Bb Rational Choice Theory Soviet General Secretary Mikhail Gorbachev and US President Ronald Reagan sign the Intermediate-range Nuclear Forces (INF) Treaty on 8 December 1987, in which the two Countries agreed to eliminate their entice arsenals of the weapons. Rational choice theory can help us understand decisions such as these. Source: © Reuters/CORBIShy do things happen in international relations? Wars start and \ X ] end, alliances are formed, states build weapons, and so on. We are interested in understanding why events like these happen. Ultimately, things happen in international relations—as in the political, eco- nomic, and social worlds more generally—as a result of decisions made by people. These people may be leaders of states, members of the legislature or military, members of nongovernmental organizations, or just simply citizens ofa country. Many people argue that to explain international conflict we need to focus on explaining decisions. But how do we explain them? To do so, many rely upon the assumption of rationality. Assuming rationality opens the door to various methodologies to explain decision making, particularly expected utility theory and game theory. This leads to an approach to studying inter- national conflict (and other areas of social science) known as rational choice theory, and perhaps more accurately as the strategic perspective (Bueno de Mesquita 2010), Rational choice has led to a number of advances in theoriz- ing about international conflict. ‘We begin by taking a closer look at what is meant by rationality, as well as discussing nonrational approaches to studying foreign policy decision making. We then explore expected utility theory and game theory, tools that become available through the assumption of rationality. Finally, we explore the bargaining model of war, an important theory that helps us understand international conflict throughout multiple stages of the process of war. Rationality ‘What is rationality? Should one assume rationality? Answers to these ques- tions have been subject to a great deal of debate. Decisions about whether an assumption of rationality is useful depend on what the assumption entails. Many different assumptions of rationality have been used. We will examine three of the most common: procedural rationality, bounded rationality, and instrumental rationality. Procedural rationality is similar to the common, everyday conception of rationality in which omniscient actors are said to make a “cool and clear- headed ends-means calculation” (Verba 1961, 95) in the course of considering all available options and choosing the best one.' This is called maximizing behavior because the procedurally rational actor chooses the alternative with the maximum gain. In addition, the best option is the alternative that ‘The label procedural rationality comes from Simon (1976). 4546 Foundations is objectively in the best interest of the decision maker. Therefore, different rational actors that are in the same situation will always make the same choice. Nonlogical influences—influences which the actor is unaware of, and if he were aware of it, he would discount it—are not allowed within proce- dural rationality. For instance, George and George (1956) argue that Wood- row Wilson’s decision making at several crucial points, including during his efforts to get the League of Nations treaty ratified by the United States Senate following World War I, was affected in large part by Wilson’s childhood relationship with his father. Other potential nonlogical influences include misperceptions, emotions, and psychological or cognitive limitations. To the extent that real actors are influenced by factors like these when making decisions, they are not procedurally rational. Herbert Simon (1957) developed a different conception of rationality, which is called bounded rationality. The idea of bounded rationality is that “human behavior is intendedly rational, but only boundedly so” (Simon 1997, 88). Bounded rationality involves what is called satisficing behav- ior, rather than maximizing behavior. Satisficing is a sequential process. Instead of examining all possible alternatives and choosing the very best | Box 3.1 | Concept in Focus: The Unitary Actor Assumption We often talk about states doing things: Germany invaded Poland in 1939, Iraq attacked Kuwait in 1990, and so on. This is known as the unitary actor assump- tion, where states are assumed to be single actors. The unitary actor assumption has a long history in studies of international relations, particularly within realism, a perspective focused on power and the structure of the international system. ‘Snyder, Bruck, and Sapin (1954, 1962) created what is called the decision mak- ing perspective and launched an important challenge on the unitary actor assump~ tion. They argued that foreign policy consists of decisions, made by identifiable decision makers; the making of decisions, therefore, is the behavioral activity that requires explanation. They also highlighted the importance of preferences and per- ceptions in explaining those decisions. Thus, in many ways, the ideas of Snyder, Bruck, and Sapin are fundamental to rational choice theory in international relations. The decision making approach was revolutionary because it challenged the ten- dency that realists have to reify the state (assuming that states have goals, fears, aspirations, etc., as if they were people). But states do not have character, and so on; states are legal entities. The individuals within each state are what matters. The state itself is not some autonomous actor apart from its people. Nonetheless, we might talk about one state invading another, but what we really mean is that the leader(s) of one state decided to invade another. Although Allison (1971) argues that the unitary actor assumption is a fundamental part of rational choice theory, the two are separate issues. Rational choice theory is ‘grounded in the decision making perspective and is rooted in the idea that people, not states, make decisions.Rational Choice Theory 47 one, options are looked at sequentially. The decision maker then evaluates each alternative to determine if it is acceptable. If so, that option is chosen; otherwise, the decision maker will keep looking until an alternative is found that is “good enough.” This is a very different notion of rationality that seems to be a much more realistic description of how people actually make decisions. Leaders of states are often affected by misperceptions when making decisions (Jervis 1976). Decision makers usually do not have enough time available to consider all possible courses of action (Allison 1971). Furthermore, people sometimes lack the cognitive ability to understand the ramifications of their decisions (George and Smoke 1974). Because of such constraints on decision mak- ing, Braybrooke and Lindblom (1963) argue that most decision making is “disjointed incrementalism,” where small decisions are made with little understanding. Psychological and Cognitive Approaches Given these limitations of procedural rationality, a number of scholars have sought to explain foreign policy decision making without assuming rationality. These psychological approaches instead employ social or cog- nitive psychology in their attempts to explain international politics. Most common are studies focusing on cognitive psychology to examine how indi- viduals’ beliefs, biases, and ability to process information lead to deviations from procedurally rational behavior. There are numerous examples of this approach. Lampton (1973) argues that the image that American leaders had of Chinese leadership played an important role in determining the outcome of crises between the United States and China. Jervis (1976) argues that perceptions and misperceptions have important impacts on international politics. Hermann and Kegley (1995) argue that leaders’ images and beliefs can explain peace between democracies. Other scholars have used elements of both psychology and rationality to explain foreign policy decision making. George (1969) introduces the idea of the operational code, which refers to a political leader’s beliefs about the nature of politics and political conflict, his views regarding the extent to which historical developments can be shaped, and his notions of cor- rect strategy and tactics. Walker, Schafer, and Young (1998) develop sys- tematic procedures for identifying leaders’ operational codes. Additionally, operational codes can be used to identify leaders’ preferences and then game theory used to explain the resulting decisions. Prospect theory is a similar framework to expected utility theory (which we will examine in more detail shortly) for analyzing decision making with risk, but there are important differences. Prospect theory highlights the importance of the decision-maker’s frame of reference when confronting a situation, People tend to be averse to risk when they anticipate gains but tend to be risk acceptant when they anticipate losses (Levy 1997). Although it is sometimes characterized as being contrary to rational choice, Carlson and48 Foundations Dacey (2006) show that prospect theory can be incorporated in a game- theoretic model. Finally, poliheuristic theory is an approach that specifically integrates cognitive and rational choice together. Poliheuristic theory views decision making as a two-stage process. At the first stage, decision makers use non- rational means to reduce the set of possible options to a manageable level, and at the second stage, they make a rational choice among the remaining options (Mintz. 