Abstract
Previous research suggests that first-strike advantages can create commitment problems that lead rational leaders to opt for war, even when both sides are perfectly informed. In this article, we develop a formal model that extends upon this work. We include a more detailed account of war processes in order to determine what factors are most likely to cause first-strike advantages that lead to war. In particular, we decompose first-strike advantages into three underlying dimensions: tactical offensive advantages, mobilization advantages, and the destructiveness of an initial attack. Our model also allows states to counterattack, rather than, as typical in the literature, assume a single-stage conflict. By studying the interaction of these factors, we uncover surprising results about wars produced by first-strike advantages. First, offensive advantages can lead to war in the absence of mobilization advantages. However, they only do so when successful attacks can destroy an extremely large proportion of the enemy's military power. In contrast, mobilization advantages always increase the likelihood of war. Surprisingly, mobilization advantages are particularly likely to lead to war when combined with defensive advantages. We demonstrate that these mechanisms can explain war initiation through an examination of the 1967 and 1973 Arab-Israeli Wars.
Countries are often concerned that adversaries will attack by surprise, thus gaining crucial advantages in the early stages of war. For example, North Atlantic Treaty Organization (NATO) member-states have recently become concerned about their ability to defend the Baltic states against a surprise Russian invasion (see Shlapak and Johnson 2016) and are deploying a multinational deterrent force in response (Economist 2016; Emmott and Siebold 2016). During the Cold War, many experts worried that a Soviet blitzkrieg attack could quickly overwhelm NATO forces in western Europe (for example Nunn and Bartlett 1977; Cohen 1988), although others believed that NATO could successful defend itself until American reinforcements arrived (for example, Mearsheimer 1982). One assessment began by stating forthrightly, “[t]he Western alliance cannot afford a surprise attack” (Betts 1981, 117). Similarly, concerns about first-strike advantages appear in discussions of cyber security (Saltzman 2013; Shaheen 2013; Farrell 2013; Slayton 2016) and debates over anti-access/area-denial and air-sea battle strategies in East Asia (for example, Keck 2013). The empirical record supports these concerns. By our count, slightly more than one-quarter (28.6 percent) of wars since 1918 began with some sort of surprise or preemptive attack.1 These include some of the deadliest wars in recent history. The Korean War, the Iran-Iraq War, and the Badme Border War have all begun through surprise attacks, representing three of the four deadliest conflicts since World War II.
Some of these wars may have been inevitable, driven by factors other than the existence of an opportune moment for a surprise attack. However, we focus on scenarios in which the potential benefit of the surprise attack is itself the proximate cause of conflict. War may occur because one or more states possess first-strike advantages—short-run benefits to initiating hostilities. Such situations are inherently destabilizing, as states may rush to use these advantages before they disappear or may preempt a rival state from gaining or making use of their own first-strike advantages.
Previous research, including that by Fearon (1995) and Powell (2006), has explored these mechanisms, demonstrating that first-strike advantages can cause rational actors to initiate wars. They do so by creating implicit power shifts that generate commitment problems. Such conflicts can occur despite perfect information on the part of both combatants. First-strike advantages may simply outweigh, or eliminate, the potential gains from a conflict-avoiding bargain (Leventoglu and Slantchev 2007, 757–59, 765–67; Tarar 2013; Wolford, Reiter, and Carrubba 2011, 561).
We refine current theoretical understandings of how first-strike advantages cause war. Why do states attack by surprise? When will states with first-strike advantages actually use them to attack immediately rather than negotiate? To answer these questions, we develop a formal model of conflict that extends previous work by including a more detailed account of war processes.
First, we consider multiple causes of first-strike advantages. States may be tempted to attack first because tactical or technological factors favor offense—the traditional “offensive advantage” considered in classic accounts of the security dilemma. States may also be tempted to attack first to make use of temporary advantages associated with operational surprise, such as the under-mobilization of their opponent. Finally, states may be tempted to attack first to preemptively damage their target's war-making capacity and tilt the balance of power in their favor, as in the example of Pearl Harbor.
Second, we allow war to extend more than one round. Doing so captures the idea that war involves back-and-forth exchanges and that fortunes on the battlefield may change midconflict. This allows us to explore how factors such as early mobilization advantages or destructiveness alter the course of a longer conflict.
By decomposing first-strike advantage into its more basic components and building in more battlefield realism than previous models, we uncover several surprising results about how these factors interact to incentivize fighting. We find the well-known argument that offensive advantages always favor war is not generally true. Offensive advantages only lead to war when the initial battles can destroy extremely large portions of the attacked state's armed forces. Instead, mobilization advantage provides the most consistent predictor of war initiation. War becomes considerably more likely when one side is more prepared than the other. In addition, when combined with a mobilization advantage, war occurs substantially more often alongside defensive advantages, which make “bite and hold” strategies viable. This causal mechanism is largely absent from previous accounts of first-strikes and rational incentives for war. We also demonstrate other interesting results: higher war costs sometimes make war more likely, and weak states may often initiate war against stronger adversaries.
First-Strike Advantages and War: Theoretical Foundations
Speculation that first-strike advantages could lead to war arose during the Cold War, particularly surrounding worries that a nuclear first-strike could destroy a country's retaliatory capacity (for example, Schelling 1960, 205–254). In the 1980s and 1990s, offense-defense theory hypothesized that blitzkrieg offenses could create incentives for preemptive war (Jervis 1978 186–99; Hopf 1991; Lynn-Jones 1995; Van Evera 1998; Van Evera 1999, 35–72, 117–239; Lieber 2000).2
Formal-theoretic analyses considerably deepened our understanding of first-strike advantages and embedded this concept more soundly in modern bargaining theories of war. Previous formal work has established that first-strike advantages cause commitment problems by creating implicit power shifts (Powell 2004; Powell 2006; Leventoglu and Slantchev 2007; Tarar 2016, 747–50). This fulfills one of Fearon's (1995, 401–4) conditions for rationalist war. By forgoing a surprise attack, a state allows bargaining or future war to occur under less favorable circumstances. Yet, empirical examination of these mechanisms has often focused on offensive advantages or the offense-defense balance, with, at best, mixed results (Fearon 1997; Gortzak, Haftel, and Sweeney 2005; Adams 2004; Lieber 2005; Nilsson 2012). In addition, some have criticized offense-defense theory for a lack of conceptual clarity. In particular, critics suggest it is difficult to differentiate offensive advantages from raw power, specific tactics, or military skill (Davis, Finel, Goddard, Evera, Glaser, Kaufmann 1998; Biddle 2001, 742, 746–47). We seek to provide a richer theoretical account of how battlefield parameters might affect the decisions of states to initiate war.
