David versus Goliath: Risk and Weaker State Confrontation

David versus Goliath: Risk and Weaker State Confrontation Confronting a more powerful rival can be a risky proposition. This paper integrates prospect theory into the growing Neoclassical Realist (NCR) literature to identify the conditions under which decision makers are most likely to accept foreign policy risks. I argue that decision makers governing regimes with low levels of political counterframing are more likely to settle into a dominant loss frame when their external security environment erodes. This increases the probability that they will initiate disputes with more powerful adversaries. To assess this proposition, I conduct a test of weaker state confrontation grounded in a NCR framework that utilizes the insights from prospect theory. Data come from the postwar era and support my hypothesis. Loss frames exert influence over the decision to initiate risky foreign policy strategies in regimes in which political counterframing is absent. “I’ll strike you down and cut off your head” ∼ David, confronting Goliath Introduction Why do weaker states confront opponents that are more powerful? Picking a fight with a superior adversary is a risky proposition. While the spoils in victory are often greater than those from prevailing over a lesser rival, greater too are the potential losses and likelihood of defeat. In the fable of David versus Goliath, David’s willingness to step forward and challenge his physically superior, better armed, and more experienced rival derived from a conviction that he possessed an unseen advantage unrelated to obvious indicators of their relative strength. In proclaiming, “The Lord will deliver you into my hands,” David indicated his belief that the balance of power actually favored his cause. Likewise, our understanding of weaker state confrontation concludes that smaller states initiate disputes when they enjoy non-obvious advantages over stronger rivals—for example, greater resolve, short-term tactical advantages, or anticipation of favorable third-party intervention. In such instances, the decision to challenge a more powerful adversary is a rational strategy and therefore theoretically unproblematic. By contrast, this paper takes at face value the proposition that asymmetric confrontation is a risky proposition and investigates the conditions under which decision makers become more acceptant of risk in the conduct of foreign policy. The argument offered here integrates prospect theory into a Neoclassical Realist (NCR) framework by combining established findings from cognitive science with basic categories of regime type. I assert that decision maker security frames, defined by changes in a state’s external security environment, influence the willingness to accept risks. Consistent with prospect theory, when a state’s security position erodes, leaders become more risk acceptant increasing the likelihood of initiating disputes with rivals that are more powerful. I also argue that this process is conditioned by the nature of political competition within the regime. Decision makers are more likely to fall prey to framing effects in states with low levels of political counterframing and are therefore more likely to adopt risky foreign policy strategies when their state experiences erosion in security. This argument is consistent with the emphasis that NCR places on the importance of domestic factors in foreign policy. It is also in line with the call to explicitly integrate authentic microfoundational assumptions into our theories of foreign policy behavior (Hudson 2005). As I demonstrate below, the resulting framework is also amenable to large-n testing. This opens new avenues of inquiry for both NCR and prospect theory that, to date, have been largely confined to case study analysis. The discussion below is organized as follows. I first briefly discuss the logic of weaker state confrontation and argue that it is a risky strategy even when it is carefully conceived and deliberate. Second, I summarize the core commitments of an NCR framework and use these to build a new theory of weaker state conflict. This section focuses on the relationship between domestic political competition and the onset of loss frames and takes the rest of NCR largely as it currently exists in the literature. Third, I conduct a statistical test of the propositions developed here. Results offer strong support for my theory. In the conclusion, I discuss the broader implications of these results for our understanding of risky foreign policy behavior and for NCR analysis more generally. Among the most promising conclusions is that NCR provides a much-needed framework for interdisciplinary scholars seeking to integrate findings from cognitive science into a foreign policy research program. Risk and Weaker State Aggression Is weaker state conflict initiation risky?1 Much of the scholarship in foreign policy does not address this question directly, despite the fact that confronting a more powerful rival initially seems like a poor strategy containing limited chances for success. For example, one explanation is that weaker states initiate risky policies because of misperception. The list of potential sources for errors in judgment are now well understood. Governments might operate under misperceptions about rival intentions (Jervis 1988), overestimate their relative capabilities (Fearon 1995), or fall prey to the human disposition favoring overconfidence (Johnson 2009). By this view, weaker state aggression is the inadvertent result of perceptual errors rather than an explicitly risky choice. A second category of argument depicts weaker initiators as deliberate and calculating, seizing advantage when they believe that their prospects for immediate success are greatest. The goal here is not outright victory but instead fait accompli.2 As Paul (1994) notes, weaker states sometimes confront more powerful rivals in order to capitalize upon short-term military and/or political advantages and then, operating from a newfound position of strength, demand a revision to the status quo before the full force of rival superiority can manifest. Success hinges on correctly identifying both the emergence of immediate tactical advantages and in accurately surmising that the target does not have the resolve to pursue the conflict to conclusion. In a similar vein, older arguments about alliance entrapment (Snyder 1984) and more recent work on third-party intervention (Chan 2010) suggest that allied support or international attempts to settle disputes can inadvertently generate moral hazards for weaker initiators, creating incentives to pursue fait accompli. However, even when the choice of confrontation is calculated, it is still reasonable to characterize weaker state initiation as a risky strategy. In the lead up to conflict, decision makers are often uncertain about rival resolve (Fearon 1995; Sechser 2010). Making judgments with imperfect information about resolve can be costly when confronting an inferior opponent, but it is potentially fatal when facing a superior rival willing to push a conflict to conclusion. In addition, research suggests that surprise attacks intended to capitalize on short-term advantages rarely yield wholly favorable outcomes and often backfire even when the balance of power favors the initiator (Helfstein 2012). Moreover, weak state strategy often has little to do with the outcome of a dispute. Even when they are successful, weaker states do not prevail because they outsmart their rivals. Instead, “Strong actors lose asymmetric conflict when they adopt the wrong strategy vis-à-vis their weaker adversaries” (Arrefuin-Toft 2005, 121). So, as Van Evera (2013) notes, while fait accompli may promise a “greater chance of political victory than quiet consultation,” it also increases the probability of open conflict because when “opponents stand firm, a collision is hard to avoid” (131–32). Although this is true for all governments, the risks in collision are greater for weaker initiators. Consider an iconic example of fate accompli, the Japanese Pearl Harbor attack against the United States. Feeling trapped by an eroding strategic position, Japanese policymakers hoped that the United States, confronting the specter of a costly European conflict, would acquiesce once Japan established military superiority in the region (Conroy and Wray 1990). Consistent with fait accompli, Japanese leaders correctly understood that there was an advantage in acting first, decisively, and with crippling force. However, while “none of the leaders were confident of Japan’s chances, especially in a lengthy war,” they nonetheless launched into the conflict with “a fatalistic attitude and only a vague hope that the United States would grow weary of the struggle after early Japanese successes and negotiate an agreement recognizing Japan’s territorial gains in East Asia” (Levi and Whyte 1997, 808). As the example suggests, the logic of fait accompli can effectively explain why a limited-aims strategy is sometimes more attractive than a total war strategy. However, the decision to confront powerful rivals in this way is still a risky endeavor, and, in the Japanese case at least, decision makers were explicitly aware of this fact. The central issue motivating this study thus remains. Why do states adopt risky strategies of confrontation? The cognitive version of NCR presented here attempts to answer this question by designing a framework to explicitly identify the combination of international and domestic conditions under which risk acceptant foreign policy is most likely to occur. A Cognitive Neoclassical Approach At its core, NCR is focused on explaining how states respond to security challenges. NCR scholars broadly accept realist assertions about the primacy of security politics and argue that shifts in the security environment initiate changes in foreign policy (Toje and Kunz 2012). As Rose (1998) puts it, security threats shape “the broad contours and general direction” of foreign policy behavior (147). Changes in the distribution of power are particularly salient (Kitchen 2010), though research points to a broad range of triggering mechanisms ranging from shifting alliance commitments (Cha 2000) to concerns over burden sharing in climate change negotiations (Purdon 2013). NCR departs from traditional realist analysis in two ways. First, NCR depicts state institutions as intervening variables linking external security imperatives to a country’s observed foreign policy. There is no perfect transmission belt that directly connects a state’s capabilities to its foreign policy needs (Kunz and Saltzman 2012). For example, policymakers are often constrained in decentralized states because power is diffused across institutions and domestic political interests are sharply divided. This can both prevent governments from expanding their global influence as national power grows (Zakaria 1999) and also renders the state unable to balance against emergent threats in a timely manner (Schweller 2008). When threatened, constrained governments are also least likely to respond with innovative foreign policies or emulate the successful strategies of others (Taliaferro 2006). Second, foreign policy decision makers are independent actors in NCR analysis.3 While elites can pressure decision makers to define security threats in terms of their own parochial interests, decision makers also act autonomously (Lobell 2009). For example, decision makers sometimes manipulate societal discourses about conflict in order to mobilize domestic opinion and coordinate political actors, thereby (partially) overcoming institutional constraints (Schweller 2009). Decision makers often attach meaning to security threats consistent with their preexisting beliefs, and then align foreign policy responses with internal constraints (Edelstein 2002). This can influence their willingness to confront rivals and contour the form that such policies might take (Dueck 2009). By this view, foreign policy decision makers are not simply rational maximizers pursuing objective state interests, but are instead political entrepreneurs with their own biases, beliefs, and cognitive deficiencies. There remains considerable debate over the degree to which NCR exists as part of a progressive realist program (Rathbun 2008; Quinn 2013). Nonetheless, NCR is now a proven foreign policy framework explaining when and how governments respond to external security challenges. Below I offer a theory of weaker state confrontation that adopts the three core assumptions in NCR analysis: salient shifts in the external security environment are the primary motivation for changes in foreign policy, domestic political institutions condition state responses to such shifts, and decision maker perceptions are an independent factor shaping foreign policy outcomes. Because weaker state conflict initiation is a risky strategy, my argument focuses on the role of decision maker frames and utilizes prospect theory to explain how an erosion in the external security environment increases the likelihood of risk acceptant foreign policies. As I discuss below, prospect theory lacks a “theory of framing” for simple decisions, let alone for the more complicated environments that often define foreign policy (Boettcher 2004). This means that prospect theory itself offers little guidance on how to identify the conditions under which decision makers are most likely to settle into a dominant frame, nor can it alone explain how intervening domestic political institutions might influence the manner in which decision makers react to a decline in security. Stated differently, because prospect theory is a model of human choice and not a theory of foreign policy, it lacks a theoretical story about the transmission belt by which changes in the security environment are processed through domestic political interests and the degree to which this shapes decision frames. Such a deductive argument is required, however, in order to construct hypotheses about foreign policy that are amenable to statistical examination. Indeed, the prospect theory literature in foreign policy is dominated by thick descriptions about the emergence of frames that, while insightful, have not accumulated into a set of broadly accepted framing principles. Previous research has identified a number of contingent factors that lead to the onset of loss frames including external shocks, shifting domestic political incentives, and/or intra-elite deliberation.4 By contrast, the discussion below demonstrates that combining NCR with prospect theory produces generalizable theoretical claims appropriate to statistical analysis. Prospect Theory, Political Competition, and Counterframes To summarize, my argument is that decision makers initially evaluate changes in the current security environment against their previous security position. When confronted with an erosion in security, policymakers confront two options. They can accept the certain losses that come with an eroding security environment by remaining in the status quo, or they can opt for a revisionist strategy of confrontation. Confrontation is a gamble because it holds the potential to improve a state’s security position, but it also contains the possibility of further losses beyond the status quo. According to prospect theory, this kind of choice framing produces risk acceptance. I argue that this is most likely to occur in regimes where policymakers operate in an environment with low levels of political counterframing. In such regimes, there are fewer relevant interests lobbying on foreign policy outcomes and fewer competing political frames defining the costs and benefits attached to alternative strategies. This institutional context is more conducive to the emergence of a single dominant frame that defines a country’s options. The result is that regimes in an eroding strategic position are most likely to take inordinate risks in their foreign policy when counterframing is absent.5 These risks include initiating disputes with more powerful rivals. Loss Frames and Risk According to prospect theory, decision makers become risk averse when confronted with choices over gains and risk acceptant when confronted with choices over losses (Kahneman and Tversky 1979a; Thaler 1980). An individual’s risk disposition is determined by “whether the outcomes are perceived as gains or losses” relative to a neutral reference point (Quattrone and Tversky 1988). Research suggests that the cognitive mechanisms underlying prospect theory emerged in response to environmental pressures in the early human environment and are thus part of a shared evolutionary legacy (McDermott et al. 2008). Indeed, more recent advances in cognitive neuroscience confirm that shifts in risk disposition are anchored to structures in the human brain (Kuhnen and Knutson 2005; Trepel et al. 2005; Tom et al. 2007) and that framing is attached to basic emotional processing (Ma et al. 2012). For foreign policy scholars, prospect theory offered a new framework to understand risk acceptant decisions that were inconsistent with expected utility explanations (Levy 1997). There is now strong evidence to suggest that when policymakers frame their options in terms of losses, the propensity to adopt risk acceptant foreign policy strategies increases significantly (McDermott 2004). This insight extends to militarized conflict (Mcdermott 1992; Taliaferro 1998), grand strategy (He and Feng 2013), civil conflict resolution (Hancock et al. 2010), and military deterrence (Berejikian 2002). The consistent finding here is that when the status quo ante becomes unacceptable the decision frame shifts to losses, and strategies that were once too risky to contemplate become more acceptable to decision makers.6 Prospect theory was initially tested using decision experiments. Laboratory subjects were confronted with two clear alternatives to the status quo—a certain outcome and a gamble—and this defined the decision frame.7 However, decision makers also assess current conditions against prior circumstances, treating the recent past as a reference point for evaluating the desirability of the status quo (Tversky and Kahneman 1974; Wright and Anderson 1989). This is particularly true when actors experience a setback because the emotional impact of losses is greater than gains and individuals are therefore slow to accept their new circumstances. As Jervis (1992: 200) notes, In international politics and in social life in general… we renormalize for gains much more quickly than we do for losses. We rapidly, if not effortlessly, adjust to good fortune and any improvement in our lives…. Neither individuals nor nations are so accepting of losses, however. We remain unhappy, unreconciled, and often bitter for a prolonged period.When states suffer an erosion in security, decision makers are slow to update to the new status quo and