2004). Instrumental Rationality Discussions of bounded rationality as well as psychological and cogni- tive approaches seem to demonstrate that the assumption of rationality in rational choice theory is very problematic. These scholars have critiqued procedural rationality, however, but that is not the rationality of rational choice theory. What then do we mean by rationality? Rational choice theory is rooted in the assumption of instrumental ratio- nality.* An instrumentally rational actor is one who, when confronted with “two alternatives which give rise to outcomes . .. will choose the one which yields the more preferred outcome” (Luce and Raiffa 1957, 50). Note that the definition does not say which alternative is better; the decision maker decides. This definition avoids many problems by not trying to dictate what decision makers should prefer. Instrumentally rational players, then, are those who always make choices they believe are consistent with their inter- ests and objectives as they define them. In other words, instrumentally ratio- nal players are purposeful players. Additionally, a rational actor must have complete and transitive prefer- ences. For preferences to be complete, for any two alternatives x and y, either xPy (x is preferred to y), yPx (y is preferred to x), or xIy (the person is indifferent between x and y). Transitivity involves preferences among at least three different outcomes. Considering the most basic case with three alternatives x, y, and z, transitivity means that if xPy and yPz, then xPz. Thus, if I prefer chocolate ice cream to vanilla, and vanilla ice cream to strawberry, then it only makes sense that I prefer chocolate ice cream to strawberry. Instrumental rationality makes no normative judgments about prefer- ences. Whether one’s preferences are “good” or “evil,” “instrumental” or “expressive,” or anything else has no impact on one’s instrumental rational- ity. Thus, a person such as Adolf Hitler who prefers to launch an attempt to conquer entire continents and wipe out entire races, and acts accordingly, is just as rational as someone like Woodrow Wilson who seeks to promote democracy and acts accordingly. Similarly, a job seeker who prefers to maximize leisure, with no regard for income, and acts accordingly, is just ¢ detailed discussions of instrumental rationality, see Zagare (1990), Quackenbush (2004), and Hindmoor (2006).Rational Choice Theory 49 as rational as another who acts according to her preference for maximizing income. Saying that someone is instrumentally rational is not paying them a compliment; it is simply saying that they act according to their prefer- ences, whatever they may be. Even so, many have argued that a variety of psychological, informational, or structural factors interfere with actors’— particularly states’—ability to act rationally. These factors, however, only interfere with procedural rationality, not instrumental rationality. Instru- mental rationality is compatible with a wide variety of supposedly limiting factors (Zagare 1990). If decisions are central to international relations, it makes sense that decision makers make those decisions with a purpose. Such purposive deci- sion making is what instrumental rationality is about. In addition, assum- ing instrumental rationality allows us to use expected utility theory and game theory, useful tools for explaining decision making. We explore each in turn, Expected Utility Theory Expected utility theory is a model of rational decision making. We will examine the basics of expected utility theory and then turn our attention to a specific application, the expected utility theory of international con- flict developed by Bueno de Mesquita (1981).) Modeling decisions using expected utility requires that we know several pieces of information about the situation we want to examine. First, we must know the possible choices that could be made. These are the various courses of action that an individual can choose from: starting a war or not, putting more money into the defense budget or into the edu- cation budget, and so on, Situations can have many alternatives. Part of setting up an expected utility model is deciding what options we should include, which means judging what the most important or salient alterna- tives were to the leaders when they made their decision. Second, we must know the possible outcomes of the situation. These are the various things that could occur in the world: various states of being or ways in which things could turn out. For example, if a state decides to start a war, it may win or it may lose. There can of course be many more out- comes than two. When we develop models we choose how many outcomes and how much detail about them we want to deal with. Third, we need to know the players’ utility or preferences for the differ- ent outcomes. Utility is the value that people attach to different outcomes. People want, or prefer, outcomes with higher utility. Essentially, utility is net benefits, or the benefits of a particular outcome minus the costs of achieving that outcome. There are different ways that utility can be measured. *For a more detailed introduction to expected utility theory, see Morrow (1994) or Resnik (1987).50 Foundations The most basic way to measure utility is at the ordinal level. Sometimes we just know that one outcome is preferred to another. For example, I like chocolate ice cream more than vanilla; I prefer winning a war to losing a war. At the ordinal level, we can simply determine the order of the actors’ preferences. A more detailed way to measure utility is at the cardinal level. Some- times we can judge exactly how much outcomes are valued on a continuous scale. This will allow us to tell how much more one outcome is worth than another. For example, we may know that winning a war is worth 1,000, while losing a war is worth -500 and remaining at peace is worth 250. By measuring utility on a cardinal scale, decision makers can make rather com- plicated comparisons and judgments within their own cardinal scale. Finally, we need to know the probability, or likelihood, that the different outcomes will occur in some situation, Often when we make a choice, we do not know for sure what the outcome will be, but we usually have some idea, that is, a rough estimate of the probability. When a very strong country starts a war, for instance, there may be 90 percent chance that it will win. Other times, if countries are closely matched, there may be 50 percent chance of winning, We sometimes discuss probability as a percentage, such as 75 percent, or as a decimal, such as 0.75. These mean the same thing. Probabilities of different outcomes that might follow some action must add up to 1 (or 100 percent). Often, we see the options of “win” and “lose.” With only two outcomes, we only need to know one probability, since we can calculate the other: p(lose) = 1 - p(win). Using expected utility suggests that people actually think not only about what they want—their preferences or the utility they get—but also the odds that they will get it, Expected utility is calculated simply be multiplying the utility of outcomes by the probability that they will actually happen and summing the results over the set of possible outcomes. So, if