Alexander B. Downes is a predoctoral fellow in national security at the John M. Olin Institute for Strategic Studies at Harvard University. He is a doctoral candidate in political science at the University of Chicago and will be a postdoctoral fellow at Stanford University’s Center for International Security and Cooperation in 2003-04.

Why do states sometimes target and kill civilians intentionally in war? According to one estimate, of the nearly 110 million warrelated deaths in the twentieth century alone, 56 percent were civilians. Noncombatants died in the greatest numbers in the twentieth century, but the practice of brutalizing civilians in wartime is, as Caleb Carr notes, "as old as warfare itself." Athens and Melos, Rome and Carthage, and the rampages of Attila the Hun, Genghis Khan, and Tamerlane across Europe and Central Asia all suggest that, as one scholar puts it, "the average war ... has not been very 'just', as far as the killingof unarmed civilians was concerned." The recent attention devoted to "collateral damage"-civilian casualties resulting from attacks aimed at military targets-obscures the fact that such incidents actually kill relatively few noncombatants. Most civilians die because someone is trying to kill them. The question is, why?

Despite the extent of the carnage, our understanding of why civilians die in such large numbers during war remains limited. A number of recent studies, however, suggest that the major perpetrators are authoritarian states. According to emerging conventional wisdom, autocracies-free of domestic institutional constraints and prone to be run by leaders not shy about killing to get their way-are more likely to engage in genocide or mass killing of civilians. A combination of liberal norms and democratic institutions, on the other hand, constrain democracies from brutalizing noncombatants. Liberal norms, for example, prohibit the harming of innocent individuals, even enemy civilians in wartime. Democratic institutions force leaders to be mindful of public opinion in making foreign policy choices. Just as fighting a costly war-or even worse, losing one-is a policy likely to result in a leader's repudiation at the ballot box, killing large numbers of civilians in combat operations is liable to provoke public censure, possibly leading to the loss of elected office by the officials responsible. Finally, liberal democracies are presumably the type of regime most sensitive to international ethical norms prohibiting intentional or disproportionate harm to noncombatants because democracies themselves abide by similar norms domestically.

Some recent evidence supports the view that democracies go to great lengths to protect civilians from harm in wartime. The number of civilians killed by aerial bombardment in the last three wars fought by the United States, for example, has dropped precipitously: the figure was 65,000 in the Vietnam War; fell to 3,000 in the first Gulf war, and reached only 500 in the war over Kosovo (although it increased slightly to 1,000 to 1,300 in Afghanistan). In none of these campaigns, moreover, were civilians the direct object of air attack. Wars waged by other democracies in the past thirty years show similar restraint. In the 1982 Falklands War, for example, Britain killed a total of five Argentine civilians. Examples can also be found in the distant past: U.S. forces did not target noncombatants in the Mexican War (1846-48), nor did the British do so in colonial wars when they confronted adversaries who fought conventionally with regular armies, such as the Sikhs in India or the Zulus in sub-Saharan Africa.

Disturbingly, however, examples of democracies victimizing civilians in war-sometimes on a massive scale-spring readily to mind. The Anglo-American naval blockade in World War I contributed to the deaths of half a million German civilians, while the Allied strategic bombing campaigns of World War II killed hundreds of thousands of Axis noncombatants. Israel used targeted terror against Arab civilians in its war of independence, resulting in the flight of about 750,000 people, and killed as many as 10,000 Lebanese noncombatants in the siege of Beirut in 1982. Nor have democracies always been humane in their conduct of counterinsurgency campaigns, as exemplified by the Second Boer War (1899-1902), U.S.-Filipino War (1899-1902), wars by the French and Americans in Indochina (1945-54 and 1965-73), and the French-Algerian War (1954-62).

Indeed, my own research shows that when a large sample of wars is examined, democracies are slightly more likely than non-democracies to target civilians, although the difference is not great enough to be significant in statistical terms. I examined every interstate war since 1815, a total of 97 armed conflicts involving 316 countries. The results of this analysis were surprising: 26 percent of the democracies intentionally targeted civilians, compared to only sixteen percent of the autocracies.

Thus we are left with a paradox. Liberal democracies respect the rights of their citizens to be free from arbitrary violence domestically and loudly condemn the human-rights violations of other states. But when democracies go to war, they, too, may target and kill enemy civilians, not only through occasional massacres but also by systematic killing, sometimes in large numbers. What explains this puzzling empirical finding? Why do all types of states seemingly behave similarly when it comes to their treatment of enemy civilians in war?