Causes of First-Strike Advantages
Previous examinations of first-strike advantages have generally conceptualized various terms (such as first-strike, offensive advantage) differently than how they are used in qualitative accounts of conflict and battlefield tactics. In particular, this research often uses the phrases first-strike advantage and offensive advantage synonymously, making it difficult to distinguish between different dimensions of the two concepts. Van Evera (1999, 37) defines a first-strike advantage as an advantage that “accrues to the first of two adversaries to use force.” In this usage, first-strike advantage acts as an umbrella term covering any setting where an initiator has a strictly higher probability of winning a war begun at a particular time than under other conditions. We believe this umbrella term can be conceptually distinguished from the battlefield factors that create the advantage. Throughout, we use the term surprise attack to refer to one state's actual aggressive action that occurs without warning. By contrast, we use the term first-strike advantage to refer to the underlying conditions that may make a surprise attack favor the initiator of conflict.
Theoretically, three factors exist that may create first-strike advantages. The first is offensive tactical advantages. While formal-theoretic work has generally used the terms first-strike advantage and offensive advantage interchangeably, considerable scholarship attaches a more specific meaning to the latter term. Jervis (1978, 187) defines an offensive advantage as when “it is easier to destroy the other's army and take its territory than it is to defend one's own.” According to this definition, offensive or defensive advantages would affect the outcomes of specific engagements rather than the war as a whole. Offensive advantage represents a tactical/operational parameter that defines when the aggressor has a higher likelihood of winning a given battle than a defender of similar military power (among others, see Van Evera 1999, 117–239; Biddle 2001; and Nilsson 2012, 470 for examples of this usage). With this connotation, offensive advantages are a force multiplier, where a given number of attacking soldiers provide greater value than an equivalent number of defending soldiers. Conversely, a defensive advantage would confer a similar force multiplier on the defending forces. Previous research suggests that defensive advantages occur more often than offensive advantages, with a three-to-one ratio of attacking to defending forces commonly cited as necessary for a successful attack (Mearsheimer 1989). When attacking forces have advantages over defense in individual engagements, first-strike advantages might arise in the overall war.3
However, offensive advantages are not the only factor that could create a first-strike advantage. Alternatively, a state may hold a mobilization advantage where they are temporarily more prepared for war than their opponent (for example, Van Evera 1999, 35–72; Tarar 2013; Lai 2004). For instance, if both sides need to mobilize reserves, gaining a head start on mobilization could temporarily allow one state to field more forces than the other. This would give them an incentive to strike immediately rather than allow their opponent to fully mobilize.
Finally, successful attacks may inflict large, disproportionate losses on the defender. Maneuver tactics, such as the German blitzkrieg, aim to disorganize and surround enemy formations. This eliminates their ability to function effectively and may capture large amounts of prisoners or equipment. Alternatively, air strikes may destroy key elements of the opponent's military power. This destruction occurred in the Japanese attack on Pearl Harbor in 1941 and the Israeli attacks on Arab air bases in 1967. The chance to eliminate portions of an opponent's power could create enough advantages in later battles to make striking first advantageous, giving a higher probability of winning the overall war.
Previous literature has studied these elements in isolation. We will include all three in a single model to study the interaction among these parameters. In some cases, the interaction of these parameters leads to surprising and novel results.
Changing Fortunes During War
After noting the distinction between first-strike advantage and its potential causes, a logical next step is to examine how and when these tactical parameters make war more favorable for initiators. Typically, formal models of first-strike advantages model the possible war outcomes with a single probability parameter. In these models, the state initiating a war through a surprise attack has a strictly higher probability of winning the conflict than if they begin the war at a later time, or if the other state attacks first (Fearon 1995, 401–4; Powell 2006, 184–85, Tarar 2013).4 This approach straightforwardly considers how various first-strike advantages, whatever their underlying cause, net out to produce an overall chance of victory.
However, as we discussed above, the battlefield factors that may create first-strike advantages—offensive advantages, mobilization advantages, or destructiveness—may operate differently at different stages of a longer-term conflict. For example, combat destructiveness may give one state a lasting advantage in later phases of a conflict, but would be irrelevant in a war lasting a single battle. Similarly, the possibility of mobilizing reserve forces means that the two sides’ relative power will change as the war progresses. Accordingly, to better analyze the interaction of these factors, we consider war as a longer process in which each battlefield factor has different effects at different stages of a conflict.
Indeed, in practice, wars generally involve more than one engagement and fortunes often shift during the course of a conflict., The attacked state failed to counterattack in only four of the sixteen wars we assess as having begun by a surprise attack (the Sinai War, the 1967 Arab-Israeli War, the 1982 Lebanon War, and the War over the Aouzou Strip).5 Previous work has recognized the theoretical value of considering changing fortunes on the battlefield. Some scholars have extended Fearon's (1995) model to include multiple rounds of combat to explore mechanisms such as information revelation (Wagner 2000; Filson and Werner 2002; Smith and Stam 2004), as well as various perfect information dynamics (Langlois and Langlois 2005; Slantchev 2003).
For similar reasons, our model includes multiple battles to capture the back-and-forth process of combat. This allows the various battlefield parameters to have varying effects at different stages of the conflict. Thus, our war phase remains probabilistic, but adds more complexity to allow nuanced interactions among parameters.
The Ultimatum Model
We now consider a formal model of war initiation that explores the factors that may create first-strike advantages and lead rational states to begin a war. We adopt a simple ultimatum framework that still captures the essential elements of Powell's (2006) analysis of power shifts. Centrally, both sides have opportunities to initiate war or to pass that opportunity to their opponent. This generates an incentive to preempt the opponent's first-strike with one's own.
Assume a revisionist state (R) desires some discrete territory with a total normalized value 1, of which it holds some status-quo portion q, with the remainder held by a target state (T).6 The revisionist may attack immediately to try to seize the entire territory or instead demand concessions from the target by threatening war. Note that, if the revisionist attacks before making a demand, it gives up any chance of a peaceful resolution. If it makes a demand, the target then has the option of accepting or refusing the bargain. If the target accepts the bargain, the status quo changes accordingly, and the game ends. Alternatively, if the target refuses the bargain, it may itself elect to immediately initiate war against the revisionist. Finally, if the target neither accepts the offer nor preemptively attacks (i.e., simply rejects the offer), the revisionist receives one more opportunity to either attack or back down.