often tenaciously cling to their previous reference point (Levy 1992). Prospect theory’s concept of a decision frame therefore extends to instances in which actors confront a choice between accepting recent changes to the status quo and a gamble (Berejikian 1997). Consistent with NCR commitments about the primacy of security politics, I assume that changes in the external security environment define the decision frame through which policymakers assess their state’s security position.8 Foreign policy actors in states that experience erosion in their security environment are therefore confronted with a loss frame as defined by prospect theory. They can do nothing and accept the known negative consequences inherent in the revised status quo, or they can opt for a gamble in an attempt to return to their previous position. Such decision makers are operating in the domain of losses and are, according to prospect theory, more likely to adopt risk acceptant strategies.9 Political Competition and the Onset of Loss Frames Political competition reduces the likelihood that any single frame will settle over a policy space.10 To the extent that the consequences of foreign policy affect domestic groups differently, the larger the number of groups engaged in competition over policy outcomes, the more likely it is that leaders will be forced to incorporate additional domestic political gains and losses when considering foreign policy strategies. Experimental evidence provides microfoundational support for this contention. Simply offering decision makers alternative frames can dislodge a settled frame (Druckman 2004; Boettcher and Cobb 2009; Kam and Simas 2010). Research also shows that when individuals are asked to reconsider their choices, they first reprocess the incentive structure. The cognitive demands of reappraisal reduce the emotional impact of negative frames, diminishing the influence of framing effects (Miu and Crişan 2011). As frame salience diminishes, decision processing shifts from emotional centers of the brain to structures associated with higher-order cognitive functioning (Benedetto et al. 2006), and the impact of decision bias is reduced. Domestic counterframing is a routine aspect of foreign policy. We know that political actors representing entrenched interests define foreign policy issues in self-serving ways (Allison and Zelikow 1999), and this includes explicit attempts to reframe foreign policy initiatives (Garrison 2001). As political competition broadens in a society, decision makers must pay closer attention to an increasingly varied set of policy demands in order to build and maintain successful governing coalitions. As groups compete, they seek to undermine evaluative frames that are not consistent with their parochial interests. Such self-interested framing is understood to be a central component of effective political manipulation in foreign policy (Maoz 1990; Levy 2000). Further, the ability of political rivals to exploit ineffective policies for partisan gain disciplines executives because they must be concerned that failed policies will underpin a shift in support to competitors (Bueno De Mesquita 1985; Fearon 1994: Bueno de Mesquita et al. 2003). Domestic political competition can therefore transform the content of a foreign policy frame. For example, strategies that might look to be a gamble in the absence of counterframing can take on the characteristics of a certain loss when counterframing is present. In the analysis below, Syria is identified as a state that slipped into the domain of losses (1982) and that lacked meaningful counterframing institutions. Bogged down by its military presence in Lebanon and increasingly marginalized buy its allies, the Syrian government was looking for a way to regain its foreign policy momentum and initiated two militarized interstate disputes (MIDs) against more powerful adversaries. One of these was with its long-time rival Iraq. The Assad regime seized Iraqi oil, embargoed Iraqi oil exports through Syria, provided military support to Iran, and publically called for the overthrow of the Iraqi regime. Confronting Iraq was a calculated risk. On the one hand, political and military assistance to Iran held the possibility of tipping the scales against Syria’s long time foe in the stalemated Iran–Iraq war. On the other hand, Syria risked further political and diplomatic isolation from traditional allies for siding against an Arab country. Syria desperately wanted continued Arab political support in the looming confrontation with Israel, and even more so it required ongoing financial assistance to subsidize its expensive military presence in Lebanon. The Assad regime ruled through military force and drew political support from a narrow slice of Syrian society, the Alawites, while the much larger Sunni population was ruthlessly repressed. As a result, there existed no potential for counterframing in Syria when Assad considered the decision to side with Iran and against an Arab neighbor. This becomes clear if we consider the alternative as a hypothetical—one in which the Sunnis had meaningful access to political power and could impose the kinds of audience costs described earlier. Added to the expected regional Arab backlash would be the negative domestic political consequences in the form of bitter Sunni opposition. The decision to support Iran would then look less like a calculated risk and more like a doomed political strategy. Assad’s options would no longer resemble a standard prospect theory decision frame. Instead of a choice between a certain loss (status quo) and a foreign policy gambit (confront Iraq), effective Sunni counterframing transforms the gamble leaving Syria two alternatives that both offer only losses. The logic of counterframing articulated here does not predict which specific domestic interests ultimately prevail. Nor is the argument that counterframing will always dislodge an already entrenched frame. Instead, changes in external security forces leaders to reevaluate their policy options, and part of that evaluation involves a consideration of domestic political consequences and constraints. Because robust political competition attaches varied and politically salient costs and/or benefits to available strategies, it increases the number of trade-offs that policymakers must confront. This can prevent the emergence of a single dominant frame. Without counterframing, leaders like Assad are free to take their cues primarily from the external security environment, increasing the likelihood that an erosion in external security produces a clear loss frame as defined by prospect theory. Hypotheses and Testing I have argued that, in the absence of counterframing institutions, erosion in the external security environment generates loss frames and increases the probability of risk acceptant foreign policy decisions. This produces the main hypothesis examined in this paper. Hypothesis 1: The onset of a loss frame increases the probability of weaker state dispute initiation in regimes that lack counterframing institutions.To evaluate this hypothesis I examine a subset of MIDs taken from the Correlates of War (COW) data set. The dependent variable is coded “1” for disputes where the initiator actions ranged from direct displays of force to war against a target in the observation year and “0” otherwise.11 The sample includes all such cases for which data are available between 1946 and 1999. Only politically relevant dyads in which states can engage one another directly are included. Decisions to join ongoing disputes are not included. Initiators are defined as those states involved in a dispute when it begins and that are on the side taking the first action. Targets are defined as those states that are originally involved in a dispute at its inception on the side that does not take the first action. Because I am interested in weaker state initiation, the unit of analysis is the directed dyad year for all pairs in which potential initiators have a lower score on the Composite Index of National Capability (CINC) than potential targets (Singer 1988). The main independent variable consists of two components. The first is the decision maker security frame and the second is an absence of counterframing institutions. Decision Maker Security Frame NCR assumes that managing shifts in the external security environment is a primary foreign policy objective, and that policymakers look to “other state’s intentions” and “changes in the relative distributions of power” as salient indicators of their security environment (Kunz and Saltzman 2012, 102). Therefore, a decision frame variable must first capture the assumption that decision makers anchor their evaluation of the status quo to changes in these dimensions. Second, to make meaningful comparisons across years, a framing variable must also provide a standardized indicator that tracks changes in a state’s security environment. Because the security potential of states is not static, this is a challenge both across states and within states over time. For example, a country can grow to become more powerful relative to its rival and nonetheless find itself in a more precarious security environment if alliances shift against it. Third, statistical testing hinges upon a key assumption about decision maker perception. If we assume that decision makers perceive real changes in their environment, then objective shifts in security define the content of the decision frame. In the laboratory, researchers assume that prospect theory subjects all understand the objective choices confronting them and that participants adopt the decision frames as constructed by investigators. This assumption is also common in rational choice, which holds that decision makers confront incentive structures observable by both participants and investigators. Finally, the discipline of behavioral economics has used objective measures to test prospect theory’s claims extensively (Barberis 2012).12 To operationalize the decision frame, I begin with Bueno De Mesquita's (1985) metric to identify a state’s security position. I construct a security continuum ranging from best to worst possible security environments for any state i by summing the utilities for conflict by all states j versus i.13 This continuum is bracketed at one end by the hypothetical interstate alliance pattern and relative power configuration that would leave i most vulnerable to defeat and at that other, end by the pattern that leaves i most secure. I define the endpoints on this continuum as the sum of expected utilities for j versus i as ∑E(Uji)max and ∑E(Uji)min, respectively. State i's observed security, ∑E(Uji), moves between the extremes of this continuum as alliance patterns and relative power shift over time.14 By this conceptualization, i is least secure when the summed expected utilities for conflict are at their maximum and i is most secure when they are at their minimum.15 I next calculate the previous year’s observed security score and measure the distance from this to the midpoint on the continuum for that year, and then take this same distance using the observed security score for the current year. When the difference between these two distances is positive, this indicates the summed regional utilities for conflict are increasing and therefore that the state is declining in security year to year. I assume that states experiencing a security decline operate in the domain of losses and are coded as lossframe =1. The result is a yearly dichotomous indicator that identifies the onset of a losses frame. The primary benefit of this conceptualization is that it provides an updating indicator for loss frame that compares a country’s current security performance to the previous year. Take, for example, the decline Syria experienced in 1982, a year in which it initiated two MIDs against adversaries that were more powerful. To construct the indicator for 1982, I first take the values for security in 1981. Syria’s security continuum ranged from 78.60 (worst security) to − 65.30 (best security), with a midpoint at 6.65. Syria’s observed security score in 1981 was 31.74 with a distance from the midpoint of 25.09. In 1982, Syria’s observed security score was 37.06, and the distance from the 1981 midpoint increased to 30.41. This indicates a 21.19 percent shift in the direction of decreased security. As a result, I code Syria as lossframe = 1 for the year 1982. Counterframing To account for the absence of counterframing institutions, I initially use two Polity IV concept variables to construct a dichotomous indicator identifying the set of states with both very low levels of political competition between domestic groups and where such groups have little ability to impose their preferences on the executive (Marshall and Jaggers 2002).16 For this set of states, there will be little counterframing, and, to the extent it exists, domestic groups have little ability to assert their frames on the executive. The variable counterframe is coded as “1” for countries where the Polity IV score for political competition is either “repressed” or “severely restricted” (represented by a “political competition” score of 1 or 2) and where executives operate freely from domestic group preferences (represented by an “executive constraint” score of 1 or 2). Otherwise, countries are given a score of “0.” This identifies a broad category of states with low levels of political counterframing. However, it does not capture fine-grained distinctions. For example, research suggests that autocratic regimes vary in the level of domestic audience costs that can hold leaders to account on foreign policy outcomes (Peceny and Beer 2003; Lai Slater 2006). Of particular interest here are so-called personalist regimes characterized by the highest levels of centralized control in the hands of an individual. This can include direct authority over the military, the governing bureaucracy, and/or the ruling party apparatus. In such states, the status of elites is subject to the whims of a leader who can remove dissenters at will. The defining characteristic of personalist regimes is that, while such regimes “are often supported by parties and militaries, these organizations have not become sufficiently developed or autonomous to prevent the leader from taking personal control of policy decisions” and are less bound by the preferences of domestic actors (Geddes 2003, 53). This is in contrast to authoritarian systems with more constrained forms of autocratic rule in which elites have access to an independent political base. Personalist regimes therefore diverge from other forms of authoritarian rule in that “access to office and the fruits of office depend much more on the discretion of the individual leader” (Geddes 2003, 51). Control over the “fruits of office” is key to my argument about political competition and counterframing. The implication here is that some leaders in nondemocratic regimes are uniquely susceptible to loss frames because the institutional infrastructure required for even minimal levels of counterframing is absent. To explore this possibility, I include in my analysis four categories of authoritarian regimes constructed by Weeks (2012). I examine the impact of loss frames on two categories of elite-constrained regimes. Political “machines” are characterized by civilian elite audiences often embedded in the dominant political party, while “juntas” are typified by military institutions that serve as an independent source of political authority within government. I also examine two categories of personalist regimes, political “boss” and military “strongman.” Here individuals enjoy their status as elites largely at the behest of the leader who controls access to the benefits of government policies. I expect that loss frames increase the probability of conflict in personalist regimes because audience costs are almost entirely absent. This produces a second hypothesis examined here. Hypothesis 2: The probability of initiation will increase with the onset of loss frames in personalist regimes, compared to all other government forms. Control Variables I also include a suite of control variables common in dyadic conflict analysis to account for alternative factors that might influence the probability that weaker states will initiate. Because trading relationships are known to affect the propensity for conflict, I include a standard measure for initiator’s trade dependence, defined as the total trade between initiator and target divided by initiator’s GDP (Oneal and Russett 1999). Security relationships can also shape foreign policy preferences. For example, similar alliance portfolios between countries indicate shared security interests, reducing the potential for conflict. In addition, weaker states that share security interests with powerful states may feel emboldened, whether they are pursuing a strategy of fait accompli or outright victory, if they anticipate political support from their more powerful allies. To control for these effects, I use Signorino and Ritter's (1999) s-score for alliance similarity between target and initiator and between initiator and the global system leader, calculated for the initiator’s home region. To account for the known impact of contiguity, I include a binary indicator for shared borders taken from Stinnett et al. (2002). Regime instability has also been identified as a source of aggressive behavior (Sirin 2011), so I include an indicator for the existence of civil war in the initiator (Gleditsch et al. 2008) and a simple dummy measure for regime durability.17 Relative capabilities are measured by the ratio of initiator over target CINC score. I expect that the probability of initiation will increase as this ratio approaches a value of one (power parity). I also include a number of controls that define the contextual relationship between initiator and target. First, Goertz and Diehl (1993) argued that some states are locked into a relationship of rivalry, so the pattern of conflict may be “influenced by the outcome or processes of previous disputes or by the prospect of future disputes between the same states” (148). Because managing rivalries is among the most salient foreign policy objectives, weaker states in rivalries with more powerful adversaries may initiate conflicts because of dynamics internal to the rivalry rather than because of an increased tolerance for risk. To account for this, I include a variable indicating the presence of an enduring rivalry using the operationalization provided by Klein, Goertz, and Diehl (2006). I also include a dichotomous variable indicating the existence of a major power dyad taken from the COW data set. Second, research suggests that the presence of nuclear weapons influences the propensity for conflict, although there remains considerable debate about the nature of this effect (for example, Horowitz 2009; Sobek, Foster, and