The Weakest Link: Credible Deterrence Threats and Alliance Entrapment

Puzzle

In the early days of the Libyan uprising, there was little US interest in becoming involved in the conflict. The United States was already involved in multiple costly wars overseas, and Libya was not necessary to the overall security strategy. Nevertheless, on March 16th, the United States became officially involved in aiding to establish a no-fly zone in Libya (United Nations Security Council 2011). Despite little strategic interest and even less desire to intervene, the United States eventually funded and carried out the majority of the operations in Libya. Some have suggested that the decision to intervene was largely a result of external pressure placed on the United States by allies within NATO, particularly France. In this case, it seems as though the United States' defensive alliance partners were able to coerce the United States into becoming involved in a conflict which did not necessarily meet the bounds of the North Atlantic Treaty. Conversely, the United States has also refused to intervene in some conflicts despite the desires of alliance members. France and Belgium lobbied the United States to help end the Rwandan genocide after France intervened in 1994 (Taha 2011, 34). The United States refused to participate and did not send troops or otherwise support its alliance members in this conflict. This situation demonstrates two competing trends in the capacity of alliances to trap a state into entering into a conflict against its will. The United States occasionally succumbs to alliance pressures, but it is not always forced to intervene against its will. So, do alliances actually force states to intervene in conflicts against their will?

The potential for defensive alliances to result in an intervention that is outside the bounds of a state's foreign policy interests has several important policy implications. The United States and other governments have used defensive alliances as a way of increasing the overall security of the state. If defensive alliances actually force states to intervene in areas against their interests, then they would actually diminish the overall security of the state rather than increase it. Since these alliances are meant to increase the security of the state, a finding that member states are drawn into more conflicts would demonstrate that these alliances are not an effective deterrent. This might have greater policy implications down the road, as states must make the decision whether or not to enter into an alliance. Thus, there are important policy issues at stake when examining whether or not defensive alliances truly minimize conflict or merely cause a state to be pulled into conflicts that they have not chosen.

Thus, this paper will attempt to determine if all states are drawn into conflicts originated by alliance members equally across different levels of state strength. There is a wide body of literature which examines the efficacy of alliances in preventing violent conflict. However, there is little research which specifically examines the internal dynamics of defensive alliances. Most of this research does not distinguish between when a third party joins in conflicts that were originated by the alliance member or conflicts that were originated by another party. While there is a large literature base concerning the efficacy of defensive alliances in protecting states against outside aggressors, there is less empirical research examining whether states in defensive alliances are forced to join in when alliance members originate a conflict. In addition, little of this literature focuses on the impact of state strength in terms of the potential for alliance entrapment. Thus, this paper will contribute to the scholarship on the efficacy of alliances as a deterrent to conflict by examining whether and how alliance members are drawn into conflicts they did not initiate.

Research Question

This paper will attempt to determine why some states are drawn into conflicts by defensive alliance members which are outside of the bounds of the defensive alliance and the best interests of the state when others are not. Alliances are supposed to demonstrate a credible threat which makes the cost of conflict higher for potential adversaries. This should hold true across different levels of state strength. Thus, defensive alliance partners should only be drawn into conflict very rarely because few potential adversaries would wish to challenge a state which has multiple partners to defend it. However, most scholars would also agree that the strength of the deterrent threat is tied to the relative strength potentially lose in the conflict. A loss of strength to any alliance member weakens the deterrent threat and makes each member less secure. This is true across all levels of state strength but is particularly worrisome for weaker states. A stronger state would lose comparatively less security when the alliance partner is weakened, but an already weak state would find itself substantially less secure. Thus, weaker states should be more willing to defend their alliance partners or join into conflicts originated by alliance partners to prevent the weakening of the alliance's deterrent threat. The deterrent threat sent by an alliance is vital in protecting the weaker state against potential attackers. Thus, this paper will attempt to provide support for two main hypotheses:

1. In cases where an alliance member originates a conflict, a great power or strong state will not necessarily feel compelled to join in the conflict because the state is stable enough that changes in the strength of alliance member do not impact the security of the state.
2. In cases where an alliance member originates a conflict, a weak or medium powered state will be compelled to join in the conflict, even when the conflict may go against the state's desires or the mandate of the alliance because changes in the strength of the alliance member impact the security of the state.