We assume any war phase proceeds the same way, regardless of whether war begins with a surprise attack or after bargaining. If an initial attack by the revisionist succeeds, we assume the revisionist has captured the entire territory. In such a case, the target may choose whether to counterattack to try and regain the territory, after which the game ends regardless of the outcome.7 Each of these battles imposes costs, cR and cT, on the revisionist and target respectively and plays out probabilistically as described below. A war begun by the target plays out similarly, with an attack by the target followed by a potential counterattack by the revisionist. Figure 1 gives a game tree with simplified payoffs. Note that all values v indicate one side's total payoff for initiating a war (their probability of holding the territory minus any costs), and w indicates their payoff for being attacked. The value |${v_{{R_0}}}$|, for example, denotes the revisionist's value from attacking in the initial “surprise” period (which may differ from their value for attacking later).
Game tree, where v and w represent the expected value of attacking and being attacked, under various circumstances
Note: Because war is costly, any matched pair of v and w sum to less than 1.
Within this simple framework, we seek to parameterize the various war payoffs more explicitly. In particular, we want to allow multiple potential causes of first-strike advantages to operate interactively but distinctly within the same formal framework. Thus, developing a function for the probability of battle victory is particularly important. The probability of victory in each engagement follows the simple function: |$\frac{{Attacker\ Strength}}{{Attacker\ Strength + \mathit{De\!f\!ender}\ Strength}}$|. We then characterize the component parts of this function in terms of several parameters that describe the tactical setting. First, we normalize the target's capabilities to 1 and the revisionist's to a single parameter b. Thus, b amounts to the relative military balance or power ratio (|$b = \frac{{Revisionist\ Strength}}{{Target\ Strength}}$|).
Second, we operationalize the three potential causes as modifiers within this equation. Offensive advantages, as conceptualized in the nonformal literature, generally describe how many offensive troops are necessary to overwhelm a given defensive force. Accordingly, we incorporate offensive advantages into this probability as a force multiplier, f, that applies equally to both sides. In each engagement, the attacker's capabilities are multiplied by the offensive advantage (f), effectively representing how many defending soldiers each attacking soldier is worth. For a value of f = 2, an attacking force of size 1 requires 2 defensive soldiers (1*f) to equalize it, representing offensive advantage. Thus, values of f less than 1 represent defensive advantage. For example, f = 1/3 would characterize a situation where 1 defending soldier balances 3 attacking soldiers.
We operationalize strategic surprise with a separate parameter in this battle probability. We assume that a surprised combatant is under-mobilized while the attacker is fully mobilized, creating a mobilization advantage for the revisionist. For practical reasons, most countries only have a portion of their forces ready for war during peacetime and may only have a fraction of their total forces stationed near the disputed territory (Tarar 2013, 344; Lai 2004, 216). The initiator's ability to choose the location and time of their first attack may create similar short-run advantages. Thus, a state attacked by surprise may be less effective in proportion to its relative unpreparedness. We thus treat mobilization advantages as a second force multiplier. We multiply an unprepared target's strength by a mobilization factor, m ∈ [0, 1], representing the proportion of the target's forces ready for combat at the beginning of the war. However, mobilization advantages are fleeting; making a demand (or attacking) alerts the target to the revisionist's hostile intentions, giving them the opportunity to mobilize more fully. Thus, the mobilization advantage only accrues in the revisionist's first battle in wars begun by a surprise attack and disappears if not taken advantage of at that moment.
Finally, we allow the initial battle to alter the military balance of the two states. As previously discussed, the initial attack may destroy a large portion of the target's forces, giving the attacker an advantage in the subsequent counterattack. The destructiveness parameter, d ∈ [0, 1], represents the proportion of the attacked state's (target's if the revisionist starts the war, revisionist's if the target starts the war) forces destroyed in the initial attack. This means that the target's capabilities equal (1 − d) if it chooses to counterattack.8 For simplicity, we assume that the losses from the initial attack are the same whether or not it occurs by surprise or after attempted bargaining.
Solution
Because the game contains perfect information, we adopt a solution concept of subgame perfect Nash equilibrium. For any given set of parameter values, the solution is unique up to indifference.9 All outcomes can happen in equilibrium: war, settlement, and reversion to the status quo. Despite relatively simple foundations, decomposing the various aspects of first-strike advantages causes the solution space to become quite complex. War relates nonmonotonically to several parameters. The overall contours of the solution are nonetheless relatively intuitive. To describe the solution, we develop a series of conditions that determine the solution space.
First, note Lemma 1, which we term the credibility condition. This defines R's best reply at the final node of the game. R has a credible threat to attack after a rejected offer if their value of beginning a war without a mobilization advantage exceeds the status quo. Otherwise, the target will not accept any offer more favorable to the revisionist than the status quo.
R will attack following a rejected bargain if vR > q.
This implies Lemma 2 regarding T's decision to initiate war, which we term the preemption condition. If this condition holds, then T prefers to attack immediately after receiving an unacceptable demand rather than simply reject the demand and await a possible attack. The preemption condition can be true either because T itself wishes to adjust the status quo or because attacking is better than defending.
T will prefer attacking over rejection if vT > wT if the credibility condition is met, and if vT > 1 − q if not.
Lemma 2 then implicitly describes the target's reservation values for rejecting a settlement, giving Proposition 1, which describes equilibrium behavior in general terms. In essence, R can make a demand equivalent to T's expected value of war as defined by the previous Lemmas 1 and 2. R will attack if the value of attacking immediately exceeds what it can demand at the bargaining table.
R's optimal offer s* equals 1 − vT if the preemption condition is met, equals 1 − wT if the preemption condition is not met and the credibility condition is met, and equals q if neither the preemption nor credibility conditions are met. R attacks by surprise in the first period if |${v_{{R_0}}} > {s^*}$| and makes its optimal offer otherwise.10
Though the overall pattern of play is straightforward when expressed in terms of the various shorthand v and w terms, considerable complexity arises when we characterize these payoffs in terms of the actual battlefield parameters introduced above. First, let us consider the term vR; the expected value of an attack by the revisionist depends upon the target's willingness to counterattack.
