Attack and Conquer? International Anarchy and the Offense-Defense-Deterrence Balance

Scholars and strategists have long argued that offense is easier than defense in some periods but harder than defense in others. ¹ In the late 1970s, George Quester and Robert Jervis took the first steps toward systematizing this claim. ² Since then, the debate about thecauses and consequences of the offense-defense balance has been one of the most active in security studies. ³

If the relative efficacy of offense and defense changes over time, states should be more vulnerable to conquest and more likely to attack one another at some times than at others. Specifically, when offense is easier than defense, defenders' military forces should be more likely to collapse or surrender when attacked, and defenders' political leaders should be more likely to surrender sovereignty in response to military threats. Thus states in offense-dominant eras should be conquered—involuntarily lose the monopoly of force over all of their territory to external rivals—more often than states in defense-dominant eras. ⁴ Given their heightened vulnerability to conquest, states in offense- dominant eras should also be more likely to act on the doctrine that the best defense is a good offense. That is, they should attack one another—conduct offensive military operations against states that have not previously attacked them—more often than states in defense-dominant eras.

Although these points have been widely discussed, neither proponents of offense-defense arguments ⁵ nor their critics have tested the effects of the offense-defense balance on the historical incidence of attack and conquest. ⁶ Instead, they have examined its effects on the offensive or defensive character of military doctrine; ⁷ the extent of arms racing, prevalence of cooperation, and nature of alliances; ⁸ and especially the incidence, severity, and duration of war. ⁹

Attention to war is easy to understand, for it is one of the most destructive human activities. But testing offense-defense arguments on the incidence ofwar is problematic because these arguments suggest that wars in offense-dominant eras should have many attacks while those in defense-dominant eras should have few. Moreover, they suggest that attacks in offense-dominant erasshould frequently culminate in conquest while those in defense-dominant eras should rarely do so. In defense-dominant eras, states with state-of-the-art capabilities should be able to declare war then wait to counterattack without imperiling their survival. Thus, although there may be many declarations ofwar, there should be few attacks. Moreover, attacks should not result in conquest unless states have outdated capabilities or strategies. In offense- dominant eras, by contrast, security should come from attacking first. Instead of declaring war, states should engage in surprise attacks, wars should involve a number of attacks by a variety of states, and attacks should frequently result in conquest.

Because offense-defense arguments suggest there should be less variation in war than in attack and conquest, the most direct way to test the core claims of these arguments is to examine the effects of the balance on the incidence of attack and conquest. That is my purpose in this article. But constructing an effective test requires more than applying existing arguments to new dependent variables. It also requires redefining the concept of the balance to distinguish between defense and deterrence dominance, as well as operationalizing the historical balance deductively, at the operational level of analysis, and in purely technological terms.

I begin by explaining the need for these refinements to the logic and application of offense-defense arguments. Then I elaborate and operationalize an argument about the technological sources of the offense-defense-deterrence balance and the balance's effects on state vulnerability to conquest and propensity to attack other states. Next, I derive hypotheses about the historical incidence of attack and conquest from this argument, as well as from alternative arguments about technology, relative capabilities, duration of great power status, and audience costs.

Then, using the...