or reduce the adversary’s ability to fight often becomes irresistible in order to economize on
manpower/equipment losses. Stalemated wars of attrition also trigger escalation to civilian
victimization out of a sense of desperation that every possible means must be employed that
might lead to victory or avoid defeat. Finally, governments engaged in state-building efforts—
involving conquest, annexation, and possible colonization of territory—may victimize groups
they believe pose a threat of future insurrection.
The findings of the quantitative analysis strongly support the view that involvement in
costly, protracted wars of attrition leads states to target enemy civilians systematically and to kill
larger numbers of them. Similarly, states that conquer enemy territory populated by people
believed to be hostile tend to engage in violent persecution of those people in order to eliminate
the threat they pose. Surprisingly, and contrary to the findings of some previous analyses, liberal
democracy neither lowers the probability of civilian victimization nor the number of enemy
civilians killed in interstate wars. No democracy has ever killed more than one million foreign
noncombatants in wartime—Nazi Germany and Imperial Japan are the only states to hold that
dubious distinction—but even these autocratic outliers do not tip the scales in favor of
democracy. In fact, democracies’ greater than average sensitivity to suffering high casualties
may make them more likely to strike at enemy civilians in wars of attrition. Finally, clashes of
civilizations are not associated with higher rates of civilian targeting or casualties.
The remainder of the paper is organized as follows. First, I define civilian victimization
and describe the four different ways I operationalize it in the analysis. Second, I discuss the
democratic peace, clash of civilizations, war of attrition, and territorial expansion explanations
for civilian victimization, and describe the measures I use as independent variables. After a brief
description of the dataset and the rules for inclusion in the analysis in the third section, I present
the results of the statistical analysis in the fourth section by dependent variable: civilian
4
Convention