The target will counterattack after any lost battle if |$\frac{{( {1 - d} )f}}{{( {1 - d} )f + b}} - {c_T} \ge 0.$|
Given Lemma 3, then |${v_R} = \frac{{bf}}{{bf + 1}}( {\frac{b}{{( {1 - d} )f + b}} - {c_R}} ) - {c_R}$| if the target counterattack condition is met, and |${v_R} = \frac{{bf}}{{bf + 1}} - {c_R}$| if not. By the same token, T's utility for being attacked, |${w_T} = \frac{1}{{1 + bf}} + \frac{{bf}}{{1 + bf}}( {\frac{{( {1 - d} )f}}{{( {1 - d} )f + b}} - {c_T}} ) - {c_T}$| if the target counterattack condition is met and |${w_T} = \frac{1}{{1 + bf}} - {c_T}$| if not.
Very similar conditions apply when the target initiates a war. Their value for initiation vT depends upon the revisionist's willingness to counterattack.
The revisionist will counterattack after any lost battle if |$\frac{{( {1 - d} )bf}}{{( {1 - d} )bf + 1}} - {c_R} \ge 0.$|
Lemma 4 implies that |${v_T} = \frac{f}{{f + b}}\ ( {\frac{1}{{1 + ( {1 - d} )bf}} - {c_T}} ) - {c_T}$| if the revisionist counterattack condition is met and |${v_T} = \frac{f}{{f + b}} - {c_T}$| otherwise.
Finally, consider the value of the initial surprise attack, |${v_{{R_0}}}$|; this occurs with mobilization advantage, altering its valuation. The target's counterattack condition remains identical, thus |${v_{{R_0}}} = \frac{{bf}}{{bf + m}}( {\frac{b}{{( {1 - d} )f + b}} - {c_R}} ) - {c_R}$| if this condition is met, and |${v_{{R_0}}} = \frac{{bf}}{{bf + m}} - {c_R}$| if not met.
Analysis and Discussion
This relatively simple model generates a highly complex solution space due to the interactive effect of the battlefield parameters on the various v and w payoffs and thus on the states’ equilibrium behavior. We now describe various interesting aspects of this solution space and characterize our substantive results.
Overall, our model provides substantial insight into how first-strike advantages may lead to war and reveals several surprising conclusions. Most notably, offensive advantages only lead to war in extreme conditions. In contrast, mobilization advantages always increase the likelihood of war and thus represent a more likely cause of war through first-strike advantages. In addition, when mobilization advantages are present, defensive rather than offensive advantages more often lead to war. The model also reveals several additional findings that are counterintuitive, particularly that higher costs can sometimes make war more likely and that weaker states may attack stronger opponents.
The Bargaining Range
To provide a general illustration of the relationship between the three battlefield parameters f, d, and m, we initially consider the simple question of when a bargaining range exists. A bargaining range fails to exist—and war is certain—when |${v_{{R_0}}} + {v_T} > 1$|. Here, no peaceful settlement is possible.11 Figure 2 shows how the various battlefield parameters may combine to eliminate any bargaining range when the balance of power is equal and war is costless.12 In all cases, a bargaining range exists below each line and fails to exist above it. As can be seen, mobilization advantages (smaller target m) are particularly likely to eliminate large swathes of the potential bargaining range. Also surprising is that a bargaining range exists where offensive advantages exist (f > 1) under many conditions, and sometimes defensive advantages eliminate the bargaining range. Figure 2 illustrates a more general result of the model—war initiation generally depends upon the interaction of various battlefield factors. In particular, the offensive advantage f exhibits nonmonotonic effects that depend strongly on its interaction with other factors.
Thresholds for existence of a bargaining range across ranges of offensive advantage and destructiveness, for varying levels of target mobilization, assuming b = 1, and cR = cT = 0.
Notes: Offensive advantage (f) is plotted on a log scale. For each value of m, a bargaining range exists below the line and fails to exist above it.
To illustrate in more detail the types of relationships that can lead to war, we will now present a series of illustrations that detail the major relationships in the game and discuss how certain parameter interactions can cause war outbreak.
Offensive Advantages Only Lead to War in Extreme Conditions
Most previous work that has examined the causes of first-strike advantages has focused on tactical offensive advantages (Van Evera 1999, 117–239; Powell 2006, 184–85). Our model shows that offensive advantages can only cause war under two restrictive conditions. First, offensive advantages can cause rational states to initiate war even absent a mobilization advantage (when m = 1). However, under no conditions can offensive advantages cause war unless also combined with high destructiveness. Successful attacks must drastically alter the balance of power in favor of the attacker for war to occur. Second, for offensive advantages to cause war, the target state must be willing to initiate war given the opportunity rather than simply reject an offer (the preemption condition is met). Thus, for offensive advantage to cause war, both sides must possess viable offensive options and prefer to use those options.
Figure 3 displays the war range over potential values of offensive advantages and destructiveness when no mobilization advantage exists (m = 1). Consistent with Figure 2, this shows that offensive advantages only lead to war when combined with destructiveness values close to 1. In other words, for offensive advantages to lead to war, a successful attack must destroy a massive proportion of the defender's military capacity. In addition, when m = 1, war only occurs when the preemption condition holds. This means that, if warned, the target would prefer to attack immediately rather than simply refuse the bargain and await a possible attack. If the preemption condition does not hold, R will always be able to do (weakly) better by making the offer s = 1 − wT. The requirement that both sides possess offensive options means that war occurs due to Powell's (2006) basic mechanism; by foregoing an attack, R passes the offensive advantage to T, who will attack, creating an inherent power shift.
Equilibrium outcomes across ranges of f (on a log scale) and d, assuming b = 1, cT = cR = 0.2, q = 0, and m = 1.
Thus, under conditions of high destructiveness, our analysis recovers the classic result that offensive advantages can lead to war. However, decomposition of first-strike advantage into three component parts reveals that this implicit power shift is not caused solely by tactical offensive advantages, but by the combination of offensive advantages and high destructiveness.
Why do offensive advantages not cause war when destructiveness is not at extreme values? The answer lies in the possibility of subsequent battles in our model. When offense is easy, so are counteroffensives. This means that any losses can be quickly and easily reversed as long as the attacked state retains a decent military capability. However, high destructiveness substantially reduces a state's ability to counterattack successfully after it is attacked. Thus, only when successful attacks dramatically alter the power balance in subsequent battles can offensive advantages lead to rational war initiation.
The empirical record seems to confirm that surprise attacks seldom inflict sufficient damage to end the war immediately. Of the sixteen wars we code as beginning through surprise attacks, in only four cases did the attacked state fail to counterattack.13 These are the Israeli attacks on Egypt in 1956 and 1967, the Israeli invasion of Lebanon, and the Chadian-Libyan War over the Aouzou Strip. In the first two cases, the Arab armies were routed and largely incapable of further resistance. However, in Lebanon, Syria could probably have continued fighting. Certainly, Libya could have easily mounted counterattacks against Chad. Moreover, in every case in which the defender did counterattack, the counterattack successfully reclaimed the captured territory. Thus, though high destructiveness may occur occasionally, attacks are unlikely to impose sufficient damage to either prevent or hold off a counterattack.