Literature Review

In order to accurately examine the question of why some alliance members are drawn into conflicts while others are not, it is important to understand this question in terms of its position within academic literature and the debates which surround it. The following schools of thought attempt to explain why conflict occurs and why deterrence fails in specific situations. This literature review will focus on three distinct schools of thought; grievance, Rubicon theory, and rational deterrence. Thus, in order to understand deterrence and its effect on defensive alliances, it is necessary to examine these schools of thought in terms of implications for policy and as representations of the decision-making process of the nation-state.

Grievance is a structure-based account of conflict. Collective action occurs, according to this school, when there is relative deprivation of access to the government, human security, or economic opportunities. Thus, conflict occurs because, "Communal groups are jointly motivated by deep-seated grievances about group status and by the situationally determined pursuit of political interests, as formulated by group leaders and political entrepreneurs" (Gurr 1993, 166-167). Groups with a strong communal identity will contribute to conflict due to narratives of grievance based in a sense of injustice created by factors like economic disadvantages, discrimination, and poverty. This model would be inadequate as an explanation of the way that defensive alliances pull alliance partners into conflicts. The "Rebel's Dilemma," as posited by Mark I. Lichbach, argues that grievances are not enough to motivate conflict as there are more costs, in terms of forgone wages, the risk of death, and possible imprisonment than there are benefits to be gained from rebelling or participating in conflict (1994, 387). Thus, it is unlikely that a weak defensive alliance member would be willing to enter into a conflict, even in support of an alliance partner. Weak alliance partners should be unwilling to join into conflict because the conflict would diminish the security of the state. The weak alliance partner would be willing to risk a small loss of security if its partner loses rather than weakening itself substantially by entering into this conflict. Thus, weak states would be unwilling to originate costly conflicts which would only make the state weaker. It is therefore necessary to examine other schools of thought for a more complete theory to explain the decision to engage in a conflict.

Rubicon theory offers another explanation for deterrence failure which may be useful when considering the hypothesis. According to this model, states operate in terms of sequential decision frames which impact their behavior. Initially, states operate under a deliberative mind-set where they are able to rationally compare options and possible outcomes in order to make the best decision. This stage is characterized by the ability to process new information, weigh costs and benefits, and make rational choices. However, after the decision to act has been made, states enter into an implemental mindset. This shift "can occur either when an actor freely chooses a policy from a menu of options or when a course of action is dictated by an external source," (Johnson and Tierney 2011, 14). In the implemental mindset, the state has become committed to a specific course of action. This means that the state will be less willing to change course or to accept new information. Any new information that is received is generally processed according to the mission of the state, i.e. new information is made to fit the pre-stated goals of the mission. This means that once states have made the decision to act, they become vulnerable to self-serving evaluations which confirm the decisions of the state, rather than presenting evidence in a useful and unbiased manner. States, therefore, become overconfident in their ability to 'win' the conflict. Johnson and Tierney argue that the shift between the deliberative and implemental stages occurs when there is a perception that war is near (2011, 15). One of the potential causes of this perception is that the state feels forced to participate in a conflict because of its alliances. This would explain why states in defensive alliances participate in conflict when alliance members are also involved.

By entering into the alliance contract, the state has in a way already passed into the implemental stage where any conflict involving an alliance member is seen as one in which the state has already agreed to participate. This means that the state should enter into any conflict in which another member of the defensive alliance is involved. While this argument seems to fit the hypothesis, it fails to explain one important intervening variable. According to this school, once the state has crossed the Rubicon of signing the defensive pact, the state should intervene in all conflicts because they perceive that they are bound to their alliance members. However, the hypotheses argue that only weaker states will intervene to protect an alliance member. The Rubicon model cannot take this discrepancy into account.