Robinson 2012; and Miller 2013).18 To control for the impact of nuclear weapons, I include an indicator for the acquisition of nuclear weapons by the target taken from Singh and Way (2004). Third, to account for the role of joint democracy on conflict, I include an indicator for joint democracy where both initiator and target have a minimum score of 6 using the Polity IV democracy scale (Marshall and Jaggers 2002). Fourth, states with low levels of counterframing also tend to score low for democracy, and there is some evidence of an autocratic peace (Peceny et al. 2002). To account for this, I include a dummy variable indicator for autocratic states coded as “1” if the Polity IV regime score for the target is less than −5 and “0” otherwise. Results and Analysis I have argued that, in the absence of counterframing institutions, an eroding strategic position creates a loss frame and produces risk acceptant foreign policy behavior. In order to test this hypothesis, I examine a data set containing politically relevant directed dyads for the period 1946–1999, where the CINC score for the initiator is lower than the target. This produces 55,700 observations and 602 MIDs (1.08 percent). The average power ratio (initiator/target) for weaker state initiation is 0.340, with a range from 0.0003 to 0.9926. I also constructed four binary indicators for each combination of lossframe x couterframe, with the reference category set to lossframe = 1 and counterframe = 0. Of particular theoretical interest is the difference between the probability of weaker state initiation in this reference category and the category lossframe = 1 and counterframe = 1. This comparison permits a direct test of the hypothesis that the onset of loss frames increases the probability of conflict where counterframing is absent. As dispute initiation is coded dichotomously, I deploy binary logistic regression with robust standard errors clustered on the dyad. To account for duration dependence, I use the method prescribed by Beck et al. (1998) of including a count variable for the number of within-dyad peace years and three cubic splines. Table 1 summarizes the regression results using the Polity IV operationalization for counterframe, with coefficients reported as odds ratios.19 Model 1 contains only the NCR variables that are of primary theoretical interest. This includes three categories of lossframe x counterframe. Model 2 introduces the suite of controls. Model 3 contains only those variables of statistical significance in Model 2. The results across Models 1–3 provide strong support for the argument presented here. The onset of loss frames combined with a lack of counterframing institutions (lossframe = 1 and counterframe = 1) exerts a consistently positive and statistically significant effect on the likelihood of weaker state conflict initiation, compared to the reference category (lossframe = 1 and counterframe = 0). In addition, several of the controls commonly associated with conflict initiation are also significant. In both Models 2 and 3, the likelihood of initiation increases as the dyad approaches power parity, when the initiator is embroiled in civil war, when the target has acquired nuclear weapons, when the two states share a border, and in the presence of an enduring rivalry. The likelihood of initiation decreases for democratic dyads and dyads with shared alliance patterns. Surprisingly, increased trade dependence is associated with an increase in initiation.20 Table 1. Weak state MID initiation Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Notes. Reference category: Lossframe = 1, Counterframe = 0. *p ≤ .05, **p ≤ .01. Table 1. Weak state MID initiation Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Notes. Reference category: Lossframe = 1, Counterframe = 0. *p ≤ .05, **p ≤ .01. Table 2 reports results for models using each of the four categories of authoritarian regimes identified above and includes the controls found to be statistically significant predictors in Table 1. Here, counterframe is coded as “1” when the regime takes the characteristics of each autocratic type and “0” otherwise. The results are interesting and suggest that there are important counterframing differences within the autocratic camp. For both categories of personalist regime, boss and strongman, the onset of a loss frame increases the probability of weaker state initiation. Consistent with expectations, personalist regimes appear to lack counterframing institutions and are therefore more susceptible to risk acceptance as their security erodes. The findings for elite constrained regimes are varied. Under loss frames, civilian machines appear to behave no differently than other regimes. This suggests the presence of frame competition by entrenched elites within the society who are also able to assert their interests to their government. By contrast, the coefficient for juntas is positive and significant. While juntas are also elite constrained, this positive association between the onset of loss frames and risk acceptant policy suggests that a single coherent security frame is more likely when government is dominated by military institutions. This may be due to the fact that, compared to their civilian counterparts, elites in juntas are more likely to share unified “beliefs about the utility and appropriateness of military force as an instrument of politics” and that they are “more likely than civilians to form ominous views of the status quo” (Weeks 2012, 333). This suggests that, rather than a lack of counterframing institutions, the narrow composition of elites in military juntas explains the onset of loss frames that produce risk acceptant foreign policy choices. Table 2. Initiation by autocratic type Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  *p ≤ .05, **p ≤ .01. Table 2. Initiation by autocratic type Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  *p ≤ .05, **p ≤ .01. Logistic regression coefficients cannot be interpreted directly as their effect is curvilinear and dependent upon specific values of the other covariates in the model. In addition, because conflict is a rare event, even relatively powerful predictors will produce only a small change in the overall probability of weaker state initiation. Table 3 provides a more intuitive way to understand the substantive impact of the coefficients by reporting the percentage change in the predicted probability of initiation when lossframe changes from “0” to “1.” Five country profiles are summarized in Table 3. Each profile reports the change in the predicted probability for a hypothetical “typical” country with continuous covariates held constant at their median values and dichotomous variables at their modal scores. Table 3 also includes values for the Syrian case described above. The onset of a loss frame for a typical country using the Polity IV indicator for counterframing increases the likelihood of risky initiation by 69.2 percent. For juntas and bosses, the effect is even larger, increasing the predicted probability by 115.4 percent and 104 percent, respectively. For strongman governments, the increase is 29.4 percent. Consistent with the fact that it was classified as a strongman regime in 1982, the onset of a loss frame for Syria increased the risk of initiating a dispute with a more powerful adversary by 27 percent. Table 3. Percentage change in predicted probability Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  p ≤ .05. *The coefficient for machine was not significant in model 4. Table 3. Percentage change in predicted probability Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  p ≤ .05. *The coefficient for machine was not significant in model 4. Although these results support NCR claims concerning the impact of loss frames, it is conceivable that a subset of dyads drive these findings. As noted above, the range of capability ratios across dyads experiencing MIDs is quite large extending from 0.0003 to 0.9926 with a mean of 0.340. Unfortunately, prospect theory offers little guidance about the magnitude of risk that foreign policymakers will accept when operating in a losses domain. Loss frames may matter less in dyads where there are large disparities in power because the consequences of initiation are more transparent, so weaker states understand that they are unlikely to prevail whatever their tolerance for risk.21 A second possibility is that the importance of loss frames diminishes as initiators approach parity with targets because they are in a stronger position to address their grievances without having to resort to conflict. The third option is that loss frames produce risk acceptance across the entire range of dyadic relative power. If, as noted above, prospect theory preferences constitute an evolutionary response to selective pressures, then we would expect them to be robust even when survival of the individual, or the state, is at stake. Using results from Model 3, Figure 1 displays the impact of loss frames across the entire range of relative power distributions for states without counterframing institutions. The results suggest that relative power mitigates the impact of loss frames only when dyads approach power parity. The upward slope in Figure 1 indicates that the predicted probability for conflict increases as the initiator approaches parity with the target, whether or not states are in a loss frame. However, the effect of lossframe is also distinct and statistically significant from the null throughout a large portion of the range of dyadic power ratios. The confidence intervals overlap when the initiator reaches 70–75 percent of the target’s power. That is, loss frames matter across the majority of relative power distributions. Importantly, this range includes the majority (85.9 percent) of weaker state initiations. Figure 1. View largeDownload slide Lossframes and dyadic relative power. Figure 1. View largeDownload slide Lossframes and dyadic relative power. In sum, these results provide support for the assertion that negative shifts in the security environment set the stage for changes in foreign policy by framing the status quo ante as a loss and increasing the attractiveness of risky strategies. The emergence of a dominant frame is contingent upon the level of political counterframing in a society. Meaningful political competition between domestic groups appears to partially inoculate decision makers against the onset of a clear decision frame. Conclusions The cognitive version of NCR presented here modifies traditional NCR analysis in several ways. First, it incorporates prospect theory to explain how changes in external security can elicit risky foreign policy choices. While the focus here is on conflict, the combination of prospect theory and NCR can be used to develop testable arguments about risk across the entire range of foreign policy domains. Second, my argument introduces variation in the level of political competition as a key domestic variable shaping elite perceptions. Competition varies across regimes and determines when dominant frames are most likely to settle over a policy issue. In this way, counterframing helps define both the meaning of changes in the external security environment and the relative attractiveness of available strategies. Third, a cognitive version of NCR enables the construction of hypotheses about foreign policy behavior that are amenable to statistical scrutiny. The first wave of NCR analysis provided mostly “theoretically informed narratives” rather than deductive theory, and the primary focus of empirical work was restricted to “counterfactual analysis” instead of identifying broad patterns of foreign policy behavior (Rose 1988, 153). However, the lingering view that the NCR enterprise is inherently restricted to case studies is incorrect.22 I do not claim that statistical testing is a superior method. Statistical testing and case analysis are complementary, and both are necessary in order to push NCR analysis into new empirical domains as part of a progressive research agenda. The results presented here also have implications for the foreign policy literature anchored to prospect theory. To the extent that the roil of domestic political competition requires policymakers to confront issues from a number of competing perspectives, counterframing appears to act as an imperfect buffer against the kind of risk acceptant foreign policy behavior identified in previous studies. This finding holds even for interstate rivalries, where a cycle of conflict dominates the relationship. A lack of domestic competition is therefore a potential liability for states. Robust counterframing appears to partially inoculate governments against unwise risks and in this way supports more reasoned foreign policy judgments. These results also suggest that risky foreign policy decisions are better explained by an NCR framework that replaces traditional rational actor assumptions with prospect theory. The notion that factors external to the state define decision maker frames is consistent with the traditional NCR assumption that shifts in the international security environment provide the initial motivation for changes in foreign policy. However, the idea of domestic loss frames is also worth investigating. For example, there is considerable debate about the degree to which domestic politics alone can produce incentives that lead to a strategy of external conflict (for example, Lai and Slater 2006). Factors like civil unrest, eroding political support, and/or leadership challenges may themselves motivate risk acceptant foreign policies. Indeed, domestic instability combined with an erosion in external security would potentially generate very salient loss frames. As Levy (1992) notes, “The combination of perceived external decline and internal insecurity may be particularly conducive to risk seeking” (287). Expanding the definition of decision maker frames to include both international and domestic factors is an important next step. Finally, perhaps the most exciting implication is that cognitive NCR should now be particularly attractive to interdisciplinary foreign policy scholars. By substituting cognitive principles in place of rationalist assumptions, NCR provides a kind of “plug-and-play” structure permitting the integration of decision science into foreign policy analysis. There is no reason to construct a new foreign policy framework for each distinct cognitive heuristic or bias. The role of loss aversion, fairness, social trust, etc. on foreign policy can now be examined under a single conceptual umbrella. Dr. Berejikian is an Associate Professor in the Department of International Affairs, a Josiah Meigs Distinguished Teaching Professor at the University of Georgia, and a Senior Fellow at the Center for International Trade and Security. 1While the conceptualization of risk in foreign policy studies continues to evolve (Clapton 2011), I adopt the conventional notion of comparative risks defined as the level of variance in outcomes attached to available strategies. A strategy is considered risky if it presents a larger variance in payoffs than an alternative. As I describe below, for states operating under an eroding security environment, weaker state confrontation is risky in that confronting a more powerful rival produces an outcome that is much better (victory) or much worse (defeat) than the status quo ante. 2See also George and Smoke (1974) who argue that, in the context of deterrence politics, fait accompli is sometimes an effective strategy. 3NCR scholarship often uses the terms elites, decision makers, and decision executives interchangeably. The purpose is to denote the relevant individuals with decision authority over a policy space. Here, I use the term decision makers in a way that is consistent with Putnam’s (1988) identification of “chief of government” and/or Lobell’s (2009) more recent “foreign policy executive.” 4Levi and Whyte (1997) identify elite deliberation as a source of loss frames in the above-mentioned Pearl Harbor case. 5I define counterframing as the attempt by actors to assert alternative decision frames—through the manipulation of reference points—consistent with their interests. In the foreign policy domain, counterframing defines the competition between domestic groups to frame policies as beneficial (gains) or detrimental (losses). 6For a comprehensive review of prospect theory research in the field of international relations, see Berejikian (forthcoming). 7For example, a typical loss frame would be one in which subjects confronted a choice between a certain loss of $80 and a gamble with an 85 percent chance losing $100 and a 15 percent of losing nothing. 8The idea that states construct evaluative frames for security has deep roots in the realist study of international politics. Balance of power (Waltz 1979), power transition theory (Organski 1968), and hegemonic conflict (Gilpin 1983) all incorporate a version of this idea. The shared insight is that governments are sensitive to changes in their security position relative to potential rivals and that it is through this lens that they assess the attractiveness of their policy options. Power transition theory, in particular, emphasizes the importance of status quo evaluations on foreign policy choice (Lemke and Reed 1996). While research has tended to focus on great powers, there is evidence that lesser states make similar assessments on a regional basis (Lemke 2002). 9This is consistent with the notion of evaluative framing. Evaluative frames set the reference point against which changes in the external environment are compared. For a conceptual treatment of the various ways in which the term “framing” is used by international relations and foreign policy scholars, see Mintz and Redd (2003). There are also additional conceptualizations in political science more generally (Chong and Druckman 2007). 10In the field of behavioral economics, the impact of counterframing competition on decision bias is already well established. Market competition forces decision makers to consider multiple strategies simultaneously because it disciplines individuals who make nonmaximizing choices. The frequency of bias over time thus diminishes as markets become more competitive (Bazerman et al. 1985; Russell and Thaler 1985). While the debate in behavioral economics continues over the exact conditions under which competition most efficiently disciplines actors for making nonmaximizing choices (for example, Rubinstein 2001), there is general agreement that competition mitigates the consequences of decision bias. Much of behavioral economics now focuses on this interaction between social institutions, competition, and the frequency with which decision heuristics take hold. 11The data were generated using the Eugene software program version 3.204 (Bennett and Stam 2000). 