These attackers may have been gambling on inflicting sufficient damage to forestall a counterattack. However, the empirical frequency and success of counterattacks suggests that attackers do need to seriously consider how they will hold any captured territory when initiating war.
For war to occur under offensive advantages, attacks must destroy a high proportion of the enemy's military forces. In addition, as war requires the preemption condition to be met, both sides must be able and willing to attack. We believe that these conditions occur relatively rarely, as shown above. Thus, contrary to the arguments made in much of the existing literature, battlefield offensive advantages are relatively unlikely to lead to war, even when they exist.
Mobilization Advantages Always Increase the Likelihood of War
In contrast to offensive advantages, mobilization advantages always increase the likelihood of war. Our analysis reveals that temporary mobilization advantages are an extremely important and highly general determinant of wars driven by first-strike advantages. The mobilization term only affects the expected value of the initial surprise attack, |${v_{{R_0}}}$|. Note that this term is always decreasing in m (regardless of whether T counterattacks or not). Thus, the initial surprise attack always becomes relatively more attractive as target mobilization decreases, no matter what the values of other parameters. This is again consistent with Figure 2.
This does not mean that mobilization advantages always lead to war. Mobilization advantages always make war relatively more attractive, but may not make it more attractive than a settlement. Depending on the other parameter values, even a complete mobilization advantage (m = 0) may not make war preferable to a negotiated settlement. Nonetheless, mobilization advantages exert a monotonic effect under all conditions.
Naturally, mobilization advantages also increase the likelihood of war due to a power shift mechanism. By offering a bargain, R tips their hand regarding their intentions and allows T time to mobilize. This form of power shift always makes an immediate attack more attractive. Moreover, mobilization advantages can cause war initiation even when the opponent has no viable threat to attack (preemption condition not met). This contrasts with offensive advantages, which only cause war when both sides have incentives to initiate and under conditions of high destructiveness.
In addition to being a more general theoretical cause of war, significant mobilization advantages likely occur more frequently than the extreme levels of destructiveness necessary for offensive advantages to cause war. As previously noted, surprise attacks are unlikely to destroy extremely large proportions of enemy forces. However, states quite often initially have only a minuscule portion of their forces ready for combat in the relevant theater. For instance, during the Falklands/Malvinas War, Great Britain had fewer than one hundred soldiers initially deployed in the Falklands (Gamba 1987, 144), allowing them to be captured easily. Thus, it seems likely that mobilization advantages are a more common cause of war than offensive battlefield advantages.
Defensive Advantages Increase the Likelihood of War When Revisionists Have Mobilization Advantages
As previously noted, most theory14 has argued that offensive advantages typically make war more likely. Interestingly, our model uncovers an interaction between mobilization and defensive advantages. When the revisionist has a mobilization advantage, often defensive rather than offensive advantages increase the likelihood of war. Figure 4 displays the war range for a significant mobilization advantage of m = 0.1.
Equilibrium outcomes across ranges of f (on a log scale) and d, assuming b = 1, cT = cR = 0.2, q = 0, and m = 0.1.
Again consistent with Figure 2, war occurs most commonly under conditions of defensive advantage (f < 1), while offensive advantages often lead to settlement except at extreme destructiveness values. For war to be attractive when initial destructiveness is low, the revisionist must not only have a good chance of succeeding in the initial attack, but also must be able to withstand or deter a counterattack. Larger defensive advantages increase the revisionist's chance of the holding any territory once captured. At the same time, while offensive advantages may make it easier to win the initial battle, they would compromise the ability to retain captured territory unless paired with extreme levels of destructiveness.
Thus, defensive advantages can actually incentivize war by making “bite-and-hold” tactics viable. A state with a mobilization advantage can attempt to seize the territory from an underprepared opponent in the initial battle and then rely on tactical defensive advantages to increase their chances of holding it in the counterattack.
Higher Costs Sometimes Make War More Likely
The above results describe the interplay among the d, f, and m terms that decompose first-strike advantages. Our model also reveals two other nonobvious findings regarding war costs and the initial balance of power. We now consider findings related to these parameters.
First, higher costs can sometimes make war more likely. Counterintuitively, increasing war costs to either the revisionist or the target may cause war when lower costs would not.15 Figure 5 displays equilibrium outcomes across the ranges of both states’ costs for one set of parameter values. As can be seen, increasing the costs for both the target and revisionist can increase the probability of war in some cases.
Equilibrium outcomes across ranges of cR and cT, assuming b = 1, f = 0.5, m = 0.1, d = 0.3, and q = 0.
It is relatively intuitive that increasing the target's costs might incentivize war. Higher costs will deter counterattacks, making war initiation more attractive for the revisionist. If the target will not counterattack, the difference in R's expected value between a surprise attack with a mobilization advantage and an attack against a prepared opponent increases. Thus, in some cases, higher costs can incentivize war by changing the target's calculus on whether to counterattack.
More surprisingly, increasing the revisionist's war costs can also make the revisionist more likely to begin a war. This occurs because the revisionist's war costs affect whether the revisionist can make a credible threat to attack after a failed bargain (whether the credibility condition is met). As these costs rise, the revisionist may become unable to make a credible war threat. Thus, they may be unable to gain anything at the bargaining table. However, combined with a mobilization advantage, a surprise attack may still have a positive expected utility, reopening the war range. Thus, by making war threats incredible, increasing the revisionist's costs can prevent them from recovering the bargaining surplus in negotiations, shifting the equilibrium from settlement to war.
Thus, increasing costs for both parties can in fact incentivize war. Based on our survey of the literature, we believe that this effect has not been previously identified. This result is particularly interesting because the bargaining model is typically used to explain why war may occur despite its inherent costs, which should incentivize negotiations. In contrast, our model displays situations in which war occurs because the war costs actually hinder negotiations.
Weak States May Initiate War against Stronger Adversaries
A final interesting trend is that, for much of the war range, weaker states will attack stronger opponents. Figure 6 displays the relationship between the power balance and offensive advantage, demonstrating this finding. This also occurs because of the credibility condition. Weaker states may be unable to achieve anything at the bargaining table because they have difficulty credibly threatening war. Thus, a revisionist's only option to improve their position may be to attack by surprise. By contrast, increasing a state's strength may make settlement more likely, because the state anticipates receiving benefits from bargaining.
