Security Studies Program Seminar

War by Other Means: the Fate of Civilians in Times of War

Benjamin Valentino
Dartmouth College

April 19, 2006

Why do some wars kill more civilians than others? First I'll review the literature and the theoretical ideas driving my project. The basic idea for this project grows out of my previous project on the cause of genocide and mass killing more generally. I didn't expect to write or study any of these subjects. Since the early 1990s, the question of genocide has been in the news due to the violence in the former Yugoslavia. Even Schelling says that the kind of violence that nuclear weapons can commit is not new; genocide is the low-tech but functional equivalent of nuclear war. Genocide is a fairly regular event.

We have spent a lot of time trying to explain nuclear war and how it might happen. But what about genocide? All of the main work had been done by sociologists and psychologists, or there was individual case work without theory. Violence was explained as an upwelling to hatred from society, which seemed odd to me given the previous studies of war I had done. US and USSR had been quietly plotting to kill one another's civilians, even though no one said hatred was welling up from the two societies. Why did we jump to the assumption that genocide and mass killing are fundamentally different? What is the strategic thinking at work behind it? There were plenty of examples of societies that experienced mass killing even when there was not a long history of hatred and discrimination. Killers and victims at times were of the same ethnic, social, and political groups. Even in Yugoslavia and Rwanda, using any objective indicator, they didn't lose a lot worse than many other societies that didn't have that level of violence. And the converse was true, there are many societies that had much worse relations without genocide, such as South Africa or Jim Crow era in the American south. Most of this violence was initiated by relatively small elites at the top and carried out by astonishingly small groups, less than 1% of adult male populations. The story of neighbor killing neighbor is not true. Not much evidence the public supported violence. What are the conditions under which leaders decide that this is a good strategy for achieving what they want?

Why do some interstate wars kill more civilians than others? I thought this would be a good place to do quantitative research because the data had already been gathered by places like COW. Surprisingly, most of this data was completely useless. Most of the quantitative work on war has been focused on the causes. Most of the literature is focused on the causes of war. But I want to understand the consequences of war. What COW provides you is the total troops across all countries in any given year. But it won't tell you how many are serving or how many are killed. COW also focuses on casualties, “battle related deaths.” There is not a lot distinguishing WW II from the Falklands. There also isn't much telling you where the COW data comes from or how to reconcile discrepancies between that and what you find. I gathered data on casualties for all wars since 1899, and there is a big skew in the data. There is high variability in the balance between military and civilian deaths and across time, in terms of where the wars occurred.

One set of arguments say that whether states target civilians depends on internal attributes of the states themselves. There are some good states and bad states that discriminate between military and civilians, and there are bad states that don't. Different people have defined these different ideas, but three main ideas are most prominent. The hypotheses are:

Hypothesis 1: states will kill fewer numbers of enemy non-combatants during war if they have ratified international treaties that restrict the killing of non-combatants. The treaty restrains them and reflects their intentions. Measure: ratification of Hague (pre-1949) or Geneva Conventions (post-1949).

Hypothesis 2: Highly democratic states will kill fewer numbers of non-combatants during war than other kinds of states. Measure: combined polity score.

Hypothesis 3: states will kill greater numbers of enemy non-combatants during war when the majority of their population is comprised of different major cultural groups than their adversary's population. Measure: Huntington's civilization coding (1 if different civilizations, 0 if same).

If you want to understand killing of civilians in war, you should be focused on the specific characteristics of the wars that states find themselves in. This violence is an instrumental tactic leaders use to achieve certain goals. Leaders target these groups because they see in cold calculated terms that doing so will help them win the war at the lowest cost. By killing enemy civilians in large numbers, they will coerce surrender from the other side. Even if the military forces are capable of fighting in the field, they may give up if civilians have suffered. Second, because civilians are ultimately responsible for producing weapons, food, and other services required to continue the war, in certain kinds of wars, targeting civilian populations will be effective. By targets civilians you are actually affecting the outcome on the battlefield.

Every state wants to win wars it is in and at the lowest possible cost. There are two main factors that determine if this tactic can work. First, every state probably has some incentive to coerce the adversary to surrender early. But in some wars, civilians are playing a more critical role in the conduct of the war itself. The second factor is the availability of less violent strategies. Generally, states prefer not to do this, not just for moral reasons, although I think those do matter. Killing civilians involves risks and costs. Killing civilians is a big diversion if military forces are also coming after you. Also, there is a big risk that in attacking civilians you won't coerce the other side; you will actually increase resistance. In fact, it can provoke retaliation in kind, and your civilians can be attacked. So for all of these reasons, plus the kernel of moral logic, states will use this strategy primarily as a last resort.

There are three indicators:

Hypothesis 4: states will kill greater numbers of enemy non-combatants during the war the more they rely on military strategies of attrition or counter-insurgency. Measure: percentage of war devoted to attrition or COIN strategies.

Hypotheses 5: states will kill greater numbers of enemy non-combatants in wars of conquest and regime change than in wars fought for more limited aims. Measure: 1 if highest war aim is conquest of regime change, 0 if not.

Hypothesis 6: states will kill greater numbers of combatants and kill them at a higher rate during wars, the longer the duration of war. Measure: length of war in days.

Control variables:

Hypothesis 7: states will kill greater numbers of enemy non-combatants during war the greater their military capabilities relative to their adversaries. Measure: index of total military personnel, expenditures, etc. relative to one another.

Hypothesis 8: states will kill greater numbers of enemy non-combatants during war the greater the population of their adversaries. Measure: total population of adversaries (logged).

The results show that good states/bad states story doesn't hold up well. Hypotheses 4-8 turned out too be much more persuasive. These factors go together.

So is this an effective tactic? Does it wind up backfiring because it diverts forces from the front, stiffens resistance, and prompts retaliation?

There are huge selection effects in choosing a strategy. The wars where these tactics are used are the most desperate wars. This is a tactic of last resort. So we shouldn't be surprised that using this tactic is not correlated with success. It's not surprising that this tactic doesn't work, because people who use it tend already to be losing. We will be controlling for some of these factors in our future research.


Rapporteur: Caitlin Talmadge

back to Wednesday Seminar Series, Spring 2006