12The opposing view is that decision makers possess unique worldviews and, therefore, subjectivity affects decision maker perceptions of the status quo and the decision frame. The possibility of a constructed frame, where only perception is operative regardless of objective circumstances, is also consistent with Kahneman and Tversky’s observation that “the reference point is the state to which one has become adapted” and there are many cases in which “the reference point is determined by events that are only imagined” (1982, 171–2). 13The data were generated using the Eugene software program version 3.204 (Bennett and Stam 2000). 14The set of states for j are those in i’s region. 15Recall, increasing values represent a decrease in security along the continuum. 16Please see Addendum B and C contained in the POLITY IV codebook for the rules used to construct the concept variables. 17Following Weeks (2012), this dummy variable is coded as “1” if the initiator has a Polity IV durability score of less than 3. 18See also the long-standing debate between Waltz and Sagan (2003). 19Controls for duration dependence excluded to save space. 20In Table 2, the results for trade dependence are inconsistent. This may suggest that the relationship between trade and weak state initiation is different than the established relationship for conflict more broadly or that it is not a stable predictor for weaker state initiation across regime types. 21Indeed, power transition theory has long argued that power parity creates an environment in which conflict is more likely and that asymmetric capability reduces initiation (Kugler and Lemke 1996). 22With the exception Levi and Whyte (1997), traditional prospect theory research on framing and foreign policy has also largely utilized case study analysis. References Allison Graham Zelikow Philip. 1999. Essence of Decision: Explaining the Cuban Missile Crisis . Vol. 2. New York: Pearson. Arrefuin-Toft Ivan 2005. How the Weak Win Wars: A Theory of Asymmetric Conflict : Cambridge, MA: Harvard University Press. Bazerman Max H. Magliozzi Thomas Neale Margaret A.. 1985. “ Integrative Bargaining in a Competitive Market.” Organizational Behavior and Human Decision Processes  35: 294– 313. Google Scholar CrossRef Search ADS   Beck Nathaniel Katz Jonathan N. Tucker Richard. 1998. “Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable.” American Journal of Political Science  42: 1260– 88. Google Scholar CrossRef Search ADS   Bell Mark S. Miller Nicholas L.. 2013. “Questioning the Effect of Nuclear Weapons on Conflict.” Journal of Conflict Resolution  59: 4– 92. Benedetto De Martino Kumaran Dharshan Seymour Ben J. Dolan Raymond. 2006. “Frames, Biases, and Rational Decision-Making in the Human Brain.” Science  313: 684– 7. Google Scholar CrossRef Search ADS PubMed  Bennett D. Scott Stam Allan C.. 2000. “Eugene: A Conceptual Manual.” International Interactions  26: 179– 204. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. 1997. “The Gains Debate: Framing State Choice.” American Political Science Review  91: 789– 805. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. 2002. “Model Building with Prospect Theory: A Cognitive Approach to International Relations.” Political Psychology  23: 759– 86. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. “Prospect Theory in International Relations.” In Oxford Bibliographies in International Relations , edited by James Patrick. New York: Oxford University Press (forthcoming). Boettcher William Cobb Michael. D.. 2009. “Don't Let Them Die in Vain”: Casualty Frames and Public Tolerance for Escalating Commitment in Iraq.” Journal of Conflict Resolution  53: 677– 97. Google Scholar CrossRef Search ADS   Boettcher William, A. 2004. “The Prospects for Prospect Theory: An Empirical Evaluation of International Relations Applications of Framing and Loss Aversion.” Political Psychology  25: 331– 62. Google Scholar CrossRef Search ADS   Bueno De Mesquita Bruce. 1981. The War Trap. New Haven, CT: Yale University Press. Bueno De Mesquita Bruce. 1985. “The War Trap Revisited: A Revised Expected Utility Model.” The American Political Science Review  79: 156– 77. Google Scholar CrossRef Search ADS   Bueno de Mesquita Bruce Smith Alastair Siverson Randolph M. Morrow James D.. 2003. The Logic of Political Survival . Cambridge, MA: The MIT Press. Cha Victor D. 2000. “Abandonment, Entrapment, and Neoclassical Realism in Asia: The United States, Japan, and Korea.” International Studies Quarterly  44: 261– 91. Google Scholar CrossRef Search ADS   Chan Steve. 2010. “Major-Power Intervention and War Initiation by the Weak.” International Politics  47: 163– 85. Google Scholar CrossRef Search ADS   Chong Dennis Druckman James N.. 2007. “Framing Theory.” Annual Review of Political Science  10: 103– 26. Google Scholar CrossRef Search ADS   Clapton William. 2011. “Risk in International Relations.” International Relations  25: 280– 95. Google Scholar CrossRef Search ADS   Conroy Hilary Wray Harry. 1990. Pearl Harbor Reexamined: Prologue to the Pacific war . edited by Conroy Hilaryand Wray Harry, 1990. Honolulu, HI: University of Hawaii Press. Druckman James. N. 2004. “Political Preference Formation: Competition, Deliberation, and the (ir) Relevance of Framing Effects.” American Political Science Review  98: 671– 86. Google Scholar CrossRef Search ADS   Dueck Colin. 2009. “Neoclassical Realism and the National Interest: Presidents, Domestic Politics, and Major Military Interventions.” In Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W. eds. Neoclassical realism, the state, and foreign policy , Cambridge, UK: Cambridge University Press 139– 169. Edelstein David M. 2002. “Managing Uncertainty: Beliefs about Intentions and the Rise of Great Powers.” Security Studies  12: 1– 40. Google Scholar CrossRef Search ADS   Fearon James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review  88: 577– 92. Google Scholar CrossRef Search ADS   Fearon James D. 1995. “Rationalist Explanations for War.” International Organization  49( 3 ): 379– 414. Google Scholar CrossRef Search ADS   Garrison Jean A. 2001. “Framing Foreign Policy Alternatives in the Inner Circle: President Carter, His Advisors, and the Struggle for the Arms Control Agenda.” Political Psychology  22: 775– 807. Google Scholar CrossRef Search ADS   Geddes Barbara. 2003. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics . Ann Arbor, MI: University of Michigan Press. George Alexander L. Smoke Richard. 1974. Deterrence in American Foreign Policy: Theory and Practice . New York: Columbia University Press. Gilpin Robert. 1983. War and Change in World Politics . Cambridge, UK: Cambridge University Press. Gleditsch Kristian Skrede Salehyan Idean Schultz Kenneth. 2008. “Fighting at Home, Fighting Abroad How Civil Wars Lead to International Disputes.” Journal of Conflict Resolution  52: 479– 506. Google Scholar CrossRef Search ADS   Goertz Gary Diehl Paul F.. 1993. “Enduring Rivalries: Theoretical Constructs and Empirical Patterns.” International Studies Quarterly . 37: 147– 71. Google Scholar CrossRef Search ADS   Hancock Landon E. Weiss Joshua N. Duerr Glen M. E.. 2010. “Prospect Theory and the Framing of the Good Friday Agreement.” Conflict Resolution Quarterly  28: 183– 203. Google Scholar CrossRef Search ADS   He Kai Feng Huiyun. 2013. Prospect Theory and Foreign Policy Analysis in the Asia Pacific: Rational Leaders and Risky Behavior . New York: Routledge. Helfstein Scott. 2012. “Backfire: Behavioral Decision Making and the Strategic Risks of Successful Surprise.” Foreign Policy Analysis  8: 275– 92. Google Scholar CrossRef Search ADS   Horowitz Michael. 2009. “The spread of nuclear weapons and international conflict: Does experience matter?” Journal of Conflict Resolution  53: 234– 57. Google Scholar CrossRef Search ADS   Hudson Valerie. 2005. “ Foreign Policy Analysis: Actor-Specific Theory and the Ground of International Relations.” Foreign Policy Analysis  1( 1): 1– 30. Google Scholar CrossRef Search ADS   Jervis Robert. 1988. “War and Misperception.” Journal of Interdisciplinary History  18: 675– 700. Google Scholar CrossRef Search ADS   Jervis Robert. 1992. “Political Implications of Loss Aversion.” Political Psychology  13: 187– 204. Google Scholar CrossRef Search ADS   Johnson Dominic D. P. 2009. Overconfidence and War . Cambridge, MA: Harvard University Press. Kahneman Daniel Tversky Amos. 1979. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica  47: 263– 91. Google Scholar CrossRef Search ADS   Kahneman Daniel Tversky Amos 1982. “The Psychology of Preferences.” Scientific American  246: 160– 73. Google Scholar CrossRef Search ADS   Kam Cindy D. Simas Elizabeth N.. 2010. “Risk Orientations and Policy Frames.” The Journal of Politics  72: . Kitchen Nicholas. 2010. “Systemic Pressures and Domestic Ideas: A Neoclassical Realist Model of Grand Strategy Formation.” Review of International Studies  36: 117– 43. Google Scholar CrossRef Search ADS   Klein James P. Goertz Gary Diehl Paul F.. 2006. “The New Rivalry Dataset: Procedures and Patterns.” Journal of Peace Research  43: 331– 48. Google Scholar CrossRef Search ADS   Kugler Jacek Lemke Douglas. 1996. Parity and War: Evaluations and Extensions of the War Ledger . Ann Arbor, MI: University of Michigan Press. Kuhnen Camelia M. Knutson Brian. 2005. “The Neural Basis of Financial Risk Taking.” Neuron  47: 763– 70. Google Scholar CrossRef Search ADS PubMed  Lai Brian Slater Dan. 2006. “Institutions of the Offensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes, 1950–1992.” American Journal of Political Science  50: 113– 26. Google Scholar CrossRef Search ADS   Lemke Douglas. 2002. Regions of War and Peace . Cambridge, UK: Cambridge University Press. Lemke Douglas Reed William. 1996. “Regime Types and Status Quo Evaluations: Power Transition Theory and the Democratic Peace.” International Interactions  22: 143– 64. Google Scholar CrossRef Search ADS   Levi Ariel S. Whyte Glen. 1997. “A Cross-Cultural Exploration of the Reference Dependence of Crucial Group Decisions under Risk: Japan's 1941 Decision for War.” Journal of Conflict Resolution  41: 792– 813. Google Scholar CrossRef Search ADS   Levy Jack S. 1992. “Prospect Theory and International Relations: Theoretical Applications and Analytical Problems.” Political Psychology  13: 283– 310. Google Scholar CrossRef Search ADS   Levy Jack S. 1997 . “Prospect Theory, Rational Choice, and International Relations.” International Studies Quarterly  41: 87– 112. Google Scholar CrossRef Search ADS   Levy Jack S. 2000. “Loss Aversion, Framing Effects, and International Conflict: Perspectives from Prospect Theory.” In Handbook of War Studies II , edited by Midlarsky M. I., 193– 221. Ann Arbor, MI: University of Michigan Press. Lobell Steven E. 2009. “Threat assessment, the state, and foreign policy: A neoclassical realist model.” In Neoclassical realism, the state, and foreign policy , edited by Lobell, Steven E., Norrin M. Ripsman, and Jeffrey W. Taliaferro. 42–74. Cambridge, UK: Cambridge University Press. Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W.. 2009. Neoclassical Realism, the State, and Foreign Policy . New York: Cambridge University Press. Ma Qingguo Feng Yandong Xu Qing Bian Jun Tang Huixian. 2012. “Brain Potentials Associated with the Outcome Processing in Framing Effects.” Neuroscience Letters  528: 110– 3. Google Scholar CrossRef Search ADS PubMed  McDermott Rose. 2004 . “Prospect Theory in Political Science: Gains and Losses from the First Decade.” Political Psychology  25: 289– 312. Google Scholar CrossRef Search ADS   McDermott Rose Fowler James H. Smirnov Oleg. 2008. “On the Evolutionary Origin of Prospect Theory Preferences.” Journal of Politics  70: 335– 50. Google Scholar CrossRef Search ADS   Jonathan Mercer, “Prospect Theory and Political Science.” Annual Review of Political Science  8: 1– 21. Kunz Barbara Saltzman I. Z.. 2012. External and Domestic Determinants of State Behavior. Neoclassical Realism in European Politics , 96– 116. Manchester: Manchester University Press. Mintz Alex Redd Steven B.. 2003. “Framing Effects in International Relations.” Synthese  135: 193– 213. Google Scholar CrossRef Search ADS   Miu Andrei C. Crişan Liviu G.. 2011. “Cognitive Reappraisal Reduces the Susceptibility to the Framing Effect in Economic Decision Making.” Personality and Individual Differences  51: 478– 82. Google Scholar CrossRef Search ADS   Maoz Zeev. 1990. “Framing the National Interest: The Manipulation of Foreign Policy Decisions in Group Settings.” World Politics  43: 77– 110. Google Scholar CrossRef Search ADS   Marshall Monty G. Keith Jaggers. 2002. “Polity IV project: Political regime characteristics and transitions, 1800–2002”. Oneal John R. Russett Bruce. 1999. “Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict.” Journal of Peace Research  36: 423– 42. Google Scholar CrossRef Search ADS   Organski A. F. K. 1968. World Politics. New York: Knopf. Paul Thazha Varkey. 1994. Asymmetric Conflicts: War Initiation by Weaker Powers . Cambridge, UK: Cambridge University Press. Peceny Mark Beer Caroline C. Sanchez-Terry Shannon. 2002. “Dictatorial Peace?” American Political Science Review  96: 15– 26. Google Scholar CrossRef Search ADS   Peceny Mark Beer Caroline C.. 2003. “Peaceful Parties and Puzzling Personalists.” American Political Science Review  97: 339– 42. Google Scholar CrossRef Search ADS   Purdon Mark. 2013. “Neoclassical Realism and International Climate Change Politics: Moral Imperative and Political Constraint in International Climate Finance.” Journal of International Relations and Development  17: 301– 38. Putnam Robert D. 1988. “Diplomacy and Domestic Politics: the Logic of Two-Level Games.” International Organization  42: 427– 60. Google Scholar CrossRef Search ADS   Quattrone George A. Tversky Amos. 1988. “Contrasting Rational and Psychological Analyses of Political Choice.” American Political Science Review  82: 719– 36. Google Scholar CrossRef Search ADS   Quinn Adam. 2013. “Kenneth Waltz, Adam Smith and the Limits of Science: Hard Choices for Neoclassical Realism.” International Politics  50: 159– 82. Google Scholar CrossRef Search ADS   Rathbun Brian. 2008. “A Rose by Any Other Name: Neoclassical Realism as the Logical and Necessary Extension of Structural Realism.” Security Studies  17: 294– 321. Google Scholar CrossRef Search ADS   Rose Gideon. 1998. “Neoclassical Realism and Theories of Foreign Policy.” World Politics  51: 144– 72. Google Scholar CrossRef Search ADS   Rubinstein Mark. 2001. “Rational Markets: Yes or No? The Affirmative Case.” Financial Analysts Journal  57: 15– 29. Google Scholar CrossRef Search ADS   Russell Thomas Thaler Richard. 1985. “The Relevance of Quasi Rationality in Competitive Markets.” The American Economic Review  75: 1071– 82. Schweller Randall L. 2009. “Neoclassical realism and state mobilization: expansionist ideology in the age of mass politics.” In Neoclassical realism, the state, and foreign policy , Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W. eds. 227– 50. Cambridge, UK: Cambridge University Press. Schweller Randall L. 2008. Unanswered Threats: Political Constraints on the Balance of Power . Princeton, NJ: Princeton University Press. Sechser Todd S. 2010. “Goliath's Curse: Coercive Threats and Asymmetric Power.” International Organization  64: 627– 60. Google Scholar CrossRef Search ADS   Singer David J. 1988. “Reconstructing the correlates of war dataset on material capabilities of states, 1816–1985.” International Interactions  14: 115– 32. Google Scholar CrossRef Search ADS   Singh Sonali Way Christopher R.. 2004. “ The Correlates of Nuclear Proliferation: A Quantitative Test.” Journal of Conflict Resolution  48: 859– 85. Google Scholar CrossRef Search ADS   Signorino Curtis S. Ritter Jeffrey M.. 1999. “Tau-b or Not Tau-b: Measuring the Similarity of Foreign Policy Positions.” International Studies Quarterly  43: 115– 44. Google Scholar CrossRef Search ADS   Sirin Cigdem V. 2011. “Is It Cohesion or Diversion? Domestic Instability and the Use of Force in International Crises.” International Political Science Review  32: 303– 21. Google Scholar CrossRef Search ADS   Snyder Glenn H. 1984. “The Security Dilemma in Alliance Politics.” World Politics  36: 461– 95. Google Scholar CrossRef Search ADS   Sobek David Foster Dennis M. Robison Samuel B.. 2012. “Conventional wisdom? The effect of nuclear proliferation on armed conflict, 1945–2001” International Studies Quarterly  56: 149– 62. Taliaferro Jeffrey W. 1998. “Quagmires in the Periphery: Foreign Wars and Escalating Commitment in International Conflict.” Security Studies  7: 94– 144. Google Scholar CrossRef Search ADS   Taliaferro Jeffrey W. 2006. “State Building for Future Wars: Neoclassical Realism and the Resource-Extractive State.” Security Studies  15: 464– 95. Google Scholar CrossRef Search ADS   Thaler Richard. 1980. “Toward a Positive Theory of Consumer Choice.” Journal of Economic Behavior & Organization  1: 39– 60. Google Scholar CrossRef Search ADS   Tom Sabrina Fox Craig. R. Trepel Christopher Poldrack Russell A.. 2007. “The Neural Basis of Loss Aversion in Decision-Making Under Risk.” Science  315: 515– 8. Google Scholar CrossRef Search ADS PubMed  Trepel Christopher Fox Craig R. Poldrack Russell A.. 2005. “Prospect Theory on the Brain? Toward a Cognitive Neuroscience of Decision under Risk.” Cognitive Brain Research  23: 34– 50. Google Scholar CrossRef Search ADS PubMed  Toje Asle Kunz Barbara eds. 2012. Neoclassical Realism in European Politics: Bringing Power Back in . Manchester, UK: Manchester University Press. Van Evera Stephen. 2013. Causes of War: Power and the Roots of Conflict . Ithaca, NY: Cornell University Press. Waltz Kenneth. 1979. Theory of International Politics . Reading, MA: Addison-Wesley Pub. Co. Waltz Kenneth Sagan Scott D. eds. 2003. The Spread of Nuclear Weapons: A Debate Renewed . New York: WW Norton. Weeks Jessica L. 2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization  62: 35– 64. Google Scholar CrossRef Search ADS   Weeks Jessica L. 2012. “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict.” American Political Science Review  106: 326– 47. Google Scholar CrossRef Search ADS   Wright William F. Urton Anderson 1989. “Effects of situation familiarity and financial incentives on use of the anchoring and adjustment heuristic for probability assessment.” Organizational Behavior and Human Decision Processes  44: 68– 82. Google Scholar CrossRef Search ADS   Zakaria Fareed. 1999. From Wealth to Power: the Unusual Origins of America's World Role . Princeton, NJ: Princeton University Press. © The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Foreign Policy Analysis Oxford University Press

David versus Goliath: Risk and Weaker State Confrontation

Loading next page...
 
/lp/ou_press/david-versus-goliath-risk-and-weaker-state-confrontation-DnrMLeBhxw
Publisher
Wiley-Blackwell
Copyright
© The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1743-8586
eISSN
1743-8594
D.O.I.