there are two possible outcomes to a choice, then we could write EU choice = PiU + P2U2 where EU;joce is the expected utility of a given choice, p; is the probability of outcome 1, U; is the utility of outcome 1, p> is the probability of outcome 2, and U; is the utility of outcome 2. We have a utility of an outcome, but an expected utility of a choice. We assume that any individual outcome, such as winning a war, has a fixed utility without any probability involved. Once we move to the possibility that multiple outcomes might follow from a particular choice, and hence to a situation where probability comes into play, then it becomes expected utility. ‘A simple example of expected utility can be from a state considering a military attack on another state, In this example, Upeace = 100, Uwin = 500, Une =-1000, and Pyjq = 0.5. Therefore, we can calculate the expected utility of attacking asRational Choice Theory 51 EU grrack = PwinUwin + PloseUtose EU yuack = 0.5(500) + 0.5(-1000) EU ypack = 250 - 500 EU pack = -250 The potential attacker has two options: to attack or not to attack. A rational state will choose the preferred alternative. In this example, EUseack = -250, whereas EUp acack = Upeace = 100. Thus, not attacking is much better for the state than attacking, so we would predict that the state would not attack. If the state’s likelihood of winning was greater, say Pyis = 0.75, how- ever, and everything else remained the same, then EUyrack = 125. This new expected utility for attacking is now larger than the expected utility of not attacking, so we would predict that the state would attack. This example is useful for illustrating the logic of expected utility. A sub- stantial limitation, however, is the question of where the numbers come from. Saying that peace is worth 100 while winning is worth 500 is nice, but what do these numbers mean? It is possible to develop ways to measure the utilities and probabilities in a meaningful way. In an early development of rational choice theory, Bueno de Mesquita (1981) did exactly that. In his book The War Trap, Bueno de Mesquita developed an expected utility theory of international conflict. Rationality implies that leaders would only start wars that they expect to gain from in some way; if they expect to suffer a net loss, then they will avoid war. Accordingly, Bueno de Mesquita developed a theory to determine whether states would expect a gain or loss against certain opponents at different points in time. If a state attacks a given opponent, they are likely to wind up fighting a war involving only two states: the initial attacker and the initial target. This is a bilateral war and is essentially the situation assumed in our example above. Some wars, however, do expand to include additional countries, so Bueno de Mesquita accounts for this possibility as well. In calculating the expected utility of a multilateral war, Bueno de Mesquita seeks to answer the following questions: Who is likely to join the war? On which side will they join? And what effect will their joining have on the outcome of the war? The resulting expected utility for a conflict in Bueno de Mesquita’s (1981) model depends on these individual expected utilities for a bilateral war and a multilateral war: EU(conflict) = EU(bilateral war) + EU(multilateral war). To this point, we have focused on the logic of Bueno de Mesquita’s theory, butitis also important to look at the evidence regarding it. Bueno de Mesquita does this by examining regional dyads (pairs of states within the same region of the world) between 1816 and 1974. For each conflict observed, he mea- sures each state’s expected utility for attacking the opponent. In the previous52 Foundations chapter, we saw how conflict can be observed. The more difficult obstacle is measuring expected utility. Bueno de Mesquita does so by developing tech- niques to measure the probability of victory and utility. To measure probability of victory, Bueno de Mesquita focused on the relative power between states. There is a common measure of power used in studies of international politics called the Composite Indicator of National Capabilities (or CINC score), which we will discuss more fully in chapter 5. Bueno de Mesquita assumes that a state’s probability of victory is essentially equivalent to its relative power in comparison with its potential opponent. Thus, ifa state has 60 percent of the power within a dyad, then its probabil- ity of victory is 0.60, whereas if the state has only 35 percent of the power in the dyad, then its probability of winning is 0.35. To measure utility, Bueno de Mesquita used states’ alliance portfolios. A state's alliance portfolio refers to its entire configuration of alliance commit- ments to other states, We can look at not only whether a state has an alliance with another country, but also what kind of alliance its.’ Then we compare the alliances of state A (the potential initiator) and state B (the potential target). It is assumed that the more similar two states’ alliance commitments are, the greater is their foreign policy similarity. For example, if state A has alliances with ten other countries, and state B also has alliances with ten other countries, and they are the same ten countries, then state A and B seem to have a lot in common, On the other hand, if B is allied to ten different countries than A is allied to (with no overlap), this would indicate that A and B have much less in common. Of course the possibilities are endless, but the basic idea remains the same. We can visualize the results by looking at a series of tables that compare conflict initiation with expected utility. Results for interstate wars are shown in Table 3.1. Bueno de Mesquita’s theory indicates that states will only initiate conflict if their expected utility for conflict is greater than zero. There were 76 cases of interstate war initiation; of these, the initiator’s expected utility was greater than or equal to zero 65 times (85.5 percent) and negative only 11 times. For the target, the trend is exactly opposite: Targets had a nonnegative expected utility only 11 times and a negative one 65 times. This is a statistically significant relationship, indicating support for Bueno de Mesquita’s theory. ee Interstate War Initiation and Expected Utility EXPECTED UTILITY SCORE INITIATOR OPPONENT Greater than or equal to zero 6 " Less than zero "1 65 Source: Bruce Bueno de Mesquita, The War Trap (New Haven, CT: Yale University Press, 1981), 129. ine alliances in more detail in chapter 6.Rational Choice Theory 53 Bueno de Mesquita (1981) also examines lesser levels of conflict. Results of comparing expected utility with uses of force are shown in Table 3.2. Out of 102 disputes involving the use of force, the initiator had a nonnegative expected utility 78 times (76.5 percent) and a negative expected utility 24 times. On the other hand, targets had an expected utility greater than or equal to zero only 14 times, and a negative one 88 times. Once again, this relationship is unlikely to arise from chance. Bueno de Mesquita examines one final level of conflict, interstate threats, results for which are shown in Table 3.3. Out of 73 interstate threats, the initiator’s expected utility was greater than or equal to zero 50 times (68.5 percent), and less than zero 23 times. Targets had a nonnegative expected utility 13 times and a negative one 60 times. As with the previous analyses, this relationship is again unlikely to arise from chance. One might also note that as we went down in the level of hostility from war, to use of force, and finally to interstate threat, expected utility had a slightly weaker effect on decisions to initiate. Bueno de Mesquita argues that this is to be expected because states are more willing to bluff (threatening even when they prefer not to fight) at lower hostility levels. Bueno de Mesquita also conducts these analyses looking at different peri- ods of time as well as different regions of the world and the results are essen- tially the same. Furthermore, he examines the relationship between expected utility, relative power, and conflict. While the relationship between power and conflict is quite similar to the relationship between expected utility and conflict, his expected utility theory greatly outperforms a power explana- tion in cases where the predictions between the two differ. eee Uses of Militarized Force and Expected Utility