2. What explanatory factors are associated with the varying rates of participation, violence, and saving of other individuals (the Schindler-syndrome)?

Diverse Characteristics of Mass Killing

Estimation

All variables here – whether or not activities involved torture, violence, without violence (abduction, gathering of individuals), death and saved (i.e., whether or not individuals were protected or hidden) – are “binary” in nature. In our statistical investigation, they were coded zero if the event did not have the attribute (for example, if there was no torture involved) and a one if it did. We approached this analysis in this manner because multiple activities could take place at the same time and place.

To conduct our analyses, we employed a technique called logistic regression. This lets the researcher discover the relationship between variables of interest and the probability that an event had the attribute designated by the variable. Only several hypotheses about the causes of killing and saving were explored (see our research proposal for a more complete discussion).

The discussion that follows is based on Table1, but highlights only the statistically significant variables.

The Occurrence of Torture

From our preliminary results, events that took place in private spaces were much more likely to involve torture than those taking place elsewhere. Events that took place in community, professional private and religious spaces were all about 10% more likely to involve torture than other spaces. Events in health spaces such as hospitals were slightly more likely to involve torture, but the overall magnitude of the change in the probability is relatively low. The same can be said of the negative effect of prefecture population on the likelihood of torture.

The Occurrence of Events Without Lethal Violence

Hardly any of the factors that we have analyzed thus far have any influence on this type of behavior. Thus far, events taking place in private spaces were about 10% more likely to occur without violence. Events in community and health spaces were slightly more likely to occur without violence.

The Occurrence of Events with Violence

Events taking place in private, community, commercial and religious spaces were slightly more likely to involve violence than those taking place in other spaces, though the magnitude of this change was quite small.

The Occurrence of Death

Events taking place in commercial and religious spaces were about 90% more likely to involve death. Those taking place in community and health spaces were 55%-65% more likely to involve death than those not taking place there. Events in Private and Professional Private spaces were about 30% more likely to involve death than those not taking place there.

Commune and prefecture population increased the likelihood that death was involved, but only slightly. Communes with the highest proportion of Tutsi were about 6% more likely to have events involving death than were communes with the lowest proportion of Tutsi. The reverse is true of the proportion of Twa population.

The Occurrence of Saving or Protecting

Events taking place in private, community, professional public, religious and health spaces were all about 5% more likely to be saved events than those not taking place in those locations. Commune and prefecture population increased the probability of people being saved by about 2%. Communes with the greatest proportion of Twa were about 5% less likely to have a saved event than those with the lowest proportion of Twa. The reverse is true of the proportion of Hutu, but the magnitude is about half as large.

Finally, and most significantly, communes with the highest proportion of Tutsi were 35% more likely to have saved events than those communes with the lowest proportion of Tutsi.