multiple systems estimation (MSE). Additionally, we consider the implications of different MI and MSE algorithms on the results, contributing to the study to the interaction between MI and MSE, which has not been fully explored. Finally, we use the estimation results to study the impacts of the conflict-related mortality on measures of population health, such as changes in life expectancy and lifespan inequality. If the data satisfy the assumptions, and the model adequately represents the dependence between lists and variation in the probabilities of observation, MSE may provide an unbiased estimate of the total number of victims of lethal violence, including those who were never documented. In this best case, estimated total numbers of victims allow us to move beyond the lower bounds on the impacts of lethal violence on population health, yielding a more complete view of the probable impacts of the violence had on population health during the conflict. Our findings have implications for accountability and historical memory, provide important context for understanding ongoing violence in Guatemala, and offer a comparison for the study of other mortality crises, such as the COVID-19 pandemic.