Yet another school of thought that could explain the reason that other alliance members frequently draw alliance partners into conflicts is rational deterrence. There are three important assumptions that characterize this school: (1) Actors have exogenous given preferences and options; (2) variation in outcomes is explained by different actors' opportunities; and (3) the state acts as a unitary, rational actor (Achen and Snidal 1989, 151). Many scholars in the rational choice school of deterrence argue that in order to prevent an aggressor state from pursuing violent action, the state must prevent a credible threat which makes the cost of engaging in violence outweigh the potential benefits (Achen and Snidal 1989, 157). Further, most authors are in agreement that membership in a defensive alliance, where states band together to increase the credible deterrent threat signal that they can send to potential aggressors, will decrease the likelihood of conflict by increasing the costs associated with a potential attack. Thus, states should enter into conflicts on behalf of their alliance partners only when (1) mandated by the terms of the alliance agreement or (2) when the conflict threatens to substantially decrease the deterrent signal that is being sent by the alliance as a whole. This means that states in defensive alliances should face fewer conflicts or threats of violence from aggressor states.

The rational deterrence model seems as though it would be more accurate than the Rubicon model because it would be able to predict both instances where states would intervene, such as when it would decrease the deterrent signal of the alliance, and when states would not intervene. The Rubicon model predicts intervention in all conflicts, which seems overly simplistic. However, while many scholars have argued that states in defensive alliances don't enter into conflict as frequently, few have examined the reasoning behind the decision of defensive alliance members when they do chose to enter certain conflicts. This paper would argue that stronger states are less likely to enter conflicts originated by alliance members because the potential that the alliance member loses would not decrease the deterrent signal as much as if a strong state intervened and lost. Alternatively, a weaker state that is more dependent on the signal that is sent by a defensive alliance would be more likely to enter into a conflict originated by an alliance partner because the diminished credible deterrence signal that would result from a potential loss would have a more substantial impact on the security of the state.

Of the three schools of thought examined here, rational deterrence seems to offer the clearest model for predicting the potential of defensive alliances to entrap weaker alliance members in conflicts that they did not originate more often than stronger alliance members. It is important to note that while these scholars have attempted to explain why deterrence fails and how alliances contribute to deterrence success or failure, there has been little examination of whether the relative strength of the alliance partner contributes to the decision to intervene. While the Rubicon model would predict that states should intervene in any conflict originated by an alliance member, the rational deterrence model would argue that intervention is only will only occur if the alliance partner is relatively weak and depends on the credible deterrent signal sent by the alliance for its security against potential aggressors.

Argument

The objective of this paper is to demonstrate that in certain situations, a defensive alliance can result in some states being pulled into conflicts that they would have little strategic interest in pursuing on their own. In order to establish the validity of this argument, it is important to situate it within an intellectual school of thought. I will use rational deterrence theory to explain the reasons why states in defensive alliances may be less secure in some situations than they would be if they were not in a defensive alliance. Rational deterrence will provide a baseline from which I can explore the following hypotheses and determine whether or not they are supported by the data.

Rational deterrence argues that states have exogenous given preferences that are tested against the preferences of other states in order to determine what the outcome in any given conflict or potential conflict would be. This means that states weigh the costs and benefits of initiating conflict before they actually do so. If the costs are higher for an aggressor state than the potential benefits, then there will be no conflict. In addition, a state can signal a deterrent threat to potential aggressor states that demonstrates the costs of a potential attack through higher military spending, political statements, or many other forms of signaling behavior (Achen and Snidal 1989, 151). The logical extension of this argument is that since higher costs mean that aggressor states are less likely to engage in hostile activities, states can and should bind together in alliances so that each can benefit from the power of the other alliance members. If this statement is true, then defensive alliances should act as deterrents to external aggression because each member of the alliance promises to defend every other alliance member if their territorial integrity is threatened. This explanation of rational deterrence as an incentive for building alliances is well established within the literature.