10.1093/fpa/orw037
Publisher site
See Article on Publisher Site

Abstract

Confronting a more powerful rival can be a risky proposition. This paper integrates prospect theory into the growing Neoclassical Realist (NCR) literature to identify the conditions under which decision makers are most likely to accept foreign policy risks. I argue that decision makers governing regimes with low levels of political counterframing are more likely to settle into a dominant loss frame when their external security environment erodes. This increases the probability that they will initiate disputes with more powerful adversaries. To assess this proposition, I conduct a test of weaker state confrontation grounded in a NCR framework that utilizes the insights from prospect theory. Data come from the postwar era and support my hypothesis. Loss frames exert influence over the decision to initiate risky foreign policy strategies in regimes in which political counterframing is absent. “I’ll strike you down and cut off your head” ∼ David, confronting Goliath Introduction Why do weaker states confront opponents that are more powerful? Picking a fight with a superior adversary is a risky proposition. While the spoils in victory are often greater than those from prevailing over a lesser rival, greater too are the potential losses and likelihood of defeat. In the fable of David versus Goliath, David’s willingness to step forward and challenge his physically superior, better armed, and more experienced rival derived from a conviction that he possessed an unseen advantage unrelated to obvious indicators of their relative strength. In proclaiming, “The Lord will deliver you into my hands,” David indicated his belief that the balance of power actually favored his cause. Likewise, our understanding of weaker state confrontation concludes that smaller states initiate disputes when they enjoy non-obvious advantages over stronger rivals—for example, greater resolve, short-term tactical advantages, or anticipation of favorable third-party intervention. In such instances, the decision to challenge a more powerful adversary is a rational strategy and therefore theoretically unproblematic. By contrast, this paper takes at face value the proposition that asymmetric confrontation is a risky proposition and investigates the conditions under which decision makers become more acceptant of risk in the conduct of foreign policy. The argument offered here integrates prospect theory into a Neoclassical Realist (NCR) framework by combining established findings from cognitive science with basic categories of regime type. I assert that decision maker security frames, defined by changes in a state’s external security environment, influence the willingness to accept risks. Consistent with prospect theory, when a state’s security position erodes, leaders become more risk acceptant increasing the likelihood of initiating disputes with rivals that are more powerful. I also argue that this process is conditioned by the nature of political competition within the regime. Decision makers are more likely to fall prey to framing effects in states with low levels of political counterframing and are therefore more likely to adopt risky foreign policy strategies when their state experiences erosion in security. This argument is consistent with the emphasis that NCR places on the importance of domestic factors in foreign policy. It is also in line with the call to explicitly integrate authentic microfoundational assumptions into our theories of foreign policy behavior (Hudson 2005). As I demonstrate below, the resulting framework is also amenable to large-n testing. This opens new avenues of inquiry for both NCR and prospect theory that, to date, have been largely confined to case study analysis. The discussion below is organized as follows. I first briefly discuss the logic of weaker state confrontation and argue that it is a risky strategy even when it is carefully conceived and deliberate. Second, I summarize the core commitments of an NCR framework and use these to build a new theory of weaker state conflict. This section focuses on the relationship between domestic political competition and the onset of loss frames and takes the rest of NCR largely as it currently exists in the literature. Third, I conduct a statistical test of the propositions developed here. Results offer strong support for my theory. In the conclusion, I discuss the broader implications of these results for our understanding of risky foreign policy behavior and for NCR analysis more generally. Among the most promising conclusions is that NCR provides a much-needed framework for interdisciplinary scholars seeking to integrate findings from cognitive science into a foreign policy research program. Risk and Weaker State Aggression Is weaker state conflict initiation risky?1 Much of the scholarship in foreign policy does not address this question directly, despite the fact that confronting a more powerful rival initially seems like a poor strategy containing limited chances for success. For example, one explanation is that weaker states initiate risky policies because of misperception. The list of potential sources for errors in judgment are now well understood. Governments might operate under misperceptions about rival intentions (Jervis 1988), overestimate their relative capabilities (Fearon 1995), or fall prey to the human disposition favoring overconfidence (Johnson 2009). By this view, weaker state aggression is the inadvertent result of perceptual errors rather than an explicitly risky choice. A second category of argument depicts weaker initiators as deliberate and calculating, seizing advantage when they believe that their prospects for immediate success are greatest. The goal here is not outright victory but instead fait accompli.2 As Paul (1994) notes, weaker states sometimes confront more powerful rivals in order to capitalize upon short-term military and/or political advantages and then, operating from a newfound position of strength, demand a revision to the status quo before the full force of rival superiority can manifest. Success hinges on correctly identifying both the emergence of immediate tactical advantages and in accurately surmising that the target does not have the resolve to pursue the conflict to conclusion. In a similar vein, older arguments about alliance entrapment (Snyder 1984) and more recent work on third-party intervention (Chan 2010) suggest that allied support or international attempts to settle disputes can inadvertently generate moral hazards for weaker initiators, creating incentives to pursue fait accompli. However, even when the choice of confrontation is calculated, it is still reasonable to characterize weaker state initiation as a risky strategy. In the lead up to conflict, decision makers are often uncertain about rival resolve (Fearon 1995; Sechser 2010). Making judgments with imperfect information about resolve can be costly when confronting an inferior opponent, but it is potentially fatal when facing a superior rival willing to push a conflict to conclusion. In addition, research suggests that surprise attacks intended to capitalize on short-term advantages rarely yield wholly favorable outcomes and often backfire even when the balance of power favors the initiator (Helfstein 2012). Moreover, weak state strategy often has little to do with the outcome of a dispute. Even when they are successful, weaker states do not prevail because they outsmart their rivals. Instead, “Strong actors lose asymmetric conflict when they adopt the wrong strategy vis-à-vis their weaker adversaries” (Arrefuin-Toft 2005, 121). So, as Van Evera (2013) notes, while fait accompli may promise a “greater chance of political victory than quiet consultation,” it also increases the probability of open conflict because when “opponents stand firm, a collision is hard to avoid” (131–32). Although this is true for all governments, the risks in collision are greater for weaker initiators. Consider an iconic example of fate accompli, the Japanese Pearl Harbor attack against the United States. Feeling trapped by an eroding strategic position, Japanese policymakers hoped that the United States, confronting the specter of a costly European conflict, would acquiesce once Japan established military superiority in the region (Conroy and Wray 1990). Consistent with fait accompli, Japanese leaders correctly understood that there was an advantage in acting first, decisively, and with crippling force. However, while “none of the leaders were confident of Japan’s chances, especially in a lengthy war,” they nonetheless launched into the conflict with “a fatalistic attitude and only a vague hope that the United States would grow weary of the struggle after early Japanese successes and negotiate an agreement recognizing Japan’s territorial gains in East Asia” (Levi and Whyte 1997, 808). As the example suggests, the logic of fait accompli can effectively explain why a limited-aims strategy is sometimes more attractive than a total war strategy. However, the decision to confront powerful rivals in this way is still a risky endeavor, and, in the Japanese case at least, decision makers were explicitly aware of this fact. The central issue motivating this study thus remains. Why do states adopt risky strategies of confrontation? The cognitive version of NCR presented here attempts to answer this question by designing a framework to explicitly identify the combination of international and domestic conditions under which risk acceptant foreign policy is most likely to occur. A Cognitive Neoclassical Approach At its core, NCR is focused on explaining how states respond to security challenges. NCR scholars broadly accept realist assertions about the primacy of security politics and argue that shifts in the security environment initiate changes in foreign policy (Toje and Kunz 2012). As Rose (1998) puts it, security threats shape “the broad contours and general direction” of foreign policy behavior (147). Changes in the distribution of power are particularly salient (Kitchen 2010), though research points to a broad range of triggering mechanisms ranging from shifting alliance commitments (Cha 2000) to concerns over burden sharing in climate change negotiations (Purdon 2013). NCR departs from traditional realist analysis in two ways. First, NCR depicts state institutions as intervening variables linking external security imperatives to a country’s observed foreign policy. There is no perfect transmission belt that directly connects a state’s capabilities to its foreign policy needs (Kunz and Saltzman 2012). For example, policymakers are often constrained in decentralized states because power is diffused across institutions and domestic political interests are sharply divided. This can both prevent governments from expanding their global influence as national power grows (Zakaria 1999) and also renders the state unable to balance against emergent threats in a timely manner (Schweller 2008). When threatened, constrained governments are also least likely to respond with innovative foreign policies or emulate the successful strategies of others (Taliaferro 2006). Second, foreign policy decision makers are independent actors in NCR analysis.3 While elites can pressure decision makers to define security threats in terms of their own parochial interests, decision makers also act autonomously (Lobell 2009). For example, decision makers sometimes manipulate societal discourses about conflict in order to mobilize domestic opinion and coordinate political actors, thereby (partially) overcoming institutional constraints (Schweller 2009). Decision makers often attach meaning to security threats consistent with their preexisting beliefs, and then align foreign policy responses with internal constraints (Edelstein 2002). This can influence their willingness to confront rivals and contour the form that such policies might take (Dueck 2009). By this view, foreign policy decision makers are not simply rational maximizers pursuing objective state interests, but are instead political entrepreneurs with their own biases, beliefs, and cognitive deficiencies. There remains considerable debate over the degree to which NCR exists as part of a progressive realist program (Rathbun 2008; Quinn 2013). Nonetheless, NCR is now a proven foreign policy framework explaining when and how governments respond to external security challenges. Below I offer a theory of weaker state confrontation that adopts the three core assumptions in NCR analysis: salient shifts in the external security environment are the primary motivation for changes in foreign policy, domestic political institutions condition state responses to such shifts, and decision maker perceptions are an independent factor shaping foreign policy outcomes. Because weaker state conflict initiation is a risky strategy, my argument focuses on the role of decision maker frames and utilizes prospect theory to explain how an erosion in the external security environment increases the likelihood of risk acceptant foreign policies. As I discuss below, prospect theory lacks a “theory of framing” for simple decisions, let alone for the more complicated environments that often define foreign policy (Boettcher 2004). This means that prospect theory itself offers little guidance on how to identify the conditions under which decision makers are most likely to settle into a dominant frame, nor can it alone explain how intervening domestic political institutions might influence the manner in which decision makers react to a decline in security. Stated differently, because prospect theory is a model of human choice and not a theory of foreign policy, it lacks a theoretical story about the transmission belt by which changes in the security environment are processed through domestic political interests and the degree to which this shapes decision frames. Such a deductive argument is required, however, in order to construct hypotheses about foreign policy that are amenable to statistical examination. Indeed, the prospect theory literature in foreign policy is dominated by thick descriptions about the emergence of frames that, while insightful, have not accumulated into a set of broadly accepted framing principles. Previous research has identified a number of contingent factors that lead to the onset of loss frames including external shocks, shifting domestic political incentives, and/or intra-elite deliberation.4 By contrast, the discussion below demonstrates that combining NCR with prospect theory produces generalizable theoretical claims appropriate to statistical analysis. Prospect Theory, Political Competition, and Counterframes To summarize, my argument is that decision makers initially evaluate changes in the current security environment against their previous security position. When confronted with an erosion in security, policymakers confront two options. They can accept the certain losses that come with an eroding security environment by remaining in the status quo, or they can opt for a revisionist strategy of confrontation. Confrontation is a gamble because it holds the potential to improve a state’s security position, but it also contains the possibility of further losses beyond the status quo. According to prospect theory, this kind of choice framing produces risk acceptance. I argue that this is most likely to occur in regimes where policymakers operate in an environment with low levels of political counterframing. In such regimes, there are fewer relevant interests lobbying on foreign policy outcomes and fewer competing political frames defining the costs and benefits attached to alternative strategies. This institutional context is more conducive to the emergence of a single dominant frame that defines a country’s options. The result is that regimes in an eroding strategic position are most likely to take inordinate risks in their foreign policy when counterframing is absent.5 These risks include initiating disputes with more powerful rivals. Loss Frames and Risk According to prospect theory, decision makers become risk averse when confronted with choices over gains and risk acceptant when confronted with choices over losses (Kahneman and Tversky 1979a; Thaler 1980). An individual’s risk disposition is determined by “whether the outcomes are perceived as gains or losses” relative to a neutral reference point (Quattrone and Tversky 1988). Research suggests that the cognitive mechanisms underlying prospect theory emerged in response to environmental pressures in the early human environment and are thus part of a shared evolutionary legacy (McDermott et al. 2008). Indeed, more recent advances in cognitive neuroscience confirm that shifts in risk disposition are anchored to structures in the human brain (Kuhnen and Knutson 2005; Trepel et al. 2005; Tom et al. 2007) and that framing is attached to basic emotional processing (Ma et al. 2012). For foreign policy scholars, prospect theory offered a new framework to understand risk acceptant decisions that were inconsistent with expected utility explanations (Levy 1997). There is now strong evidence to suggest that when policymakers frame their options in terms of losses, the propensity to adopt risk acceptant foreign policy strategies increases significantly (McDermott 2004). This insight extends to militarized conflict (Mcdermott 1992; Taliaferro 1998), grand strategy (He and Feng 2013), civil conflict resolution (Hancock et al. 2010), and military deterrence (Berejikian 2002). The consistent finding here is that when the status quo ante becomes unacceptable the decision frame shifts to losses, and strategies that were once too risky to contemplate become more acceptable to decision makers.6 Prospect theory was initially tested using decision experiments. Laboratory subjects were confronted with two clear alternatives to the status quo—a certain outcome and a gamble—and this defined the decision frame.7 However, decision makers also assess current conditions against prior circumstances, treating the recent past as a reference point for evaluating the desirability of the status quo (Tversky and Kahneman 1974; Wright and Anderson 1989). This is particularly true when actors experience a setback because the emotional impact of losses is greater than gains and individuals are therefore slow to accept their new circumstances. As Jervis (1992: 200) notes, In international politics and in social life in general… we renormalize for gains much more quickly than we do for losses. We rapidly, if not effortlessly, adjust to good fortune and any improvement in our lives…. Neither individuals nor nations are so accepting of losses, however. We remain unhappy, unreconciled, and often bitter for a prolonged period.When states suffer an erosion in security, decision makers are slow to update to the new status quo and often tenaciously cling to their previous reference point (Levy 1992). Prospect theory’s concept of a decision frame therefore extends to instances in which actors confront a choice between accepting recent changes to the status quo and a gamble (Berejikian 1997). Consistent with NCR commitments about the primacy of security politics, I assume that changes in the external security environment define the decision frame through which policymakers assess their state’s security position.8 Foreign policy actors in states that experience erosion in their security environment are therefore confronted with