EXPECTED UTILITY SCORE INITIATOR OPPONENT Greater than or equal to zero 78 14 Less than zero 24 88 Source: Bruce Bueno de Mesquita, The War Trap (New Haven, CT: Yale University Press, 1981), 130. Sees Interstate Threats and Expected Utility EXPECTED UTILITY SCORE INITIATOR OPPONENT Greater than or equal to zero 50 13 Less than zero 23 60 Source: Bruce Bueno de Mesquita, The War Trap (New Haven, CT: Yale University Press, 1981), 130.54 Foundations In summary, Bueno de Mesquita’s expected utility theory of international conflict is well supported by the historical record, even if its predictions are not perfect. This indicates empirical support for the basic logic of expected utility theory and rational choice. It also indicates that measuring the utili- ties and probabilities necessary for calculating expected utility is possible, even if the measures developed are not perfect. Expected utility theory, however, has one crucial limitation: It does not account for strategic interaction between decision makers. Rather, expected utility theory focuses on each actor’s choice in a vacuum. Of course in the teal world, leaders of states must take into account a variety of other poten- tial decision makers: other states, members of their own government, inter- national organizations, the domestic population, and many others. While most theories do not account for all of these different decision makers, it is important to be able to understand this strategic interaction between mul- tiple actors. To do so, we turn our attention to game theory. Game Theory Game theory is the analysis of how decision makers interact in decision making to take into account reactions and choices of the other decision makers. Although its title uses the word theory, game theory not a unified empirical theory. Rather, itis a methodology for examining strategic behav- ior among interacting and interdependent unit: We call it game theory because we use it to examine games of strategy, as opposed to games of skill (such as a 100-meter dash) or games of chance (such as Roulette). A strategic game consists of a set of players, a set of actions available to each player, and the preferences for each player over the outcomes that may arise. We call them games because we think of actors making decisions as anticipating how the other will respond, much as we do in a game of chess or checkers. In game theory, the interacting units are called players. Who or what con- stitutes a player, however, is a determination left to the individual analyst. Normally, the identification of players depends on the analyst’s purpose and the specific research questions addressed. One could, for example, conceive of the 2011 conflict in Libya as a two-person game between Libya and the National Transitional Council, as a two-person game between the NATO alliance and Libya, or as a three-person game among Libya, NATO, and the National Transitional Council. An even more detailed analysis that rec- ognized important policy differences that separated the United States from France and Germany, however, might specify five (or more) players. In other words, both the number and identification of players are extra-game-theoretic decisions that, in principle, should be driven by theoretical considerations, empirical factors, or both rather than by methodological considerations. SNumerous textbooks provide an overview of game theory. Game theory texts that focus on applications in political science include Gates and Humes (1997), McCarty and Meirowitz (2007), and Morrow (1994).Rational Choice Theory 55 A realist who was also a game theorist would likely limit the specification of players to states. But players need not be so restricted. For example, one could employ game theory from within Allison’s (1971) organizational pro- cess model to focus on, say, the decision-making process within the United States government. In this game, the players—such as the White House, the Departments of State and Defense, and other important bureaucratic departments and agencies—would be different organizations within the government, rather than entire governments. But the analysis, nonetheless, could still be game theoretic. Regardless of how they are specified, the players are assumed to make choices that, along with the choices of the other players, lead to spe- cific outcomes. The specification of outcomes, like that of the players, is another important judgment call that an analyst must make. Thus, a general analysis of conflict might simply specify “negotiation” as a pos- sible outcome. By contrast, a more fine-grained study might distinguish between a negotiated outcome that favored one side and one that favored the other. In game theory, the players are assumed to evaluate (subjectively) the utility (or worth) of each possible outcome, and to make choices that are instrumentally rational. This is not to say, though, that the players evalu- ate the outcomes in the same way. In fact, many conflicts are traceable to disputes about the value of specific outcomes. For example, prior to the 1999 Kosovo War, the Kosovo Liberation Army (KLA) believed that by pressing its case against Serbia, an independent Kosovo would result. Serbian leader Slobodan Milosevic disagreed. Milosevic thought he could use force to stamp out the separatist movement in Kosovo and, in the pro- cess, consolidate his internal political standing. To minimally specify a game, the players, the choices available to them, the consequences of their choices (i.e., the outcomes), and the players’ utility for each outcome must be specified. Morrow (1994, 57) correctly observes that “the design of the game .. . [is] .. . the single most important decision in modeling.” Note, however, that these are decisions that must be made before the analytic tools of game theory can be used. This means that the explanatory power of any game-theoretic analysis depends less on the meth- odology itself than on the theoretical sensitivity of the analyst. Nonetheless, game theory provides a potentially useful methodological framework for analyzing international conflict. Games can be shown in two basic forms. First, there is the strategic form, also known as the normal form. If a game is reduced from the actual display of what paths actors may follow to a matrix of choices of outcomes, it is said to be in strategic form. Second is the extensive form, where the game is represented as a tree of choices. Generally, the strategic form is useful for illustrating simultaneous games, whereas the extensive form is useful for illustrating sequential games. Any game, however, can be shown in either form. We will examine the logic of game theory, including the basics of how to solve games, before examining applications of game theory to explaining international conflict.56 Foundations Strategic Form Games We begin our examination of strategic form games by looking at Prisoner’s Dilemma, the most famous of all games in game theory. The basic story of this game is that the police arrest two people suspected of commit- ting a serious crime. The district attorney, however, does not have enough evidence to convict them without a confession. So the police place the sus- pects into separate cells and make the following offer to each individually: “If you confess and your partner does not, then you'll be the state’s witness and get off scot-free and we'll throw the book at your partner. But if your partner squeals and you don’t, then I'll throw the book at you. If you both refuse to confess, then I'll till be able to convict you of more minor charges, s0 you'll still go to jail. Ifyou both confess, then you'll both be convicted of the more serious charges, but I'll make sure that you get a lighter sentence than if you keep your mouth shut.” Each of the suspects, A and B, has two options. They can either cooperate with the other player by keeping quiet and telling the police nothing, or they can defect from cooperation with their partner by confessing to the police. Since there are two players each with two strategies, four outcomes are pos- sible, Each player’s best outcome is to defect while the other cooperates—the temptation outcome. Each player’s worst outcome is to cooperate while the other defects—the sucker outcome, For the two outcomes in between, mutual cooperation would be best for both suspects since they would only be con- vieted of a more minor crime. The resulting game is shown in