However, this paper argues that there is a codicil to this argument. In order for a defensive alliance to act as an effective deterrent to external aggression, the alliance must be able to pose a credible threat, which will prevent states outside of the alliance from threatening its members. Thus, defensive alliances must present a strong enough deterrent threat to protect even the weakest members of the alliance. This means that the weak or medium powered states within an alliance are particularly invested in maintaining the strength of the deterrent signal sent by the alliance. The deterrent signal that is sent by the alliance is the combination of each state's relative strength and military capability, thus any loss of strength by one alliance member decreases the efficacy of signal sent by the alliance as a whole. Since weaker states depend on this signal to protect the state from potential adversaries, there is an inherent obligation for lower powered states to ensure that the signal remains strong. Conversely, great powers or strong states depend less on the deterrent signal sent by the alliance and more on the military and economic strength of the state itself. Thus, weaker states within a defensive alliance would be drawn into conflicts where the other alliance members are the aggressor because if the alliance member should lose, the credibility of the alliance's deterrent threat would be diminished. Even though the mandate of the defensive alliance does not cover entry into non-defensive conflicts, states involved in defensive alliances could find themselves involved in conflicts originated by weaker alliance members. Figure 1 demonstrates the logic of this argument visually.

The present study will examine three main variables which will be quantified in detail in the Research Design section of this paper. First, the independent variable is the state's membership in a defensive alliance with a conflict originator. There is also a control variable that will be tested through this paper. Second, the present study will control for whether or not a weak or strong member state originated the conflict. This variable will require a quantification of "low," "medium" or "high" levels of state strength. Finally, the dependent variable is whether the state originated the conflict. Essentially, this describes whether the state started the conflict or merely entered the conflict after it was initiated. Since each data entry represents a state participant in a conflict, each state included in the study was involved as either an originator or a joiner. This allows the study to determine which states within a defensive alliance are more likely to join a conflict which has already been initiated by an alliance partner. These three variables will help to determine whether or not the hypothesis is supported by historical examples.

If the argument is proven to be valid, the data should show that states with "low" or "medium" strength are more likely to join in a conflict originated by an alliance member. States with small populations, less military funding, and fewer available troops depend more on their alliance partners to send a credible threat to potential adversaries, so it is more likely that these states will act to prevent other alliance partners from losing strength as a result of entry into a conflict. In order for the alliance to send a valid deterrent signal, each state within the alliance must maintain a certain level of strength. Thus, states of low or medium power will be unwilling to risk the potential for a partner to lose and decrease the value of the alliance's deterrent signal. However, state of high power or great powers are less reliant on their alliance partner's strength to protect them from potential adversaries. These states have large militaries and are able to spend large amounts of money on military purchases. This means that the military in each of these states is capable of sending a deterrent signal that is not dependent on the strength of alliance partners. Therefore, great powers are less tied to their alliance partners and rely on them less to bolster their own strength. According to this logic, the data ought to demonstrate that low and medium powers are more likely to join into a conflict originated by an alliance member, where high power states are able to behave more independently.

However, if the evidence demonstrates that alliance members are equally likely to enter into a conflict originated by another alliance member, whether weak or strong, then the argument would not be supported. This evidence would contribute to the validity of the Rubicon theory because it argues that once a state has entered into a defensive alliance, it has crossed a Rubicon, having committed to helping the alliance member regardless of the situation. So, the state would participate in any conflict originated by any alliance partner, regardless of its relative strength or weakness. Alternatively, should the data demonstrate that few or none of the weaker defensive alliance partners joined in a conflict originated by an alliance member, then that could be seen as support for the grievance model. The weaker state knew that the cost of conflict would be too high, so it would be unwilling to enter into a conflict that is technically outside of the scope of the alliance at the risk of diminishing its own security.