Equilibrium outcomes across f and the balance of power b, both on log scales, for cR = cT = 0.1, m = 0.1, d = 0, and q = 0.
Some previous literature suggested that war occurs most often when the relative power of potential adversaries is nearly equal (Bremmer 1992, 313–14, 327–28, 337–38; Geller 1993; Wayman 1996; Moul 2003; Reed 2003; Hwang 2010). Further, we would intuitively expect that substantially weaker states would view their chances as slim and be especially reluctant to initiate war. However, weaker states somewhat frequently initiate conflicts against significantly stronger adversaries. In sixteen out of fifty-six wars (28.6 percent) since 1918, the initiator had a Composite Index of National Capability (CINC) score less than half that of the country they attacked.16 Thus, our model can help account for the seemingly odd behavior of weak states attacking stronger opponents. Because weak states may be unable to gain much, if anything, at the bargaining table, taking advantage of a target's low mobilization by launching a surprise attack may be their only way to change the status quo.
Certainly, other mechanisms could account for war initiation by weaker states.17 If nothing else, occasionally informational problems may be so severe that the weaker state does not recognize the power asymmetry. In addition, Chan (2010) argued that weak states might initiate war to spur great power intervention. Kadera and Morey (2008) show that weaker states may prefer war when faced with guns versus butter tradeoffs, as war may be their only chance to compete with adversaries. However, we still find it notable that for much of the war range in our model weaker states are more likely to initiate war than stronger states. Accordingly, first-strike advantages, and particularly attempts at bite-and-hold tactics, present an additional explanation for why weak states may rationally attack stronger opponents.
Case Studies
The previous sections discuss factors that may influence states’ decisions when opportunities for advantageous first-strikes exist. Here, we demonstrate that the model and its findings can aid our understanding of real-world conflicts by examining two conflicts between the same combatants: the 1967 Arab-Israeli War and the 1973 Arab-Israeli War.18 The 1967 Arab-Israeli War demonstrates the interaction between offensive advantages and destructiveness. The 1973 Arab-Israeli War demonstrates the interaction between defensive advantages and mobilization. Together, these case studies demonstrate the importance of considering interactions among various components of first-strike advantages.
The 1967 Arab-Israeli War
On June 5, 1967, after three weeks of tension, Israel launched a massive air attack that destroyed most of the Egyptian and Syrian air forces on the ground. Israel then conducted major ground offensives that routed the Arab armies and captured the Sinai Peninsula within six days (as well as the West Bank and Golan Heights). This case presents a relatively clear instance in which offensive advantages led to war. Israel's air attacks and blitzkrieg offensives effectively eliminated the ability of the Arab forces to resist. Thus, Israel brought the war to a quick and favorable end despite Arab preparations. Israel's initial attack was not merely victorious but enormously damaging, representing the combination of offensive advantages and extreme destructiveness that can lead to war in our model. At the same time, Israel's inability to maintain its mobilization for an extended period also influenced their decision to attack, equivalently to problems in our model.
The terrain and Israel's force structure in this war created offensive advantages on the ground. The relatively open desert, particularly on the southern front against Egypt, was particularly conducive to rapid maneuver and blitzkrieg-style offensives. In addition, the Israeli army, with its highly proficient armored forces (Rabin 1996, 102), was particularly suited to taking advantage of this possibility. Less than a decade earlier, the Israeli armored corps had clearly demonstrated these offensive possibilities during their lightning victory in the 1956 Sinai War (Oren 2002, 11–12). Israeli leaders were confident that they could mount a successful offensive, but were continually worried about being vulnerable defensively (Oren 2002, 122–23). In addition, Israeli leaders expressed fear that Egypt would attack (Dayan 1976, 343–44; Rabin 1996, 71), although Egypt had in fact scrapped any offensive plans (Shemesh 2008, 197–204). Thus, Israel felt a greater imperative to attack first.
Israel did successfully damage its opponent's ability to continue fighting. This power shift occurred in two stages. First, Israel's surprise air attacks on June 5 essentially eliminated the Egyptian and Syrian air forces. This gave Israel complete air superiority for the rest of the war, providing them with a crucial advantage (Oren 2002, 176, 195). Secondly, the ground offensives routed the Arab armies, and particularly the Egyptian forces that represented Israel's primary foe. During the war, Israel destroyed about 85 percent of Egypt's military hardware (Oren 2002, 215–16, 258, 273, 305). With their military capabilities effectively destroyed, the Arab states essentially lost the ability to keep fighting. Israeli generals displayed confidence that they could achieve these results in the days ahead of the war (Oren 2002, 80, 151, 172.). Moreover, military plans focused first on destroying the Egyptian forces' ability to fight and only secondarily on capturing territory (Rabin 1996, 78, 86, 96–98, 101–3; Dayan 1976, 330–31, 339).
In short, Israel's ability to devastate Egypt's war-making capacity in their surprise attack played a major role in both their decision to attack and in the swift resolution of the conflict. This case fits the classic pattern of Powell (2006) and offense-defense theory in one respect: attacking dominated defending in this setting. Our model refines this explanation; we show that researchers must consider the high potential destructiveness of Israel's first-strike, in combination with offensive advantages, in understanding this and similar examples.
Finally, note that during the crisis leading to the 1967 Arab-Israeli War, Israel was unusually mobilized for war. With the mobilization of reserves during the crisis, Israel had mobilized a full one-fifth of their labor force. Therefore, the crisis essentially paralyzed the Israeli economy (Oren 2002, 17; Maoz 2006, 81; “Memorandum for the Record” 1967). Thus, Israeli leaders felt that time was not on their side even if Egypt did not attack, as demobilization would become necessary (Rabin 1996, 77, 85, 89; Dayan 1976, 288, 316, 345). This would eliminate Israel's temporary advantages, equivalent (in relative terms) to allowing Egypt to mobilize in our model.
The 1973 Arab-Israeli War19
A second conflict between Israel and Egypt (as well as Syria) demonstrates how a combination of mobilization and defensive advantages can cause commitment problems that lead to war. As previously noted, this represents a fundamentally different combination of battlefield factors than those considered by previous theories of first-strike advantages.