Figure 3.1. Within each cell are two numbers: The number before the comma identifies the payoff for the row player (A in this case), while the number after the comma identifies the payoff for the column player (B). ‘What will the players do? We determine that by solving the game, or more formally, by finding the equilibrium or equilibria of the game. There are different types of equilibria (or equilibrium concepts) that we can use, each of which is appropriate under different circumstances. The most fun- damental equilibrium concept is the Nash equilibrium, named after John Nash, an important pioneer of game theory and subject of the 2001 movie A Beautiful Mind, A Nash equilibrium is a set of strategies where each player plays her best response to the other’s strategy; neither player has an incen- tive to unilaterally switch strategies. Figure 3.1 | Prisoners’ Dilemma Cooperate (keep quiet) __Defect (confess) Cooperate (keep quiet) (3,3) (1,4) Defect (confess) (4,1) (2,2)°Rational Choice Theory 57 ‘We can identify this in a straightforward manner. Starting with suspect ‘A, we can see that if B chooses to cooperate, A will get 3 by also cooperating but 4 by defecting. Thus, choosing to defect is his best response to coopera- tion. If B chooses to defect, A will get 1 by cooperating but 2 by defecting. Once again, choosing to defect is A’s best response. Thus, A has what is known as a dominant strategy. A dominant strategy is a strategy that is always a player’s best response to whatever strategy the opponent might do. In contrast, a dominated strategy is a strategy that is never a best response to the opponent’s strategy. Since Prisoner’s Dilemma is a symmetric game, it is easy to see that B also has a dominant strategy to defect. Therefore, there is only one Nash equilibrium: mutual defection, which we could also write as (defect, defect). Note that we should always write an equilibrium in terms of strategies, not in terms of utilities, that is, not as (2, 2), because there can be more than one outcome with the same payoffs to the players. So why is this game called a dilemma? It is a very straightforward game to solve, where each player's rational choice is obvious because each has a dominant strategy. If you look more closely at the game, you can see that if the players would both cooperate, then they would each get their second- best outcome, 3, but by both defecting, then they both get their second- worst outcome, 2. They would both be better off than they are by defecting; being rational leads them to a worse outcome. Prisoner’s Dilemma has been applied to a variety of topics in interna- tional relations, including arms races, deterrence, and international trade. One of the common questions is how can we get cooperation between the players in a Prisoner’s Dilemma situation? The short answer is that, if the players only play the game once, then there is no way to get them to cooper- ate short of changing their payoffs and thus the game. If, however, the game is repeated again and again, then cooperation emerges as a rational possibil- ity (Axelrod 1984). Exploring cooperation between states in this way has enabled scholars to better understand the impact of international institu- tions, such as the United Nations, on international cooperation and conflict (Keohane 1984; Oye 1986). ‘We will examine one more strategic form game before moving on. Chicken is another classic game that helps to further illustrate game theory. The basic story of Chicken is two people (teenage boys, most likely) drive cars straight at each other until one of them swerves. If one swerves while the other drives straight, then he is the chicken (loses), while the one who drives straight wins, Thus, the best outcome is to drive straight (defect) while the other swerves (cooperates). The second best outcome for each player is if they both swerve; although neither wins, they aren’t the chicken either. Being the chicken— swerving while the other drives straight—is the second-worst outcome. The worst outcome for both players is if neither swerves; although neither is a chicken, they both likely die in a head-on collision. Apparently, it is better to bea live chicken than a dead duck. The resulting game is shown in Figure 3.2,58 Foundations Figure 3.2 Meir} B Cooperate (swerve) _Defect (drive straight) Cooperate (swerve) (3,3) (2,4)" Defect (drive straight) (4,2) (4, 1) Considering A’s choices first, we can see that if B cooperates then A’s best response is to defect (getting 4 rather than 3). On the other hand, if B defects, A gets only 1 by also defecting but gets 2 by cooperating; A’s best response is to cooperate. Thus, we can see that A’s best response is to. choose the opposite of what B does. Similarly, B wants to choose the oppo- site of what A chooses. Thus, there are two pure-strategy Nash equilibria in Chicken: (defect, cooperate) and (cooperate, defect). Chicken has been applied to models of deterrence, crisis bargaining, and other topics in inter- national conflict. This covers the basics of strategic form games and Nash equilibria. How- ever, there is much more to these games than we have examined here. First, not all games are 2 x 2 games like Prisoner's Dilemma and Chicken because the players might have more than two strategies or there might be more than two players, The logic of finding Nash equilibria, however, remains the same. Second, we have only been looking at pure strategies, when a player chooses a particular strategy (such as cooperate or defect) with certainty. Yet it is also possible for players to employ mixed strategies, where they randomly choose between their pure strategies according to specified probabilities (e.g., cooperate two-thirds of the time and defect one-third of the time). Extensive Form Games The second primary way to illustrate a game is in the extensive form. Extensive form games are also called dynamic games because they explic- itly model the sequence of decisions—an aspect of time—within the game. Extensive form games also make it easier to examine situations in which there are three or more players or there are multiple options available to the decision makers. For example, instead of simply having options to cooper- ate or defect, states may also have options to escalate or to select not only whether to make a demand, but how much of one to make. Although the strategic form is the classical way to represent games, the extensive form has become more common in recent years. To illustrate the extensive form, we will revisit Prisoner’s Dilemma. Recall that in the story above, the suspects were separated into separate rooms and presented with their options separately. Thus, each player had to decide whether to cooperate or defect without knowing the choice of the other. In effect, their moves were simultaneous, which nicely fits theRational Choice Theory 59 strategic form of Figure 3.1. But what if their moves were sequential, that is, what if suspect A had to decide first, and then suspect B could decide after observing what A had done? This new situation is illustrated in the extensive form game of Figure 3.3. Extensive form games are divided into decision nodes, where the player mak- ing the decision at that point must choose one of the alternatives, which are represented as branches extending from the node. At node 1, suspect A has to choose between C (cooperate) along the left branch or D (defect) along the right branch. If A cooperates, then it is B’s choice at node 2. B can also choose C along the left branch or D along the right branch. If A defects, then B chooses at node 3, where B can again choose between C and D. Regardless of B’s choice, a terminal node is reached where the game ends and payoffs are assigned. Payoffs (or utilities) are separated by a comma; the number before the comma is the payoff for the first mover (suspect A in this game), and the number after the comma is the payoff for the second mover (suspect B in this game). The payoffs are the same for the different outcomes as in Figure 3.1; mutual cooperation is (3, 3), mutual defection is (2, 2), and one player cooperating while the other defects is either (4, 1) or (1, 4) with the defecting player getting 4. We can solve extensive form games using a technique called backward induction. Backward induction involves determining the rational choice that each player will make at each node of the game, beginning at the end of game and