a loss frame as defined by prospect theory. They can do nothing and accept the known negative consequences inherent in the revised status quo, or they can opt for a gamble in an attempt to return to their previous position. Such decision makers are operating in the domain of losses and are, according to prospect theory, more likely to adopt risk acceptant strategies.9 Political Competition and the Onset of Loss Frames Political competition reduces the likelihood that any single frame will settle over a policy space.10 To the extent that the consequences of foreign policy affect domestic groups differently, the larger the number of groups engaged in competition over policy outcomes, the more likely it is that leaders will be forced to incorporate additional domestic political gains and losses when considering foreign policy strategies. Experimental evidence provides microfoundational support for this contention. Simply offering decision makers alternative frames can dislodge a settled frame (Druckman 2004; Boettcher and Cobb 2009; Kam and Simas 2010). Research also shows that when individuals are asked to reconsider their choices, they first reprocess the incentive structure. The cognitive demands of reappraisal reduce the emotional impact of negative frames, diminishing the influence of framing effects (Miu and Crişan 2011). As frame salience diminishes, decision processing shifts from emotional centers of the brain to structures associated with higher-order cognitive functioning (Benedetto et al. 2006), and the impact of decision bias is reduced. Domestic counterframing is a routine aspect of foreign policy. We know that political actors representing entrenched interests define foreign policy issues in self-serving ways (Allison and Zelikow 1999), and this includes explicit attempts to reframe foreign policy initiatives (Garrison 2001). As political competition broadens in a society, decision makers must pay closer attention to an increasingly varied set of policy demands in order to build and maintain successful governing coalitions. As groups compete, they seek to undermine evaluative frames that are not consistent with their parochial interests. Such self-interested framing is understood to be a central component of effective political manipulation in foreign policy (Maoz 1990; Levy 2000). Further, the ability of political rivals to exploit ineffective policies for partisan gain disciplines executives because they must be concerned that failed policies will underpin a shift in support to competitors (Bueno De Mesquita 1985; Fearon 1994: Bueno de Mesquita et al. 2003). Domestic political competition can therefore transform the content of a foreign policy frame. For example, strategies that might look to be a gamble in the absence of counterframing can take on the characteristics of a certain loss when counterframing is present. In the analysis below, Syria is identified as a state that slipped into the domain of losses (1982) and that lacked meaningful counterframing institutions. Bogged down by its military presence in Lebanon and increasingly marginalized buy its allies, the Syrian government was looking for a way to regain its foreign policy momentum and initiated two militarized interstate disputes (MIDs) against more powerful adversaries. One of these was with its long-time rival Iraq. The Assad regime seized Iraqi oil, embargoed Iraqi oil exports through Syria, provided military support to Iran, and publically called for the overthrow of the Iraqi regime. Confronting Iraq was a calculated risk. On the one hand, political and military assistance to Iran held the possibility of tipping the scales against Syria’s long time foe in the stalemated Iran–Iraq war. On the other hand, Syria risked further political and diplomatic isolation from traditional allies for siding against an Arab country. Syria desperately wanted continued Arab political support in the looming confrontation with Israel, and even more so it required ongoing financial assistance to subsidize its expensive military presence in Lebanon. The Assad regime ruled through military force and drew political support from a narrow slice of Syrian society, the Alawites, while the much larger Sunni population was ruthlessly repressed. As a result, there existed no potential for counterframing in Syria when Assad considered the decision to side with Iran and against an Arab neighbor. This becomes clear if we consider the alternative as a hypothetical—one in which the Sunnis had meaningful access to political power and could impose the kinds of audience costs described earlier. Added to the expected regional Arab backlash would be the negative domestic political consequences in the form of bitter Sunni opposition. The decision to support Iran would then look less like a calculated risk and more like a doomed political strategy. Assad’s options would no longer resemble a standard prospect theory decision frame. Instead of a choice between a certain loss (status quo) and a foreign policy gambit (confront Iraq), effective Sunni counterframing transforms the gamble leaving Syria two alternatives that both offer only losses. The logic of counterframing articulated here does not predict which specific domestic interests ultimately prevail. Nor is the argument that counterframing will always dislodge an already entrenched frame. Instead, changes in external security forces leaders to reevaluate their policy options, and part of that evaluation involves a consideration of domestic political consequences and constraints. Because robust political competition attaches varied and politically salient costs and/or benefits to available strategies, it increases the number of trade-offs that policymakers must confront. This can prevent the emergence of a single dominant frame. Without counterframing, leaders like Assad are free to take their cues primarily from the external security environment, increasing the likelihood that an erosion in external security produces a clear loss frame as defined by prospect theory. Hypotheses and Testing I have argued that, in the absence of counterframing institutions, erosion in the external security environment generates loss frames and increases the probability of risk acceptant foreign policy decisions. This produces the main hypothesis examined in this paper. Hypothesis 1: The onset of a loss frame increases the probability of weaker state dispute initiation in regimes that lack counterframing institutions.To evaluate this hypothesis I examine a subset of MIDs taken from the Correlates of War (COW) data set. The dependent variable is coded “1” for disputes where the initiator actions ranged from direct displays of force to war against a target in the observation year and “0” otherwise.11 The sample includes all such cases for which data are available between 1946 and 1999. Only politically relevant dyads in which states can engage one another directly are included. Decisions to join ongoing disputes are not included. Initiators are defined as those states involved in a dispute when it begins and that are on the side taking the first action. Targets are defined as those states that are originally involved in a dispute at its inception on the side that does not take the first action. Because I am interested in weaker state initiation, the unit of analysis is the directed dyad year for all pairs in which potential initiators have a lower score on the Composite Index of National Capability (CINC) than potential targets (Singer 1988). The main independent variable consists of two components. The first is the decision maker security frame and the second is an absence of counterframing institutions. Decision Maker Security Frame NCR assumes that managing shifts in the external security environment is a primary foreign policy objective, and that policymakers look to “other state’s intentions” and “changes in the relative distributions of power” as salient indicators of their security environment (Kunz and Saltzman 2012, 102). Therefore, a decision frame variable must first capture the assumption that decision makers anchor their evaluation of the status quo to changes in these dimensions. Second, to make meaningful comparisons across years, a framing variable must also provide a standardized indicator that tracks changes in a state’s security environment. Because the security potential of states is not static, this is a challenge both across states and within states over time. For example, a country can grow to become more powerful relative to its rival and nonetheless find itself in a more precarious security environment if alliances shift against it. Third, statistical testing hinges upon a key assumption about decision maker perception. If we assume that decision makers perceive real changes in their environment, then objective shifts in security define the content of the decision frame. In the laboratory, researchers assume that prospect theory subjects all understand the objective choices confronting them and that participants adopt the decision frames as constructed by investigators. This assumption is also common in rational choice, which holds that decision makers confront incentive structures observable by both participants and investigators. Finally, the discipline of behavioral economics has used objective measures to test prospect theory’s claims extensively (Barberis 2012).12 To operationalize the decision frame, I begin with Bueno De Mesquita's (1985) metric to identify a state’s security position. I construct a security continuum ranging from best to worst possible security environments for any state i by summing the utilities for conflict by all states j versus i.13 This continuum is bracketed at one end by the hypothetical interstate alliance pattern and relative power configuration that would leave i most vulnerable to defeat and at that other, end by the pattern that leaves i most secure. I define the endpoints on this continuum as the sum of expected utilities for j versus i as ∑E(Uji)max and ∑E(Uji)min, respectively. State i's observed security, ∑E(Uji), moves between the extremes of this continuum as alliance patterns and relative power shift over time.14 By this conceptualization, i is least secure when the summed expected utilities for conflict are at their maximum and i is most secure when they are at their minimum.15 I next calculate the previous year’s observed security score and measure the distance from this to the midpoint on the continuum for that year, and then take this same distance using the observed security score for the current year. When the difference between these two distances is positive, this indicates the summed regional utilities for conflict are increasing and therefore that the state is declining in security year to year. I assume that states experiencing a security decline operate in the domain of losses and are coded as lossframe =1. The result is a yearly dichotomous indicator that identifies the onset of a losses frame. The primary benefit of this conceptualization is that it provides an updating indicator for loss frame that compares a country’s current security performance to the previous year. Take, for example, the decline Syria experienced in 1982, a year in which it initiated two MIDs against adversaries that were more powerful. To construct the indicator for 1982, I first take the values for security in 1981. Syria’s security continuum ranged from 78.60 (worst security) to − 65.30 (best security), with a midpoint at 6.65. Syria’s observed security score in 1981 was 31.74 with a distance from the midpoint of 25.09. In 1982, Syria’s observed security score was 37.06, and the distance from the 1981 midpoint increased to 30.41. This indicates a 21.19 percent shift in the direction of decreased security. As a result, I code Syria as lossframe = 1 for the year 1982. Counterframing To account for the absence of counterframing institutions, I initially use two Polity IV concept variables to construct a dichotomous indicator identifying the set of states with both very low levels of political competition between domestic groups and where such groups have little ability to impose their preferences on the executive (Marshall and Jaggers 2002).16 For this set of states, there will be little counterframing, and, to the extent it exists, domestic groups have little ability to assert their frames on the executive. The variable counterframe is coded as “1” for countries where the Polity IV score for political competition is either “repressed” or “severely restricted” (represented by a “political competition” score of 1 or 2) and where executives operate freely from domestic group preferences (represented by an “executive constraint” score of 1 or 2). Otherwise, countries are given a score of “0.” This identifies a broad category of states with low levels of political counterframing. However, it does not capture fine-grained distinctions. For example, research suggests that autocratic regimes vary in the level of domestic audience costs that can hold leaders to account on foreign policy outcomes (Peceny and Beer 2003; Lai Slater 2006). Of particular interest here are so-called personalist regimes characterized by the highest levels of centralized control in the hands of an individual. This can include direct authority over the military, the governing bureaucracy, and/or the ruling party apparatus. In such states, the status of elites is subject to the whims of a leader who can remove dissenters at will. The defining characteristic of personalist regimes is that, while such regimes “are often supported by parties and militaries, these organizations have not become sufficiently developed or autonomous to prevent the leader from taking personal control of policy decisions” and are less bound by the preferences of domestic actors (Geddes 2003, 53). This is in contrast to authoritarian systems with more constrained forms of autocratic rule in which elites have access to an independent political base. Personalist regimes therefore diverge from other forms of authoritarian rule in that “access to office and the fruits of office depend much more on the discretion of the individual leader” (Geddes 2003, 51). Control over the “fruits of office” is key to my argument about political competition and counterframing. The implication here is that some leaders in nondemocratic regimes are uniquely susceptible to loss frames because the institutional infrastructure required for even minimal levels of counterframing is absent. To explore this possibility, I include in my analysis four categories of authoritarian regimes constructed by Weeks (2012). I examine the impact of loss frames on two categories of elite-constrained regimes. Political “machines” are characterized by civilian elite audiences often embedded in the dominant political party, while “juntas” are typified by military institutions that serve as an independent source of political authority within government. I also examine two categories of personalist regimes, political “boss” and military “strongman.” Here individuals enjoy their status as elites largely at the behest of the leader who controls access to the benefits of government policies. I expect that loss frames increase the probability of conflict in personalist regimes because audience costs are almost entirely absent. This produces a second hypothesis examined here. Hypothesis 2: The probability of initiation will increase with the onset of loss frames in personalist regimes, compared to all other government forms. Control Variables I also include a suite of control variables common in dyadic conflict analysis to account for alternative factors that might influence the probability that weaker states will initiate. Because trading relationships are known to affect the propensity for conflict, I include a standard measure for initiator’s trade dependence, defined as the total trade between initiator and target divided by initiator’s GDP (Oneal and Russett 1999). Security relationships can also shape foreign policy preferences. For example, similar alliance portfolios between countries indicate shared security interests, reducing the potential for conflict. In addition, weaker states that share security interests with powerful states may feel emboldened, whether they are pursuing a strategy of fait accompli or outright victory, if they anticipate political support from their more powerful allies. To control for these effects, I use Signorino and Ritter's (1999) s-score for alliance similarity between target and initiator and between initiator and the global system leader, calculated for the initiator’s home region. To account for the known impact of contiguity, I include a binary indicator for shared borders taken from Stinnett et al. (2002). Regime instability has also been identified as a source of aggressive behavior (Sirin 2011), so I include an indicator for the existence of civil war in the initiator (Gleditsch et al. 2008) and a simple dummy measure for regime durability.17 Relative capabilities are measured by the ratio of initiator over target CINC score. I expect that the probability of initiation will increase as this ratio approaches a value of one (power parity). I also include a number of controls that define the contextual relationship between initiator and target. First, Goertz and Diehl (1993) argued that some states are locked into a relationship of rivalry, so the pattern of conflict may be “influenced by the outcome or processes of previous disputes or by the prospect of future disputes between the same states” (148). Because managing rivalries is among the most salient foreign policy objectives, weaker states in rivalries with more powerful adversaries may initiate conflicts because of dynamics internal to the rivalry rather than because of an increased tolerance for risk. To account for this, I include a variable indicating the presence of an enduring rivalry using the operationalization provided by Klein, Goertz, and Diehl (2006). I also include a dichotomous variable indicating the existence of a major power dyad taken from the COW data set. Second, research suggests that the presence of nuclear weapons influences the propensity for conflict, although there remains considerable debate about the nature of this effect (for example, Horowitz 2009; Sobek, Foster, and Robinson 2012; and Miller 2013).18 To control for the impact of nuclear weapons, I include an indicator for the acquisition of nuclear weapons by the target taken from Singh and Way (2004). Third, to account for the role of joint democracy on conflict, I include an indicator for joint democracy where both initiator and target have a minimum score of 6 using the Polity IV democracy scale (Marshall and Jaggers 2002). Fourth, states with low levels of counterframing also tend to score low for democracy, and there is some evidence of an autocratic peace (Peceny et al. 2002). To account for this, I include a dummy variable indicator for autocratic states coded as “1” if the Polity IV regime score for the target is less than −5 and “0” otherwise. Results and Analysis I have argued that, in the absence of counterframing institutions, an eroding strategic position creates a loss frame and produces risk acceptant foreign policy behavior. In order to test this hypothesis, I examine a data set containing politically relevant directed dyads for the period 1946–1999, where the CINC score for the initiator is lower than the target. This produces 55,700 