WikiSummary, the Social Science Summary Database

Gartzke: War is in the error term

From WikiSummary, the Free Social Science Summary Database


Gartzke. 1999. War is in the error term. International Organization 53 (summer): 567-587.

In Brief

If we take Fearon's arguments seriously, we arrive at a paradox: There must be some element of randomness that explains war. Consider the universe of possible conflicts (every possible dyad). Theories about capabilities rule out some of them. Theories about motives rule out more. Theories of bluffing (e.g. Fearon's) rule out some more. But we've still got a few cases that we cannot theorize about within a rationalist framework. These theories all give only necessary conditions, not sufficient conditions, for war. In cases that satisfy all these necessary conditions, the occurrence of war is random. Moreover, if we accept Fearon's argument about uncertainty, then we cannot possibly identify a sufficient condition for war within a rationalist framework--because we can never predict whether bluffing will be met with a raise or a fold.

When we do a study, we should have peace as Y, not X. All our IR theories identify necessary (not sufficient) conditions for war; turn this around, and it means they all identify sufficient conditions for peace. So when one of these is present, you will have peace; when none is, you will have war or peace.

Gartzke vs Fearon

Gartzke's argument refines Fearon's (and includes and excellent summary of Fearon's work). You must understand Fearon to understand Gartzke.

Gartzke summarizes Fearon's three Xs: incentives to bluff, commitment problems, and issue indivisibility. Fearon argued that issue indivisibility (Fearon's X3) has no empirical effect. Gartzke adds that commitment problems (Fearon's X2) don't get us far either. They might help explain war initiation, but they would make it harder to explain war termination.

For example, suppose A (declining state) attacks B (rising state) to prevent B from becoming dominant. Bargaining can't prevent this conflict, because B can't credibly commit not to attack A in the future. Suppose A cripples B, ensuring that it won't outpace B. But the conflict won't end if the thing that started it hasn't gone away; and A now cannot credibly commit not to abuse B. So B would have to fight to the death, which rarely happens. So preventive wars will always be total war. Since states would know this, there would be a built in deterrent against preventive wars, making bargaining more likely to succeed.

Thus, X2 has problems. As Gartzke puts it: "The logic of preventive war thus implies one of three conclusions. First, if a solution exists to the commitment problem and states are assumed to be fully informed, then ex ante bargaining can occur for the same reasons discussed elsewhere by Fearon. Second, if a solution exists but states are hampered by uncertainty and incentives to bluff, then preventive war is really just a special case of Fearon's first explanation. Third, if no solution to the commitment problem exists, then the costs of such contests are presumably extremely high. States are likely to anticipate the destructiveness of preventive war and avoid it in all but extreme situations." [see Wagner 2000]

Main Argument

"Given uncertainty and incentives to bluff, there are NO factors that lead the mechanisms explaining the occurrence of war to systematically produce one outcome over another. Properly understood, the causal mechanisms that explain the occurrence of war from crises in large samples are stochastic."


  • Y = War (as opposed to a bargain)
  • X = Fearon's X1, incentives to bluff. But this yields only a probabilistic argument. We cannot know whether war will occur in a specific case; war includes a stochastic (random, thus unknowable) element.

Role of Uncertainty

A central point: If uncertainty matters, then something must be random.

Gartzke claims that uncertainty (bluffing) can lead to either war or peace. Thus, there is some random element in warfare. For example, assume that two states are bargaining over territory. Each has some "reservation price," which is the amount that it must be offered (in bargaining) to make the bargain better than the war. For example, if a state values a territory at $50, but war will cost $30, then it will accept any bargain that leaves it with at least $30. Assume at least one state has an incentive to bluff about its reservation price:

  • "Given the preceding conditions, there are at least two reasons why war cannot always occur. First, some states will be satisfied with the offers they receive in bargaining. They will prefer accepting the bargains to fighting. If states fail to make offers or make offers that no opponent will accept, then at least some of the time such states unnecessarily bear the burden of costly contests. Since states are always better off accepting the outcomes of costly contests ex ante, any offer within the Pareto space (the range of outcomes between each state's "ideal" outcome) will be preferred by at least one opponent to the option of war. Second, if uncertainty and incentives to bluff always lead to war, then states that prefer a bargained outcome based on their capabilities and resolve will simply reveal themselves to their opponents. Yet all states prefer a bargain to a costly contest that yields the same outcome."
  • So within a rationalist framework, you can't go any further toward explaining war.
  • But Gourevitch (1999) argued that people rely on ideology when they run out of information. Could we look at ideology to predict whether a state raises or folds when facing a bluff? Well, to a degree. But the players also know their ideologies. And if, based on these, they could be sure whether war would occur, then they wouldn't fight--there wouldn't be uncertainty. So ideology can only be yet another necessary condition (like capabilities, motive, and bluffing) that reduces the number of situations in which conflict is possible.
  • Couldn't we theorize about how these beliefs affect whether we think the other guy is bluffing (and by how much)? Gartzke concedes this, to a point: "Explaining why some states overestimate reservation prices and others underestimate them explains why some states fight and others do not and constitutes an extension and refinement of Fearon's first rationalist explanation for war." But there will still be a random element remaining.

Implications for Future Research

Study what creates uncertainty and what doesn't. This might bring you towards domestic politics (e.g. democracies are more open, so there's less uncertainty when dealing with them; also, ideology, bureaucratic procedures, and so forth might help reduce uncertainty). But we still can't eliminate the uncertainty, we can only reduce it.