working backward to the beginning. Beginning at node 3, we can see that B will choose D to get 2 rather than 1. At node 2, B once again will choose D, getting 4 rather than 3. Now we can move to A’s choice at node 1. IF A chooses D, she knows that B will also choose D, so A will get 2; if A chooses C, she knows that B will again choose D, so A will get only 1. Thus, A’s rational choice is to choose D. fee Sequential Prisoners’ Dilemma VA \ Ve N (3,3) (1,4) (4.1) (2,2)60 Foundations When we use backward induction to solve a game, we find the subgame perfect equilibrium. In the sequential Prisoner's Dilemma game, the sub- game perfect equilibrium is (defect; defect, defect); A’s strategy is defect, and B’s strategy is defect, defect (in game theory, a strategy is a complete plan of action for play of a game, so a player’s strategy must indicate her decision at every possible decision node in the game). So even though we have made the game sequential, the equilibrium outcome remains mutual defection. The subgame perfect equilibrium concept was developed by Reinhard Selten (1975) and is a refinement of Nash equilibrium that is more consis- tent with the assumption of instrumental rationality. Nash equilibrium is the basis of all solution concepts for noncooperative games because only outcomes that are associated with Nash equilibria are consistent with instrumentally rational choices by all players in a game. Some Nash equilib- ria, however, are based on instrumentally irrational choices. Consider the example game developed by John Harsanyi (1977) and shown in Figure 3.4. There are two players, A and B, each of whom makes one choice. Backward induction allows the course of rational play to easily be seen, At node 2, player B will choose D, securing 1 = 2 rather than the 0 received by choosing C. Knowing this, player A must make a choice at node 1 between C (u = 1) or D (1 = 2, since B will choose D). Thus, the rational choice is for each player to choose D, as indicated by the arrows in Figure 3.4. This is the subgame perfect equilibrium. Although the outcome DD is the only subgame perfect equilibrium, there are two Nash equilibria. (I indicate outcomes by using two letters, indicat- ing the strategies used by A and B to reach the outcome. Thus, outcome CD indicates that A cooperates and B defects, while outcome DD indicates that both defect.) The Nash equilibria, which are more easily identified in strategic form (Figure 3.5), are CC and DD. The equilibrium CC, however, eee Extensive-Form Representation of Harsanyi's Game A Node 1 (1.3) eI Node | (0,0) (2,2)Rational Choice Theory 61 Game Strategic-Form Representation of Harsany c D tat Cc (1, 3)* (1, 3) D (0, 0) (2, 2)" relies on player B’s irrational threat to choose C if A chooses D. This occurs because Nash equilibria do not exclude irrational threats off the equilibrium path. The equilibrium path is the path from the start of the game through each decision node until the equilibrium outcome is reached. For the DD equilibrium, the equilibrium path goes from the start of the game (node 1) through node 2 to the end, The equilibrium path for the CC equilibrium, however, only includes node 1 because A’s choice of C at node 1 ends the game; thus, B’s choice at node 2 is off the equilibrium path. Subgame perfect equilibria are those Nash equilibria that are based on instrumentally rational choices at every decision point. If there is a differ- ence between the Nash and subgame perfect equilibria of a particular game, we should always rely on the subgame perfect equilibrium as the correct solution. One simplification that we have assumed in our examination of game theory is that each player knows the opponent’s preferences. This is known as having complete information. We can also examine games of incomplete information, or asymmetric information, where at least one side has some uncertainty about the opponent, usually their preferences. With games of incomplete information, we need to use a different solution concept. A com- mon one is the Perfect Bayesian equilibrium, which ensures that actors make rational choices at each decision node and update their beliefs about oppo- nents using Bayes’s rule when possible. Using Game Theory Game theory can be used in several different ways. First, game theory could be used as a strictly normative tool to evaluate the efficacy of compet- ing policy prescriptions. For example, Zagare (1983) uses a game-theoretic framework to analyze the 1973 crisis in the Middle East, finding that the decision of the Nixon administration to place US strategic forces on alert was not only justifiable but was also more efficacious than the less provoca- tive approach of reassuring the Soviets of a US willingness to compromise. Game-theoretic models could also be employed descriptively to explain single cases that are intrinsically interesting or otherwise important. Bates et al. (1998) call this approach “analytic narratives,” where the theory and a case study are tightly integrated. Game theory is frequently utilized in this way. For example, Zagare’s (1979) analysis of the 1954 Geneva Conference helps to reconcile the62 Foundations well-known but unexplained discrepancy between the public and private policy pronouncements of US decision makers both before and during the negotia- tions that ended the Franco-Vietminh War. Zagare (1982) uses the same game- theoretic framework to eliminate competing, seemingly plausible, explanations of the Geneva Conference advanced by Thakur (1982). A game-theoretic analysis of a single historical case could also be thought as an inductive step taken to facilitate the development of a general theory. If they are to be useful, formal models, game-theoretic or otherwise, cannot be fashioned out of whole cloth. Game theory provides a useful framework for developing comparable case studies, which, in turn, could serve as a guide in the construction of more refined models or more powerful theories. Game theory is most commonly used within the study of international rela- tions to develop general theories. Such theories strive to model general political processes such as those associated with crisis bargaining, alliance formation, and war. There have been a large number of such applications in the study of international conflict. We now turn our attention to one important example. The International Interaction Game A prominent example of a game-theoretic model of international con- flict is the International Interaction Game (IG) developed by Bueno de Mesquita and Lalman in their 1992 book War and Reason. In this model (see Figure 3.6), two actors (State A and State B) choose whether to make a EX“ International Interaction Game 1 Cap, War, Nego Cap, War, Cap, War, Nego Cap, War, Source: Bruce Bueno de Mesquita and David Lalman, War and Reason: Domestic and Interna- tional Imperatives (New Haven, CT: Yale University Press, 1992), 30.Rational Choice Theory 63 a demand on the other, and given that demands have been made, whether or not to use force to pursue their aims.‘ If neither state makes a demand, the status quo results (SQ). If both states make demands and choose to use force, then war commences (War, or Wary, depending on who used force first). If a state backs down to a demand, it is said to acquiesce (Acq, or ‘Acqs), but if it backs down to a use of force, it is said to capitulate (Cap, or Caps). Finally, if both states make demands but neither resorts to the use of force, negotiation results (Nego). In the domestic constraints variant of the game, which receives the most empirical support, the decision to make a demand is the result of unspecified domestic processes. There are seven basic assumptions that form the heart of Bueno de Mesquita and Lalman’s theory. They assume that decision makers are instrumentally rational; that negotiations and war involve a lottery in which each player has a chance to win or lose with some probability; that achiev- ing one’s demands represents a gain from the status quo; and that each out come has a set of costs, benefits, or both associated with it. (For a complete discussion of the assumptions, see Bueno de Mesquita and Lalman 1992, 40-6; or Bueno de Mesquita 2010, 151.) These assumptions limit the pos- sible preference orderings as shown in Table 3.4. There are eight different outcomes that can occur from the game. The most preferred outcome takes a value of 8, while the least preferred outcome