observations and 602 MIDs (1.08 percent). The average power ratio (initiator/target) for weaker state initiation is 0.340, with a range from 0.0003 to 0.9926. I also constructed four binary indicators for each combination of lossframe x couterframe, with the reference category set to lossframe = 1 and counterframe = 0. Of particular theoretical interest is the difference between the probability of weaker state initiation in this reference category and the category lossframe = 1 and counterframe = 1. This comparison permits a direct test of the hypothesis that the onset of loss frames increases the probability of conflict where counterframing is absent. As dispute initiation is coded dichotomously, I deploy binary logistic regression with robust standard errors clustered on the dyad. To account for duration dependence, I use the method prescribed by Beck et al. (1998) of including a count variable for the number of within-dyad peace years and three cubic splines. Table 1 summarizes the regression results using the Polity IV operationalization for counterframe, with coefficients reported as odds ratios.19 Model 1 contains only the NCR variables that are of primary theoretical interest. This includes three categories of lossframe x counterframe. Model 2 introduces the suite of controls. Model 3 contains only those variables of statistical significance in Model 2. The results across Models 1–3 provide strong support for the argument presented here. The onset of loss frames combined with a lack of counterframing institutions (lossframe = 1 and counterframe = 1) exerts a consistently positive and statistically significant effect on the likelihood of weaker state conflict initiation, compared to the reference category (lossframe = 1 and counterframe = 0). In addition, several of the controls commonly associated with conflict initiation are also significant. In both Models 2 and 3, the likelihood of initiation increases as the dyad approaches power parity, when the initiator is embroiled in civil war, when the target has acquired nuclear weapons, when the two states share a border, and in the presence of an enduring rivalry. The likelihood of initiation decreases for democratic dyads and dyads with shared alliance patterns. Surprisingly, increased trade dependence is associated with an increase in initiation.20 Table 1. Weak state MID initiation Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Notes. Reference category: Lossframe = 1, Counterframe = 0. *p ≤ .05, **p ≤ .01. Table 1. Weak state MID initiation Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Logistic regressions   Coefficients as odds ratios   Model 1  Model 2  Model 3 (robust standard error)  Lossframe  Counterframe        0  0  0.872  0.793*  0.799  (0.995)  (0.096)  (0.096)  0  1  1.184  0.823  0.844  (0.145)  (0.111)  (0.111)  1  1  1.857**  1.382**  1.411**  (0.206)  (0.167)  (0.168)  Civil war      1.667**  1.163**      (0.330)  (0.321)  Durability      0.810        (0.096)    Major dyad      1.405        (0.347)    Trade dependence      5.218*  5.274*      (4.230)  (4.188)  Target nukes      1.279*  1.315*      (0.162)  (0.161)  S-score      0.475**  0.432**      (0.066)  (0.056)  S-score with system leader      0.812        (0.198)    Power ratio      2.295**  2.524**      (0.453)  (0.476)  Contiguity      3.835**  3.964**      (0.474)  (0.486)  Democratic dyad      0.427**  0.417**      (0.078)  (0.075)  Enduring rivalry      4.264**  4.413**      (0.592)  (0.602)  Constant    0.182  0.076  0.067    (0.021)  (0.015)  (0.012)  Observations    46,422  44,077  44,077  Notes. Reference category: Lossframe = 1, Counterframe = 0. *p ≤ .05, **p ≤ .01. Table 2 reports results for models using each of the four categories of authoritarian regimes identified above and includes the controls found to be statistically significant predictors in Table 1. Here, counterframe is coded as “1” when the regime takes the characteristics of each autocratic type and “0” otherwise. The results are interesting and suggest that there are important counterframing differences within the autocratic camp. For both categories of personalist regime, boss and strongman, the onset of a loss frame increases the probability of weaker state initiation. Consistent with expectations, personalist regimes appear to lack counterframing institutions and are therefore more susceptible to risk acceptance as their security erodes. The findings for elite constrained regimes are varied. Under loss frames, civilian machines appear to behave no differently than other regimes. This suggests the presence of frame competition by entrenched elites within the society who are also able to assert their interests to their government. By contrast, the coefficient for juntas is positive and significant. While juntas are also elite constrained, this positive association between the onset of loss frames and risk acceptant policy suggests that a single coherent security frame is more likely when government is dominated by military institutions. This may be due to the fact that, compared to their civilian counterparts, elites in juntas are more likely to share unified “beliefs about the utility and appropriateness of military force as an instrument of politics” and that they are “more likely than civilians to form ominous views of the status quo” (Weeks 2012, 333). This suggests that, rather than a lack of counterframing institutions, the narrow composition of elites in military juntas explains the onset of loss frames that produce risk acceptant foreign policy choices. Table 2. Initiation by autocratic type Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  *p ≤ .05, **p ≤ .01. Table 2. Initiation by autocratic type Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  Logistic regressions   Coefficients as odds ratios    Model 1  Model 2  Model 3  Model 4  (Robust standard error)    Boss  Junta  Strongman  Machine  Lossframe  Counterframe          0  0  0.777**  0.738**  0.702**  0.681**  (0.076)  (0.068)  (0.067)  (0.064)  0  1  0.832  0.834  1.123  0.747  (0.150)  (0.210)  (0.196)  (0.154)  1  1  1.683**  1.790**  1.476**  0.689  (0.240)  (0.378)  (0.241)  (0.136)  Civil war    1.567*  1.583*  1.493*  1.570*    (0.305)  (0.308)  (0.294)  (0.306)  Trade dependence    1.014  1.016  1.016*  1.017*    (0.008)  (0.008)  (0.008)  (0.008)  Target nukes    1.407**  1.418**  1.413**  1.400**  (0.164)  (0.166)  (0.166)  (0.164)  S-score    0.419**  0.399**  0.393**  0.399**  (0.051)  (0.049)  (0.490)  (0.050)  Power ratio    2.513**  2.555**  2.515**  2.622**    (0.453)  (0.460)  (0.455)  (0.475)  Contiguity    4.506**  4.579**  4.476**  4.470**    (0.414)  (0.512)  (0.531)  (0.528)  Democratic dyad    0.414**  0.401**  0.413**  0.379**    (0.072)  (0.069)  (0.071)  (0.065)  Enduring rivalry    4.154**  4.125**  4.059**  4.109**    (0.560)  (0.551)  (0.547)  (0.556)  Constant    0.070  0.076  0.077  0.083    (0.012)  (0.013)  (0.132)  (0.014)  Observations    44,592  44,592  44,592  44,592  *p ≤ .05, **p ≤ .01. Logistic regression coefficients cannot be interpreted directly as their effect is curvilinear and dependent upon specific values of the other covariates in the model. In addition, because conflict is a rare event, even relatively powerful predictors will produce only a small change in the overall probability of weaker state initiation. Table 3 provides a more intuitive way to understand the substantive impact of the coefficients by reporting the percentage change in the predicted probability of initiation when lossframe changes from “0” to “1.” Five country profiles are summarized in Table 3. Each profile reports the change in the predicted probability for a hypothetical “typical” country with continuous covariates held constant at their median values and dichotomous variables at their modal scores. Table 3 also includes values for the Syrian case described above. The onset of a loss frame for a typical country using the Polity IV indicator for counterframing increases the likelihood of risky initiation by 69.2 percent. For juntas and bosses, the effect is even larger, increasing the predicted probability by 115.4 percent and 104 percent, respectively. For strongman governments, the increase is 29.4 percent. Consistent with the fact that it was classified as a strongman regime in 1982, the onset of a loss frame for Syria increased the risk of initiating a dispute with a more powerful adversary by 27 percent. Table 3. Percentage change in predicted probability Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  p ≤ .05. *The coefficient for machine was not significant in model 4. Table 3. Percentage change in predicted probability Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  Change in loss frame (0/1)   Country profile  Difference  Polity Lv  +69.2  Machine  −8.4*  Junta  +115.4  Boss  +104.0  Strongman  +29.4  Syria 1982  +27.0  p ≤ .05. *The coefficient for machine was not significant in model 4. Although these results support NCR claims concerning the impact of loss frames, it is conceivable that a subset of dyads drive these findings. As noted above, the range of capability ratios across dyads experiencing MIDs is quite large extending from 0.0003 to 0.9926 with a mean of 0.340. Unfortunately, prospect theory offers little guidance about the magnitude of risk that foreign policymakers will accept when operating in a losses domain. Loss frames may matter less in dyads where there are large disparities in power because the consequences of initiation are more transparent, so weaker states understand that they are unlikely to prevail whatever their tolerance for risk.21 A second possibility is that the importance of loss frames diminishes as initiators approach parity with targets because they are in a stronger position to address their grievances without having to resort to conflict. The third option is that loss frames produce risk acceptance across the entire range of dyadic relative power. If, as noted above, prospect theory preferences constitute an evolutionary response to selective pressures, then we would expect them to be robust even when survival of the individual, or the state, is at stake. Using results from Model 3, Figure 1 displays the impact of loss frames across the entire range of relative power distributions for states without counterframing institutions. The results suggest that relative power mitigates the impact of loss frames only when dyads approach power parity. The upward slope in Figure 1 indicates that the predicted probability for conflict increases as the initiator approaches parity with the target, whether or not states are in a loss frame. However, the effect of lossframe is also distinct and statistically significant from the null throughout a large portion of the range of dyadic power ratios. The confidence intervals overlap when the initiator reaches 70–75 percent of the target’s power. That is, loss frames matter across the majority of relative power distributions. Importantly, this range includes the majority (85.9 percent) of weaker state initiations. Figure 1. View largeDownload slide Lossframes and dyadic relative power. Figure 1. View largeDownload slide Lossframes and dyadic relative power. In sum, these results provide support for the assertion that negative shifts in the security environment set the stage for changes in foreign policy by framing the status quo ante as a loss and increasing the attractiveness of risky strategies. The emergence of a dominant frame is contingent upon the level of political counterframing in a society. Meaningful political competition between domestic groups appears to partially inoculate decision makers against the onset of a clear decision frame. Conclusions The cognitive version of NCR presented here modifies traditional NCR analysis in several ways. First, it incorporates prospect theory to explain how changes in external security can elicit risky foreign policy choices. While the focus here is on conflict, the combination of prospect theory and NCR can be used to develop testable arguments about risk across the entire range of foreign policy domains. Second, my argument introduces variation in the level of political competition as a key domestic variable shaping elite perceptions. Competition varies across regimes and determines when dominant frames are most likely to settle over a policy issue. In this way, counterframing helps define both the meaning of changes in the external security environment and the relative attractiveness of available strategies. Third, a cognitive version of NCR enables the construction of hypotheses about foreign policy behavior that are amenable to statistical scrutiny. The first wave of NCR analysis provided mostly “theoretically informed narratives” rather than deductive theory, and the primary focus of empirical work was restricted to “counterfactual analysis” instead of identifying broad patterns of foreign policy behavior (Rose 1988, 153). However, the lingering view that the NCR enterprise is inherently restricted to case studies is incorrect.22 I do not claim that statistical testing is a superior method. Statistical testing and case analysis are complementary, and both are necessary in order to push NCR analysis into new empirical domains as part of a progressive research agenda. The results presented here also have implications for the foreign policy literature anchored to prospect theory. To the extent that the roil of domestic political competition requires policymakers to confront issues from a number of competing perspectives, counterframing appears to act as an imperfect buffer against the kind of risk acceptant foreign policy behavior identified in previous studies. This finding holds even for interstate rivalries, where a cycle of conflict dominates the relationship. A lack of domestic competition is therefore a potential liability for states. Robust counterframing appears to partially inoculate governments against unwise risks and in this way supports more reasoned foreign policy judgments. These results also suggest that risky foreign policy decisions are better explained by an NCR framework that replaces traditional rational actor assumptions with prospect theory. The notion that factors external to the state define decision maker frames is consistent with the traditional NCR assumption that shifts in the international security environment provide the initial motivation for changes in foreign policy. However, the idea of domestic loss frames is also worth investigating. For example, there is considerable debate about the degree to which domestic politics alone can produce incentives that lead to a strategy of external conflict (for example, Lai and Slater 2006). Factors like civil unrest, eroding political support, and/or leadership challenges may themselves motivate risk acceptant foreign policies. Indeed, domestic instability combined with an erosion in external security would potentially generate very salient loss frames. As Levy (1992) notes, “The combination of perceived external decline and internal insecurity may be particularly conducive to risk seeking” (287). Expanding the definition of decision maker frames to include both international and domestic factors is an important next step. Finally, perhaps the most exciting implication is that cognitive NCR should now be particularly attractive to interdisciplinary foreign policy scholars. By substituting cognitive principles in place of rationalist assumptions, NCR provides a kind of “plug-and-play” structure permitting the integration of decision science into foreign policy analysis. There is no reason to construct a new foreign policy framework for each distinct cognitive heuristic or bias. The role of loss aversion, fairness, social trust, etc. on foreign policy can now be examined under a single conceptual umbrella. Dr. Berejikian is an Associate Professor in the Department of International Affairs, a Josiah Meigs Distinguished Teaching Professor at the University of Georgia, and a Senior Fellow at the Center for International Trade and Security. 1While the conceptualization of risk in foreign policy studies continues to evolve (Clapton 2011), I adopt the conventional notion of comparative risks defined as the level of variance in outcomes attached to available strategies. A strategy is considered risky if it presents a larger variance in payoffs than an alternative. As I describe below, for states operating under an eroding security environment, weaker state confrontation is risky in that confronting a more powerful rival produces an outcome that is much better (victory) or much worse (defeat) than the status quo ante. 2See also George and Smoke (1974) who argue that, in the context of deterrence politics, fait accompli is sometimes an effective strategy. 3NCR scholarship often uses the terms elites, decision makers, and decision executives interchangeably. The purpose is to denote the relevant individuals with decision authority over a policy space. Here, I use the term decision makers in a way that is consistent with Putnam’s (1988) identification of “chief of government” and/or Lobell’s (2009) more recent “foreign policy executive.” 4Levi and Whyte (1997) identify elite deliberation as a source of loss frames in the above-mentioned Pearl Harbor case. 5I define counterframing as the attempt by actors to assert alternative decision frames—through the manipulation of reference points—consistent with their interests. In the foreign policy domain, counterframing defines the competition between domestic groups to frame policies as beneficial (gains) or detrimental (losses). 6For a comprehensive review of prospect theory research in the field of international relations, see Berejikian (forthcoming). 7For example, a typical loss frame would be one in which subjects confronted a choice between a certain loss of $80 and a gamble with an 85 percent chance losing $100 and a 15 percent of losing nothing. 8The idea that states construct evaluative frames for security has deep roots in the realist study of international politics. Balance of power (Waltz 1979), power transition theory (Organski 1968), and hegemonic conflict (Gilpin 1983) all incorporate a version of this idea. The shared insight is that governments are sensitive to changes in their security position relative to potential rivals and that it is through this lens that they assess the attractiveness of their policy options. Power transition theory, in particular, emphasizes the importance of status quo evaluations on foreign policy choice (Lemke and Reed 1996). While research has tended to focus on great powers, there is evidence that lesser states make similar assessments on a regional basis (Lemke 2002). 9This is consistent with the notion of evaluative framing. Evaluative frames set the reference point against which changes in the external environment are compared. For a conceptual treatment of the various ways in which the term “framing” is used by international relations and foreign policy scholars, see Mintz and Redd (2003). There are also additional conceptualizations in political science more generally (Chong and Druckman 2007). 