On October 6, 1973, Egypt and Syria launched a coordinated surprise attack on Israel to regain territories lost in the 1967 Arab-Israeli War. By 1973, Israel had clearly demonstrated military superiority over the Egyptian forces in previous conflicts such as the 1956 Sinai War, 1967 Arab-Israeli War, and the 1967–1970 War of Attrition, particularly in armored and air warfare (Bar-Joseph 2005, 15–17). Egypt elected to attack Israel in 1973 nonetheless. Their reasons and plans closely map onto our model parameters, particularly the surprising interaction of defensive advantages and low target mobilization that makes war attractive for weaker states. Interestingly, both sides perceived a defensive advantage with respect to the disputed territory, yet this did not prevent war.
Israeli Defensive Advantages
Even though the overall terrain was similar to that of the 1967 Arab-Israeli War, Israel's capture of the Sinai Peninsula changed the situation significantly. The new border on the Suez Canal presented a major defensive obstacle, hindering Egypt's ability to attack successfully without operational surprise. Concrete walls between one and three meters above water level lined the Suez Canal, making it more difficult to construct bridges and operate amphibious equipment. Along the bank, Israel had constructed an earthen berm up to twenty meters high. Egyptian infantry would have to scale this barrier before advancing inland and would have to blast gaps through it to enable heavier forces to cross. Thirty-five fortified strongholds provided protection to Israeli forces firing at the crossing units. Finally, the Egyptians believed that Israel had constructed a system to spew oil onto the canal that when lit would literally turn the canal into a burning moat (Shazly 1980, 7–9). Combined, these obstacles would make any heavily opposed crossing nearly impossible. Our model shows how this fact significantly contributed to Egypt's initial aggression. No threat of war could possibly be credible given that it would put Israeli forces on alert, allowing them to effectively use these defensive tools. Thus, Egypt could not possibly have achieved any of its aims at the negotiating table.
Mobilization and Surprise
While Egypt was in no position to demand a settlement from Israel, they were presented with a golden opportunity to strike against only a fraction of Israeli forces. Egypt held short-term advantages because of the Israeli Defense Force's (IDF) heavy reliance on civilian reserves. At the beginning of the war, Israel had only a single armored brigade and one infantry battalion deployed close to the canal and only two more armored brigades deployed in the entire Sinai Peninsula. Thus, at the time of the attack, Israel had only 451 soldiers and three tanks deployed along the canal, supported by a further seventy-five tanks and fifty artillery pieces to meet the Egyptian assault (Bar-Joseph 2005, 205–9). By contrast, a day into hostilities Israel had deployed an additional six armored brigades near the canal, increasing its number of tanks to about 960 (Shazly 1980, 236). These additional forces arrived only after Egypt's initial crossing.
Egyptian Defensive Advantages
However, Egypt not only needed to cross the canal successfully, but to hold the territory gained. The Egyptian plan relied heavily on infantry anti-tank weapons, particularly RPG-7 rocket-propelled grenades and AT-3 anti-tank missiles. These weapons, which allow individual soldiers to destroy tanks, are particularly effective as defensive weapons because infantry in defensive positions are less exposed to enemy fire. Egypt went to great lengths to bolster its anti-tank defenses. For instance, they stripped all units not directly involved in the initial crossings of these weapons to provide more for the initial assault waves (Bar-Joseph 2005, 24; Shazly 1980, 34). In addition, Egypt deliberately restricted their initial advance to three miles, where they would entrench their positions. This allowed the Egyptian army to concentrate their initial forces, and their anti-tank weapons, in limited areas, giving them their best chance of repelling the initial counterattacks (Shazly 1980, 34). Egypt also restricted their own advance to protect their forces from Israeli air attack. Between 1967 and the October War, Egypt had acquired advanced air-defense weapons, in particular SA-2 and SA-3 surface-to-air missiles (Shazly 1980, 21). Emplaced on the Egyptian side of the canal, these missiles effectively prevented the Israeli Air Force from operating near the canal. Thus, Egypt had significantly improved its defensive capabilities since the previous conflict. Egypt's careful attention to these details reveals how defensive advantages can incentivize war through bite-and-hold tactics.
While Israel ultimately breached the defensive line, their ultimate victory required a close and difficult battle. Israel also sustained heavy losses in their first attempts to recover the disputed territory (Adan 1980, 83, 153). Although Egypt's effort to hold the territory ultimately failed, their assessment of how the various tactical considerations made war a reasonable choice closely matches our model. Egypt's belief in the defensibility of their forward position, combined with the opportunity to attack an under-mobilized opponent, made this war seem like a reasonable gamble. Had Egypt not suffered their own significant losses during an ill-advised October 14 effort to advance further into the Sinai, it is plausible that they might have successfully held the captured territory (Maoz 2006).
Conclusion and Implications
We developed a formal model that extends upon previous studies of first-strike advantages by including a more detailed account of war processes. In particular, we decomposed the abstract concept first-strike advantage into three underlying dimensions: offensive tactical advantage, mobilization advantage, and destructiveness of an initial attack. We then examined how these causes interact in a model with multiple battles rather than just a single stage.
Past work sometimes treats first-strike advantage as interchangeable with these possible underlying reasons. Considering the various causes more explicitly in a simple crisis bargaining framework leads to several surprising conclusions. Most notably, while offensive advantages can lead to war when attacks are able to destroy large proportions of enemy capacity, offensive advantages do not create first-strike advantages under all conditions. By contrast, mobilization advantages do always produce first-strike advantages; they may lead to bite-and-hold tactics when combined with defensive advantages and thus make war more likely. In addition, we found that under some conditions weaker states are more likely to initiate war and that, in some cases, higher war costs may actually make war more likely.
The important role mobilization advantages play in our model suggests several directions for future research. First, our findings imply that we should pay additional attention to the determinants of military force structure (for example, Sechser and Saunders 2010) and how states may resolve the guns and butter tradeoff (for example, Kadera and Morey 2008). Similarly, our model may have implications for understanding the relationship between conflict and regime type (for instance, see Oneal and Russett 1999; Schultz 1999; Weeks 2012; Valentino, Huth, and Croco 2010) or economic systems (see Gartzke 2007; Barbieri 1996). It is possible that democracies or highly industrialized states are particularly prone to heavy reliance on military reserve systems and therefore are more likely under-mobilized. Similarly, states that rely on military alliances for their security may also be more vulnerable to the bite-and-hold tactics we describe, which suggests implications for the literature on coalitions, alliances, and alliance politics (Leeds 2003a, 2003b; Leeds and Savun 2007, Fang, Johnson, and Leeds 2014; Smith 1995; Weitsman 2003, 2004; Henke 2017). The model also has obvious implications for the importance of geographical terrain on conflict (Toft 2014; Carter 2010; Goddard 2006, Vasquez and Henehan 2001; Hensel and Mitchell 2005; Rasler and Thompson 2006), particularly terrain that tends to produce defensive advantages.