takes a value of 1. Preference Restrictions for the International Interaction Game OUTCOME __RESTRICTIONON ORDERING —_ POSSIBLE PREFERENCE RANK Buea} sq > Acq, Cap, 7to3 Acq, > all other outcomes 8 Acq, > Cap, 5to2 Nego > Acqi, Cap, War, Wan, 7to5 Cap, > War, War, 7to3 War, > War, 5to2 Cap, — 4to1 War, = 4to1 Note: All restrictions are in terms of Nation i's preferences. Source: Bruce Bueno de Mesquita and David Lalman, War and Reason: Domestic and Interna- tional Imperatives (New Haven, CT: Yale University Press, 1992), 47. A demand placed by State i is labeled D', whereas the choice to not make a demand is ~D', Similarly, the use of force is F', and the decision to not use force is ~F', where i denotes the state making the decision, Lowercase letters are used when a state has a second opportunity to make a demand or use force (for example, State A’s choice at node 4).64 Foundations These preference restrictions follow from Bueno de Mesquita and Lalman’s assumptions. Forcing the opponent to acquiesce is a low-cost way to achieve one’s demand, and is therefore preferred over all other outcomes; its value is always 8. Negotiation is assumed to be preferred to four other outcomes, so it can fall anywhere between the second-best outcome (prefer- ence rank 7) and the fourth-best outcome (preference rank 5). Given these preference restrictions, there are 2,704 different pairs of preferences for the two players. Thus, the International Interaction Game provides a general theory that accounts for 2,704 different strategic situations. Bueno de Mesquita and Lalman define several different player types that identify important characteristics of states in the International Interaction Game. A hawk is a state that would like to force his opponent to capitulate rather than negotiate (Cap, > Nego), while a dove is a state that prefers to negotiate with its opponent (Nego > Cap). A retaliator prefers to respond to an attack with a war rather than capitulate (War, > Cap,), while a surren- derer prefers to give in (capitulate) rather than fight (Cap, > War). Finally, a pacific dove is a state who is both a dove and a surrenderer. The conditions leading to war in the International Interaction Game are shown by the basic war theorem. The basic war theorem comes directly from examining the subgame perfect equilibria of the International Interac- tion Game. The theorem tells us that war is the sole equilibrium (in com- plete and perfect information) if Player A prefers to initiate a war than acquiesce (War, > Acqa), Player A prefers to capitulate than fight a war started by B (Cap, > Wary), Player B prefers to fight a war started by A than capitulate (Wary > Caps), and Player B prefers to force A to capitulate than negotiate (Cap, > Nego). So war can only occur with a hawk-retaliator against a surrenderer. The surrenderer’s weakness, rather than making war less likely, actually makes it more likely. The IG can help us understand the relationship between uncertainty and war. War can occur as an outcome of the IIG through several different paths; there are many different preference orderings that can lead states into war. War can happen even without uncertainty, as the basic war theorem demonstrates. Many scholars have argued that uncertainty makes war more likely while certainty makes war less likely. By contrast, the IIG identifies conditions under which increases in uncertainty make war more likely and conditions under which increases in uncertainty make war less likely. If at least one of the four conditions of the basic war theorem are not met, but decision makers perceive that all four are satisfied because each has incom- plete information about the preferences of the other, then uncertainty makes war more likely. If, however, the four conditions are met, then uncertaintyRational Choice Theory 65 about the preferences of an opponent can lead to choices that do not result in war even though war would result if there was complete information. Thus, uncertainty can either make war more likely or less likely, depending upon the circumstances. Bueno de Mesquita and Lalman also use the IG to examine other rela- tionships. Controversially, they argue that dissatisfaction with the status quo has no effect on the likelihood of conflict. They do so because war can occur in the IIG even when both sides value the status quo very highly and does not necessarily occur even when the status quo is poorly valued by the two sides. They argue that the IIG also sheds light on the impacts that democracy and power have on international conflict. ‘As with any other theory, it is important to test Bueno de Mesquita and Lalman’s theory with empirical evidence. Bueno de Mesquita and Lalman (1992) develop techniques similar to the techniques Bueno de Mesquita (1981) used to measure expected utility. This enables them to conduct a series of quantitative tests of their theory using data from Europe between 1816 and 1970. Bennett and Stam (2000a, 2000b, 2004) conduct a broader series of quantitative tests of the International Interaction Game across all regions of the world from 1816 to 1992. Through a series of empirical tests, the predictions made by the International Interaction Game are generally supported by the evidence. Bargaining Model of War The bargaining model of war refers to a variety of individual models that, despite having certain differences, share important common features, partic- ularly the inclusion of the size of demands as part of the game and the inclu- sion of war as part of the bargaining process (Reiter 2003). It is a rational choice framework that is a focus of much recent research on international conflict that has been effective in explaining the entire process of war. The bargaining model focuses squarely on an important puzzle regarding inter- national conflict. War is an extremely costly way for states to settle their disputes. And although countries have conflicts of interest all the time, only some disputes are resolved by force. So why go to war? Given the human and material costs of military conflict, why do states sometimes wage war rather than resolve their disputes through negotiations? Fearon (1995) used a bargaining model to examine rationalist explana- tions of the causes of war. As long as war is costly, some bargain should exist that both sides would prefer to actually fighting. If so, war can only arise as a result of some bargaining failure. To see this, consider the bargain- ing model shown in Figure 3.7. There are two states, A and B, engaged in bargaining over some issue, such as territory. The good is divided between the two states, with x indicating the share for State A (therefore, 1 - x is the share for State B). Although x is shown to the left of point p in the figure, x can take any value between 0 and 1. State A’s ideal point—an actor’s most66 Foundations etiewal The Bargaining Range Bargaining Range: Deals that both prefer to war —————— Deals that A prefers to war. ———> <¢—— Deals that B prefers to war 0 pea x Pp ptb u! State B's ideal point State A’s ideal point preferred division of the issue—is at the right edge of the figure, where x = 1, while state B’s ideal point is to the left where x = 0. The further we go away from someone’s ideal point, the less satisfied they are. We assume that State A would win a war between the two sides with probability p. Since war is costly, there are also cost terms: a is the expected cost that A would have to pay in war, while b is the expected costs for State B. For example, India and Pakistan disagree about Kashmir; although each controls only a portion of the territory, they claim the entirety as their own. India’s ideal point is for India to control all of Kashmir while Pakistan con- trols none of it (x = 1), whereas Pakistan’s ideal point is exactly the opposite (x =0). Although each side wants to have all of the territory, neither side actually does. Why not? Examining this question more generally gets to the heart of the bargain- ing model of war. We will assume that the winner of a war would be able to take all of the good in question. Since the probability that A wins is b, the expected value following a war is p. Since war is costly, however, each state's expected value for war is diminished. Accordingly, the expected value of war for State A is p - a, while the expected value of war for State B is p + b (notice that the cost terms move each state’s expected value of war