States tend to target noncombatants in wartime when defensive advantages on the battlefield cause (or threaten to cause) armed conflicts to become costly, protracted wars of attrition. States are sensitive to costs and seek to minimize the loss of human life and the time it takes to achieve their war objectives. Furthermore, leaders and ordinary citizens are predisposed to favor the lives of their own people over those in other states. According to a recent survey by the International Red Cross, overwhelming majorities of people in all types of societies and cultures believe that purposefully targeting civilians or failing to discriminate between combatants and noncombatants is wrong. But when their state actually gets into a fight, these attitudes change: people become willing to breach the laws of war and sacrifice the lives of enemy civilians if doing so promises to ensure the survival of their own state or preserve the lives of their soldiers. Political leaders, too, are pulled towards conserving the lives of their own citizens, even if this means sacrificing innocents on the other side. As George Kennan once put it, "Government is an agent, not a principal. Its primary obligation is to the interests of the national society it represents." The interests of mankind as a whole rate, if anything, a distant second.

But why attack civilians? How does targeting them contribute to victory? When human costs escalate or appear to loom on the horizon, or when wars bog down into prolonged struggles, governments seek to shorten the war and lower their own casualties, while still endeavoring to attain their objective. Unfortunately for civilians, targeting noncombatants offers a way to achieve both of these goals. On the one hand, civilians' morale and willingness to support a war effort is thought to decrease when they are attacked. This logic was most famously articulated by Italian interwar air theorist Giulio Douhet who-ironically, in hindsight-called for punishing air attacks on urban centers as a way to shorten wars and make them more humane. The population, in his view, unable to withstand the terror and destruction of aerial attack, would rise up and demand an end to the war. But targeting civilians can degrade not only the enemy's will to continue the struggle but also its ability to fight. The rationale behind the American firebombing of Japanese cities in the Pacific War, for example, was to incapacitate Japan's war economy not by destroying factories, which were too hard and costly to hit, but by killing industrial workers so there would be no one to work in the plants. Finally, targeting civilians also accomplishes the straightforward goal of removing them from a particular piece of territory.

In practice, these various motivations translate into two general forms of civilian victimization. Coercive victimization occurs in protracted wars of attrition, both conventional and guerrilla, as costs escalate and each side searches for ways to convince the other to abandon the struggle. Violence against civilians in this scenario is not intended to wipe out entire populations but rather is a tool to coerce the enemy leadership to give up. In conventional wars coercive victimization takes the form of aerial bombardment of civilians or naval blockade and siege designed to starve civilian populations. In guerrilla wars, conversely, the tactics of coercive victimization are those that prevent or deter noncombatants from supporting the insurgency, such as population concentration, destruction of crops, village burnings, or massacres.

Eliminationist victimization, on the other hand, occurs in wars of conquest or when a war erupts between intermingled groups within one state. It tends to happen immediately rather than escalate over time because one or both sides expect high future costs of occupation, believing that a particular group cannot be reconciled to its rule and represents a permanent threat of revolt. Therefore, the attacker moves preemptively to eliminate that group via expulsion or, in extreme circumstances, mass murder. Eliminationist victimization often takes the form of ethnic cleansing but also occurs in the advanced stages of guerrilla wars when one side believes that the civilian population is ineradicably committed to supporting its opponent.

Certain factors tend to favor defense over attacking, which results in costlier wars. When these factors exist, civilian victimization is more likely. They include: (1) an even military balance; (2) technologies, terrain, or strategies that favor defending over attacking, such as guerrilla or siege warfare; (3) the escalation of states' war aims, which induces fiercer resistance in the opponent; (4) one side's intention to annex and colonize its adversary's territory; and (5) the belief that a population cannot be controlled.

This understanding of civilian victimization helps explain why combatants target civilians in some wars but not in others. In World War I, for example, all participants expected a short, victorious war but in reality were confronted with a protracted and bloody slugfest. Unable to prevail in the trench warfare on the continent and unwilling to abandon their goals of overthrowing Prussian militarism and restoring a balance of power on the continent, British leaders tightened the naval blockade of the Central Powers, denying food to hungry civilian populations. Germany, too, attempted to sever Britain's lines of communication with U-boats, albeit with much less success, and launched the first extended strategic bombing campaign, killing 1,336 British civilians. Similarly, in World War II, faced with a costly invasion of Japan's home islands, American leaders turned to strategic bombing as a means to achieve that goal at low cost, a campaign that culminated in the atomic bombings of Hiroshima and Nagasaki.

America's recent wars, by contrast, have all been fought at low cost against weak opponents: 147 Americans died in the first Gulf war, 38 were killed in Afghanistan, while none at all perished in Kosovo. Moreover, an adherence to international norms protecting noncombatants in war has spread and grown stronger. At present, therefore, the normative environment and America's overwhelming power both point in the same direction, allowing the United States to act on its liberal beliefs and eschew targeting civilians intentionally. As long as the United States continues to fight low-cost wars, civilians will remain off-limits. Should the United States actually encounter an opponent who can inflict casualties on American forces, however, civilians may again come into the crosshairs.