Future research can also consider the role that private information plays in creating first-strike advantages, thus expanding on previous literature on private information as a source of war (Fearon 1995; Leventoglu and Tarar 2008; Lindsey 2015). This similarly suggests the importance of intelligence success and failures (Bar-Joseph and McDermont 2017) in explaining war, particularly as there are always some signs of military mobilization or offensive preparations. Similarly, previous research suggests that intergovernmental organizations (IGOs) and peacekeeping missions may have significant effects on conflict (Shannon, Morey, and Boehmke 2010; Chapman and Wolford 2010; Fortna 2004), and both may also limit mobilization advantages. By providing information about military mobilization or introducing delays in war initiation, IGOs and peacekeepers may be able to prevent mobilization advantages from emerging and thus deter war.
Practically, our analysis suggests two alternate ways of deterring potential revisionists from initiating war. First, states could maintain sufficient defensive forces in disputed regions to make it difficult for a revisionist to capture the territory in the first place. While possible, maintaining the degree of mobilization necessary to deter attacks may prove difficult in practice. Second, states could maintain robust counteroffensive forces to recapture any lost territory, thus reducing the chances that a revisionist could both capture and retain disputed territories. A possible added advantage of this second option is that counteroffensive forces are less vulnerable to destruction in the initial battle than defending forces stationed in the disputed territory. Thus, to deter attacks on the Baltic states, NATO might find developing robust counteroffensive options more beneficial than stationing large defensive forces in the Baltics.20 Similarly, the model helps to understand contemporary concerns about potential anti-access/area-denial technologies and strategies in East Asia or the Persian Gulf. As anti-access/area-denial technologies tend to increase defensive advantages, other countries’ development of these technologies would limit the United States and its allies’ counteroffensive options. Thus, these technologies might incentivize bite-and-hold tactics and encourage aggression by regional powers.
Footnotes
We draw this conclusion by coding whether wars in the Correlates of War database began with either a direct surprise attack on the armed forces of one of the combatants or the seizure of territory by stealth that led directly to armed conflict. The wars coded as beginning by surprise are the following: the Hungarian Adversaries War, World War II, the Korean War, the Sinai War, the Assam War, the 1967 Arab-Israeli War, the Bangladesh War, the 1973 Arab-Israeli War, the Sino-Vietnamese Punitive War, the Iran-Iraq War, the Falklands/Malvinas War, the War over Lebanon, the war over the Aouzou Strip, the Gulf War, the Badme Border War, and the Kargil War. These were coded based on information in Sarkees and Wayman (2010) and Phillips and Axelrod (2005). There are several additional wars that could plausibly be described as beginning with surprise attacks, but are less clearly so than those listed.
Wagner (2007) provides a recent counterpoint to the common theoretical claim that offensive advantages favor war, arguing that by limiting the risk of total defeat, defensive advantages may reduce the costs of conflict and thus make war more likely. Our analysis concurs with this revisionist take on the older conventional wisdom, although via a different causal mechanism.
In our analysis below, we consider offensive advantages as a symmetric parameter that applies to both parties equivalently, reflecting, for example, the current state of military or operational technology. Settings with asymmetric offensive advantages are no doubt possible empirically, but would essentially be incorporated into the basic power ratio. That is, if one side enjoyed both an offensive advantage when attacking and a defensive advantage when attacked, this is little different than saying that it is simply more powerful than its opponent.
The one major exception to this trend is Leventoglu and Slantchev (2007) who do model a surprise attack war as consisting of a series of battles, but do not unpack the battlefield parameters we consider here.
As with the above identification of wars begun by surprise attacks, we draw this claim from Sarkees and Wayman (2010) and Phillips and Axelrod (2005).
We assume the disputed issue is a piece of territory both because territory is more amenable to direct conquest through surprise attacks and because previous research has suggested that territorial disputes are particularly likely to lead to war (for example, Vasquez and Henehan 2001; Hensel and Mitchell 2005; Rasler and Thompson 2006).
We also considered a more complex Markovian framework in which further battles are possible, with similar results as those below.
For modeling simplicity, this assumes that the attacker takes no losses in the initial battle. Thus, the parameter d essentially represents how much more losses the target takes than the revisionist. Formally, if we allow both sides to take losses in a revisionist attack dR and dT, |$d = \frac{{{d_T} - {d_R}}}{{1 - {d_R}}}$|. If the target begins the war, this would be |$\frac{{{d_R} - {d_T}}}{{1 - {d_T}}}$|.
When R's optimal offer is the status quo, they may be indifferent between this and unacceptable offers that result in rejection and backing down.
Because war is costly, R never makes a suboptimal offer to attempt to invite T to preemptively strike; that is, 1 − vT ≥ wR whenever the credibility condition is met, thus R always makes its optimal offer.
War may also occur when preferable to optimal settlements even when a bargaining range does exist, but is guaranteed when one does not. Thus, this gives a sufficient condition for war.
Increasing costs would narrow the range when bargaining breakdown was certain. When m = 1, the bargaining breakdown range thus represents the maximum war range. However, when m < 1, war can occur even when there is a bargaining range as attempted bargaining means ceding the mobilization advantage even for a future revisionist attack.
As before, see Sarkees and Wayman (2010) and Phillips and Axelrod (2005).
As previously noted, Wagner (2007) represents an important exception.
This mechanism explains the lower strip of the war range present in Figure 4.
Cf. Correlates of War (Sarkees and Wayman 2010) and NMC data v4.0 (Singer, 1987).
We should also note that Paul (1994) has previously suggested first-strike advantages as an explanation for war initiation by weak states.
Both cases involve multiple combatants. In each case, we will focus on the Israeli-Egyptian front, as this represents the primary axis of competition.
In addition to the sources cited in this section, see Badri, Magdoub, and Din Zohdy (1978), Rabinovich (2004), and Asher (2009).
See Blank (2017) for a similar argument. In addition, Shlapak and Johnson (2016) acknowledge the possibility of counterattacks, although they express skepticism about whether NATO could make the threat of a counterattack credible. Their doubts thus line up with our model.
References
Author notes
Authors’ note: We would like to thank Jaroslav Tir, David Bearce, Carew Boulding, Jin Yeub Kim, Cody Eldredge, and Tim Passmore, along with the anonymous reviewers and International Studies Quarterly’s editors for many helpful comments.