away from their ideal point). Since State A’s expected value of war is p - a, A prefers any value of that is greater than p - a to war. In other words, A prefers any deal that is to the right of p - a over war. Similarly, B prefers any value of x that is less than (ie., to the left of) p + b to war. Therefore, both sides prefer any divi- sion of the good that is between p - a and p + b over war: This is the bar- gaining range, As long as war is costly, a bargaining range will exist where both sides prefer settlement to war. Thus, war can only result from some breakdown of bargaining, Bargaining failures can arise from three sources: incomplete information, commitment problems, and indivisible issues. We will explore each in turn.Rational Choice Theory 67 Fearon (1995) argues that the most important source of bargaining fail- ure is incomplete information with incentives to misrepresent. Incomplete information can come from several sources. States can be uncertain about each other’s willingness to fight because they disagree about the probability of victory (p) or because they disagree about the costs of fighting (a or b). One might wonder why the states cannot just tell each other about their capabilities and resolve, thus resolving the information problem. The basic answer is that states possess incentives to misrepresent or bluff. Private information about the opponent's capabilities, morale, outside help, and quality of plans can all lead to different beliefs about the prob- ability of winning. But if states revealed their private information, then it might undermine their advantage. For example, prior to the Six-Day War in 1967, Egypt and the other Arab states were confident in their ability to beat Israel. Israel was also confident in victory. Although they were greatly outnumbered in personnel, tanks, and aircraft, Israel felt that it had other advantages that numbers could not account for. In particular, Israel had a plan, called Focus (Moked), for achieving air supremacy right at the begin- ning of the war by destroying the air forces of Egypt, Syria, and Lebanon. The plan was initiated on the morning of 5 June, the first day of the war. Nearly all of Israel’s 196 combat aircraft were committed to the airstrike, with only 12 being held back to patrol Israeli airspace. In a quick succes- sion of attacking waves, Israeli aircraft struck fourteen Egyptian, one Iraqi, two Jordanian, and five Syrian air bases. The attack was a massive success for Israel, destroying 452 Arab aircraft for the loss of 19 Israeli aircraft (Hammel 1992; Oren 2002). Israel could not reveal this plan or other rea- sons why it was confident in victory because doing so would undermine its advantage. Even if states do not have secret plans, strategies, or capabilities that give them reason to be confident of victory, they have incentives to pretend that they do. Thus, there are strong incentives for states to misrepresent their private information. Fearon (1995) argues that incentives to misrepresent were impor- tant in leading to the Russo-Japanese War in 1904 and World War I in 1914. Commitment problems are also an important source of bargaining fail- ures. Commitment problems arise when states cannot credibly promise to cooperate with agreements they reach. Thus, at least one side has an incen- tive to renege on its promise once the other side fulfills its part of the deal. Promises are credible only when it is in the self-interest of actors to carry them out. Commitment problems are likely when bargaining over issues that affect future power, when power is changing exogenously, and when there are first strike advantages. When one side is rapidly growing in strength, its bargaining position is being improved continuously as time goes on. This creates a commitment problem because agreements that they agree to now will not be as good as they should be able to get in the future. This in turn creates powerful preventive motivations for opponents of such states to68 Foundations prevent such dramatic shifts. For example, Fearon argues that commitment problems were an important cause of World War I as “German leaders were willing to run serious risks of global conflict in 1914 . . . [because] they feared the consequences of further growth of Russian military power, which appeared to them to be on a dangerous upward trajectory” (1995, 407). The final source of bargaining failures identified by Fearon (1995) is indi- visible issues. Indivisible issues exist when circumstance dictates that the winner gets everything and the loser gets nothing. The bargaining model assumes that all issues are easily divided into increments. If issues are indi- Visible, there may not be possible divisions that lie within the bargaining space. For instance, Jerusalem is claimed as a capital and holy city by several groups and faiths, How can it be divided? Nonetheless, Fearon (1995) dis- counts indivisibilities as an important source of bargaining failure because virtually all issues are divisible in practice, especially if side payments are used to compensate the party that does not have direct control of the good. A variety of bargaining models have further explained the outbreak of international conflict, looking at the influence of factors such as military mobilizations (Slantchev 2005), public commitment (Tarar and Leventoglu 2009), and international organizations (Chapman and Wolford 2010). Bar- gaining models have also been developed to explain war outcomes and ter~ mination, The basic logic is that war terminates when the information gap between the two sides closes enough to create a bargaining range (Filson and Werner 2002, 2004; Powell 2004; Slantchev 2003; Smith 1998; Smith and Stam 2004; Wagner 2000). Slantchev (2003) labels this the principle of convergence. We will explore how the bargaining model enables us to explain war termination in more detail in chapter 12. One of the greatest strengths of the bargaining model of war is that it allows explanations of war onset, duration, outcome, and termination all in one integrated theo- retical model (Filson and Werner 2002). Conclusion The fundamental reason why wars and other forms of international con- flict occur is that leaders decide to fight. Furthermore, decisions by leaders of states and other key people drive the ending of wars, the formation of alliances, and many other phenomena of interest related to international conflict and security. Rational choice theory focuses on explaining these decisions by assuming that decision makers are rational. Although there are many different notions of what it means to be rational, rational choice theory is based on the assumption of instrumental rationality. Instrumen- tal rationality simply means that people make decisions according to their preferences. ‘Assuming instrumental rationality opens the way for formal, mathemati- cal modeling, but it does not require it. Expected utility theory and game theory are formal modeling tools of rational choice. The specific examplesRational Choice Theory 69 of theories developed using expected utility theory (Bueno de Mesquita 1981) and game theory (Bueno de Mesquita and Lalman 1992) that we examined in this chapter are two of many theories of international con- flict that scholars have developed using these tools—especially game theory. Game theory is especially useful because it focuses on explaining strategic interaction, which is an integral part of international politics. In particular, the bargaining model of war is an important game-theoretic framework that has become increasingly common for explaining international conflict. Rational choice theory is useful in establishing the logic of theories by ensuring logical consistency. That is certainly one of the primary strengths of formal models such as game theory. One concern about game-theoretic models is that some have not been subjected to empirical testing. Regardless of the rigor of our logic, we still need to examine evidence. We now turn our attention to a variety of specific topics related to international conflict, beginning with contiguity and territory. Key Concepts Bargaining model of war Bounded rationality Dominant strategy Expected utility theory Game theory Ideal point Instrumental rationality Nash equilibrium Operational code Poliheuristic theory Procedural rationality Prospect theory Psychological approaches Subgame perfect equilibrium Unitary actor assumption