10In the field of behavioral economics, the impact of counterframing competition on decision bias is already well established. Market competition forces decision makers to consider multiple strategies simultaneously because it disciplines individuals who make nonmaximizing choices. The frequency of bias over time thus diminishes as markets become more competitive (Bazerman et al. 1985; Russell and Thaler 1985). While the debate in behavioral economics continues over the exact conditions under which competition most efficiently disciplines actors for making nonmaximizing choices (for example, Rubinstein 2001), there is general agreement that competition mitigates the consequences of decision bias. Much of behavioral economics now focuses on this interaction between social institutions, competition, and the frequency with which decision heuristics take hold. 11The data were generated using the Eugene software program version 3.204 (Bennett and Stam 2000). 12The opposing view is that decision makers possess unique worldviews and, therefore, subjectivity affects decision maker perceptions of the status quo and the decision frame. The possibility of a constructed frame, where only perception is operative regardless of objective circumstances, is also consistent with Kahneman and Tversky’s observation that “the reference point is the state to which one has become adapted” and there are many cases in which “the reference point is determined by events that are only imagined” (1982, 171–2). 13The data were generated using the Eugene software program version 3.204 (Bennett and Stam 2000). 14The set of states for j are those in i’s region. 15Recall, increasing values represent a decrease in security along the continuum. 16Please see Addendum B and C contained in the POLITY IV codebook for the rules used to construct the concept variables. 17Following Weeks (2012), this dummy variable is coded as “1” if the initiator has a Polity IV durability score of less than 3. 18See also the long-standing debate between Waltz and Sagan (2003). 19Controls for duration dependence excluded to save space. 20In Table 2, the results for trade dependence are inconsistent. This may suggest that the relationship between trade and weak state initiation is different than the established relationship for conflict more broadly or that it is not a stable predictor for weaker state initiation across regime types. 21Indeed, power transition theory has long argued that power parity creates an environment in which conflict is more likely and that asymmetric capability reduces initiation (Kugler and Lemke 1996). 22With the exception Levi and Whyte (1997), traditional prospect theory research on framing and foreign policy has also largely utilized case study analysis. References Allison Graham Zelikow Philip. 1999. Essence of Decision: Explaining the Cuban Missile Crisis . Vol. 2. New York: Pearson. Arrefuin-Toft Ivan 2005. How the Weak Win Wars: A Theory of Asymmetric Conflict : Cambridge, MA: Harvard University Press. Bazerman Max H. Magliozzi Thomas Neale Margaret A.. 1985. “ Integrative Bargaining in a Competitive Market.” Organizational Behavior and Human Decision Processes  35: 294– 313. Google Scholar CrossRef Search ADS   Beck Nathaniel Katz Jonathan N. Tucker Richard. 1998. “Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable.” American Journal of Political Science  42: 1260– 88. Google Scholar CrossRef Search ADS   Bell Mark S. Miller Nicholas L.. 2013. “Questioning the Effect of Nuclear Weapons on Conflict.” Journal of Conflict Resolution  59: 4– 92. Benedetto De Martino Kumaran Dharshan Seymour Ben J. Dolan Raymond. 2006. “Frames, Biases, and Rational Decision-Making in the Human Brain.” Science  313: 684– 7. Google Scholar CrossRef Search ADS PubMed  Bennett D. Scott Stam Allan C.. 2000. “Eugene: A Conceptual Manual.” International Interactions  26: 179– 204. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. 1997. “The Gains Debate: Framing State Choice.” American Political Science Review  91: 789– 805. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. 2002. “Model Building with Prospect Theory: A Cognitive Approach to International Relations.” Political Psychology  23: 759– 86. Google Scholar CrossRef Search ADS   Berejikian Jeffrey D. “Prospect Theory in International Relations.” In Oxford Bibliographies in International Relations , edited by James Patrick. New York: Oxford University Press (forthcoming). Boettcher William Cobb Michael. D.. 2009. “Don't Let Them Die in Vain”: Casualty Frames and Public Tolerance for Escalating Commitment in Iraq.” Journal of Conflict Resolution  53: 677– 97. Google Scholar CrossRef Search ADS   Boettcher William, A. 2004. “The Prospects for Prospect Theory: An Empirical Evaluation of International Relations Applications of Framing and Loss Aversion.” Political Psychology  25: 331– 62. Google Scholar CrossRef Search ADS   Bueno De Mesquita Bruce. 1981. The War Trap. New Haven, CT: Yale University Press. Bueno De Mesquita Bruce. 1985. “The War Trap Revisited: A Revised Expected Utility Model.” The American Political Science Review  79: 156– 77. Google Scholar CrossRef Search ADS   Bueno de Mesquita Bruce Smith Alastair Siverson Randolph M. Morrow James D.. 2003. The Logic of Political Survival . Cambridge, MA: The MIT Press. Cha Victor D. 2000. “Abandonment, Entrapment, and Neoclassical Realism in Asia: The United States, Japan, and Korea.” International Studies Quarterly  44: 261– 91. Google Scholar CrossRef Search ADS   Chan Steve. 2010. “Major-Power Intervention and War Initiation by the Weak.” International Politics  47: 163– 85. Google Scholar CrossRef Search ADS   Chong Dennis Druckman James N.. 2007. “Framing Theory.” Annual Review of Political Science  10: 103– 26. Google Scholar CrossRef Search ADS   Clapton William. 2011. “Risk in International Relations.” International Relations  25: 280– 95. Google Scholar CrossRef Search ADS   Conroy Hilary Wray Harry. 1990. Pearl Harbor Reexamined: Prologue to the Pacific war . edited by Conroy Hilaryand Wray Harry, 1990. Honolulu, HI: University of Hawaii Press. Druckman James. N. 2004. “Political Preference Formation: Competition, Deliberation, and the (ir) Relevance of Framing Effects.” American Political Science Review  98: 671– 86. Google Scholar CrossRef Search ADS   Dueck Colin. 2009. “Neoclassical Realism and the National Interest: Presidents, Domestic Politics, and Major Military Interventions.” In Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W. eds. Neoclassical realism, the state, and foreign policy , Cambridge, UK: Cambridge University Press 139– 169. Edelstein David M. 2002. “Managing Uncertainty: Beliefs about Intentions and the Rise of Great Powers.” Security Studies  12: 1– 40. Google Scholar CrossRef Search ADS   Fearon James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review  88: 577– 92. Google Scholar CrossRef Search ADS   Fearon James D. 1995. “Rationalist Explanations for War.” International Organization  49( 3 ): 379– 414. Google Scholar CrossRef Search ADS   Garrison Jean A. 2001. “Framing Foreign Policy Alternatives in the Inner Circle: President Carter, His Advisors, and the Struggle for the Arms Control Agenda.” Political Psychology  22: 775– 807. Google Scholar CrossRef Search ADS   Geddes Barbara. 2003. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics . Ann Arbor, MI: University of Michigan Press. George Alexander L. Smoke Richard. 1974. Deterrence in American Foreign Policy: Theory and Practice . New York: Columbia University Press. Gilpin Robert. 1983. War and Change in World Politics . Cambridge, UK: Cambridge University Press. Gleditsch Kristian Skrede Salehyan Idean Schultz Kenneth. 2008. “Fighting at Home, Fighting Abroad How Civil Wars Lead to International Disputes.” Journal of Conflict Resolution  52: 479– 506. Google Scholar CrossRef Search ADS   Goertz Gary Diehl Paul F.. 1993. “Enduring Rivalries: Theoretical Constructs and Empirical Patterns.” International Studies Quarterly . 37: 147– 71. Google Scholar CrossRef Search ADS   Hancock Landon E. Weiss Joshua N. Duerr Glen M. E.. 2010. “Prospect Theory and the Framing of the Good Friday Agreement.” Conflict Resolution Quarterly  28: 183– 203. Google Scholar CrossRef Search ADS   He Kai Feng Huiyun. 2013. Prospect Theory and Foreign Policy Analysis in the Asia Pacific: Rational Leaders and Risky Behavior . New York: Routledge. Helfstein Scott. 2012. “Backfire: Behavioral Decision Making and the Strategic Risks of Successful Surprise.” Foreign Policy Analysis  8: 275– 92. Google Scholar CrossRef Search ADS   Horowitz Michael. 2009. “The spread of nuclear weapons and international conflict: Does experience matter?” Journal of Conflict Resolution  53: 234– 57. Google Scholar CrossRef Search ADS   Hudson Valerie. 2005. “ Foreign Policy Analysis: Actor-Specific Theory and the Ground of International Relations.” Foreign Policy Analysis  1( 1): 1– 30. Google Scholar CrossRef Search ADS   Jervis Robert. 1988. “War and Misperception.” Journal of Interdisciplinary History  18: 675– 700. Google Scholar CrossRef Search ADS   Jervis Robert. 1992. “Political Implications of Loss Aversion.” Political Psychology  13: 187– 204. Google Scholar CrossRef Search ADS   Johnson Dominic D. P. 2009. Overconfidence and War . Cambridge, MA: Harvard University Press. Kahneman Daniel Tversky Amos. 1979. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica  47: 263– 91. Google Scholar CrossRef Search ADS   Kahneman Daniel Tversky Amos 1982. “The Psychology of Preferences.” Scientific American  246: 160– 73. Google Scholar CrossRef Search ADS   Kam Cindy D. Simas Elizabeth N.. 2010. “Risk Orientations and Policy Frames.” The Journal of Politics  72: . Kitchen Nicholas. 2010. “Systemic Pressures and Domestic Ideas: A Neoclassical Realist Model of Grand Strategy Formation.” Review of International Studies  36: 117– 43. Google Scholar CrossRef Search ADS   Klein James P. Goertz Gary Diehl Paul F.. 2006. “The New Rivalry Dataset: Procedures and Patterns.” Journal of Peace Research  43: 331– 48. Google Scholar CrossRef Search ADS   Kugler Jacek Lemke Douglas. 1996. Parity and War: Evaluations and Extensions of the War Ledger . Ann Arbor, MI: University of Michigan Press. Kuhnen Camelia M. Knutson Brian. 2005. “The Neural Basis of Financial Risk Taking.” Neuron  47: 763– 70. Google Scholar CrossRef Search ADS PubMed  Lai Brian Slater Dan. 2006. “Institutions of the Offensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes, 1950–1992.” American Journal of Political Science  50: 113– 26. Google Scholar CrossRef Search ADS   Lemke Douglas. 2002. Regions of War and Peace . Cambridge, UK: Cambridge University Press. Lemke Douglas Reed William. 1996. “Regime Types and Status Quo Evaluations: Power Transition Theory and the Democratic Peace.” International Interactions  22: 143– 64. Google Scholar CrossRef Search ADS   Levi Ariel S. Whyte Glen. 1997. “A Cross-Cultural Exploration of the Reference Dependence of Crucial Group Decisions under Risk: Japan's 1941 Decision for War.” Journal of Conflict Resolution  41: 792– 813. Google Scholar CrossRef Search ADS   Levy Jack S. 1992. “Prospect Theory and International Relations: Theoretical Applications and Analytical Problems.” Political Psychology  13: 283– 310. Google Scholar CrossRef Search ADS   Levy Jack S. 1997 . “Prospect Theory, Rational Choice, and International Relations.” International Studies Quarterly  41: 87– 112. Google Scholar CrossRef Search ADS   Levy Jack S. 2000. “Loss Aversion, Framing Effects, and International Conflict: Perspectives from Prospect Theory.” In Handbook of War Studies II , edited by Midlarsky M. I., 193– 221. Ann Arbor, MI: University of Michigan Press. Lobell Steven E. 2009. “Threat assessment, the state, and foreign policy: A neoclassical realist model.” In Neoclassical realism, the state, and foreign policy , edited by Lobell, Steven E., Norrin M. Ripsman, and Jeffrey W. Taliaferro. 42–74. Cambridge, UK: Cambridge University Press. Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W.. 2009. Neoclassical Realism, the State, and Foreign Policy . New York: Cambridge University Press. Ma Qingguo Feng Yandong Xu Qing Bian Jun Tang Huixian. 2012. “Brain Potentials Associated with the Outcome Processing in Framing Effects.” Neuroscience Letters  528: 110– 3. Google Scholar CrossRef Search ADS PubMed  McDermott Rose. 2004 . “Prospect Theory in Political Science: Gains and Losses from the First Decade.” Political Psychology  25: 289– 312. Google Scholar CrossRef Search ADS   McDermott Rose Fowler James H. Smirnov Oleg. 2008. “On the Evolutionary Origin of Prospect Theory Preferences.” Journal of Politics  70: 335– 50. Google Scholar CrossRef Search ADS   Jonathan Mercer, “Prospect Theory and Political Science.” Annual Review of Political Science  8: 1– 21. Kunz Barbara Saltzman I. Z.. 2012. External and Domestic Determinants of State Behavior. Neoclassical Realism in European Politics , 96– 116. Manchester: Manchester University Press. Mintz Alex Redd Steven B.. 2003. “Framing Effects in International Relations.” Synthese  135: 193– 213. Google Scholar CrossRef Search ADS   Miu Andrei C. Crişan Liviu G.. 2011. “Cognitive Reappraisal Reduces the Susceptibility to the Framing Effect in Economic Decision Making.” Personality and Individual Differences  51: 478– 82. Google Scholar CrossRef Search ADS   Maoz Zeev. 1990. “Framing the National Interest: The Manipulation of Foreign Policy Decisions in Group Settings.” World Politics  43: 77– 110. Google Scholar CrossRef Search ADS   Marshall Monty G. Keith Jaggers. 2002. “Polity IV project: Political regime characteristics and transitions, 1800–2002”. Oneal John R. Russett Bruce. 1999. “Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict.” Journal of Peace Research  36: 423– 42. Google Scholar CrossRef Search ADS   Organski A. F. K. 1968. World Politics. New York: Knopf. Paul Thazha Varkey. 1994. Asymmetric Conflicts: War Initiation by Weaker Powers . Cambridge, UK: Cambridge University Press. Peceny Mark Beer Caroline C. Sanchez-Terry Shannon. 2002. “Dictatorial Peace?” American Political Science Review  96: 15– 26. Google Scholar CrossRef Search ADS   Peceny Mark Beer Caroline C.. 2003. “Peaceful Parties and Puzzling Personalists.” American Political Science Review  97: 339– 42. Google Scholar CrossRef Search ADS   Purdon Mark. 2013. “Neoclassical Realism and International Climate Change Politics: Moral Imperative and Political Constraint in International Climate Finance.” Journal of International Relations and Development  17: 301– 38. Putnam Robert D. 1988. “Diplomacy and Domestic Politics: the Logic of Two-Level Games.” International Organization  42: 427– 60. Google Scholar CrossRef Search ADS   Quattrone George A. Tversky Amos. 1988. “Contrasting Rational and Psychological Analyses of Political Choice.” American Political Science Review  82: 719– 36. Google Scholar CrossRef Search ADS   Quinn Adam. 2013. “Kenneth Waltz, Adam Smith and the Limits of Science: Hard Choices for Neoclassical Realism.” International Politics  50: 159– 82. Google Scholar CrossRef Search ADS   Rathbun Brian. 2008. “A Rose by Any Other Name: Neoclassical Realism as the Logical and Necessary Extension of Structural Realism.” Security Studies  17: 294– 321. Google Scholar CrossRef Search ADS   Rose Gideon. 1998. “Neoclassical Realism and Theories of Foreign Policy.” World Politics  51: 144– 72. Google Scholar CrossRef Search ADS   Rubinstein Mark. 2001. “Rational Markets: Yes or No? The Affirmative Case.” Financial Analysts Journal  57: 15– 29. Google Scholar CrossRef Search ADS   Russell Thomas Thaler Richard. 1985. “The Relevance of Quasi Rationality in Competitive Markets.” The American Economic Review  75: 1071– 82. Schweller Randall L. 2009. “Neoclassical realism and state mobilization: expansionist ideology in the age of mass politics.” In Neoclassical realism, the state, and foreign policy , Lobell Steven E. Ripsman Norrin M. Taliaferro Jeffrey W. eds. 227– 50. Cambridge, UK: Cambridge University Press. Schweller Randall L. 2008. Unanswered Threats: Political Constraints on the Balance of Power . Princeton, NJ: Princeton University Press. Sechser Todd S. 2010. “Goliath's Curse: Coercive Threats and Asymmetric Power.” International Organization  64: 627– 60. Google Scholar CrossRef Search ADS   Singer David J. 1988. “Reconstructing the correlates of war dataset on material capabilities of states, 1816–1985.” International Interactions  14: 115– 32. Google Scholar CrossRef Search ADS   Singh Sonali Way Christopher R.. 2004. “ The Correlates of Nuclear Proliferation: A Quantitative Test.” Journal of Conflict Resolution  48: 859– 85. Google Scholar CrossRef Search ADS   Signorino Curtis S. Ritter Jeffrey M.. 1999. “Tau-b or Not Tau-b: Measuring the Similarity of Foreign Policy Positions.” International Studies Quarterly  43: 115– 44. Google Scholar CrossRef Search ADS   Sirin Cigdem V. 2011. “Is It Cohesion or Diversion? Domestic Instability and the Use of Force in International Crises.” International Political Science Review  32: 303– 21. Google Scholar CrossRef Search ADS   Snyder Glenn H. 1984. “The Security Dilemma in Alliance Politics.” World Politics  36: 461– 95. Google Scholar CrossRef Search ADS   Sobek David Foster Dennis M. Robison Samuel B.. 2012. “Conventional wisdom? The effect of nuclear proliferation on armed conflict, 1945–2001” International Studies Quarterly  56: 149– 62. Taliaferro Jeffrey W. 1998. “Quagmires in the Periphery: Foreign Wars and Escalating Commitment in International Conflict.” Security Studies  7: 94– 144. Google Scholar CrossRef Search ADS   Taliaferro Jeffrey W. 2006. “State Building for Future Wars: Neoclassical Realism and the Resource-Extractive State.” Security Studies  15: 464– 95. Google Scholar CrossRef Search ADS   Thaler Richard. 1980. “Toward a Positive Theory of Consumer Choice.” Journal of Economic Behavior & Organization  1: 39– 60. Google Scholar CrossRef Search ADS   Tom Sabrina Fox Craig. R. Trepel Christopher Poldrack Russell A.. 2007. “The Neural Basis of Loss Aversion in Decision-Making Under Risk.” Science  315: 515– 8. Google Scholar CrossRef Search ADS PubMed  Trepel Christopher Fox Craig R. Poldrack Russell A.. 2005. “Prospect Theory on the Brain? Toward a Cognitive Neuroscience of Decision under Risk.” Cognitive Brain Research  23: 34– 50. Google Scholar CrossRef Search ADS PubMed  Toje Asle Kunz Barbara eds. 2012. Neoclassical Realism in European Politics: Bringing Power Back in . Manchester, UK: Manchester University Press. Van Evera Stephen. 2013. Causes of War: Power and the Roots of Conflict . Ithaca, NY: Cornell University Press. Waltz Kenneth. 1979. Theory of International Politics . Reading, MA: Addison-Wesley Pub. Co. Waltz Kenneth Sagan Scott D. eds. 2003. The Spread of Nuclear Weapons: A Debate Renewed . New York: WW Norton. Weeks Jessica L. 2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization  62: 35– 64. Google Scholar CrossRef Search ADS   Weeks Jessica L. 2012. “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict.” American Political Science Review  106: 326– 47. Google Scholar CrossRef Search ADS   Wright William F. Urton Anderson 1989. “Effects of situation familiarity and financial incentives on use of the anchoring and adjustment heuristic for probability assessment.” Organizational Behavior and Human Decision Processes  44: 68– 82. Google Scholar CrossRef Search ADS   Zakaria Fareed. 1999. From Wealth to Power: the Unusual Origins of America's World Role . Princeton, NJ: Princeton University Press. © The Author (2016). Published by Oxford University Press on behalf of the International Studies Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Foreign Policy AnalysisOxford University Press

Published: Apr 27, 2016

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off