This research will add to the academic debates about defensive alliances by examining a variable which has been understudied in the literature to date. The idea that state strength is an intervening variable that determines whether a state will join when a defensive alliance member originates a conflict is one that has not been tested. Additionally, this research could have several important implications for policy-making. Should the hypothesis be supported, it would perhaps indicate that entering into a conflict with weaker states is not necessarily in the best interest of the state because weaker states would be more likely to band together in support of an alliance partner. Previously, states have used defensive alliances to protect weaker states, but if the hypothesis were true, then entering into these alliances would actually decrease the security of the weaker state because they would be increasingly drawn into conflicts originated by alliance partners. Thus, weak states would be much less willing to enter into defensive alliances in the future.

Research Design

In order to test the claims made in the hypothesis, three datasets from the Correlates of War Project will be used to operationalize the three main variables examined in this paper. The dependent variable in this examination is whether or not a state originated a conflict. The Correlates of War Project's Militarized Interstate Dispute (MID) dataset was used to determine which states originated a conflict versus merely participating or joining (Ghosn 2004). This dataset contains one entry per participant in each conflict for any given set of years. In order to make the dataset more manageable, this study was limited to cases in the most recent five-year period available, 1995 to 2000. While this study will only cover a period of five years, the methods of statistical analysis will still be appropriate because of the large number of cases, which number 535. It is true that there are many events outside of this timeframe that would prove interesting to examine in light of the hypotheses, but the nature of this paper meant that a larger statistical analysis was not possible within the time allotted. In addition, the decision to utilize the most recent five year period was in an effort to avoid selection bias. Rather than choosing a period of time with heavy NATO activity, like the period of the Cold War, utilizing the most recent time period available meant that I was not able to skew the results by picking a specific timeframe that would conform to the hypothesis. This variable was coded as a nominal variable where states that originated the conflict had a value of 1 and states that did not originate the conflict had a value of 0.

The independent variable in this study is whether or not a state is in a defensive alliance with a conflict originator. Membership is defined as having a signed treaty with one or more other nations, which obligates the state to defend its alliance partners should they be attacked by states outside of the alliance. The Formal Alliances dataset compiled for the Correlates of War database was used to operationalize this variable (Gibler 2004). This dataset provides a list of defensive alliances in the form of a dyadic list of states in an alliance with an entry for each year. Thus, each entry in the MID dataset was examined to determine which states were in an alliance with an originator in the year of the conflict. A variable was created for each of the 535 cases that described whether the state was in a defensive alliance with the originator of that specific conflict when it began. Of course, this means that the state that originated the conflict is also often included as an alliance member since in many situations, there were multiple originators in an alliance with each other. Obviously, there are other factors than alliance membership which determine whether a state will enter into a conflict. However, the goal of this paper is to determine whether a state's alliance membership is statistically significant to their decision to enter into a given conflict.

Finally, state strength is operationalized using the National Material Capabilities dataset from Correlates of War (Singer 1987). In order to gain a more complex view of state strength, the Composite Index of National Capability (CINC) scores were used. This score is a combination of six different variables that influence state strength: iron and steel production, military expenditure, military personnel, energy consumption, total population, and urban population. These factors are by no means exhaustive and there are other factors that could influence state strength. However, this operationalization is defended by the Correlates of War Database, which is well respected in the field of international relations and does include some non-military factors, including energy consumption, total population, and urban population. This makes the operationalization of state strength fairly inclusive. The present study defines weak states as those with low numbers in each of these categories (Singer 1987, 115). Conversely, strong states would have high levels of these different factors. Since rational deterrence requires a state to send a signal of strength to other states, the definition of weak and strong is based on factors which would increase the cost of attack to an aggressor state like the strength of the military, i.e. factors that demonstrate the relative credibility of the state's deterrent threat. This control variable was organized into an ordinal level variable with three different categories: low, medium, and high. In order to determine the cut-off points for each of these categories, a histogram of the data was examined (Appendix 1). States characterized as having low state strength had CINC scores of less than .004, medium states had scores of .004-.016, and high states had scores of higher than .016. The histogram of cases clearly demonstrates a distinct group of high power states that is separate from the medium and low states, so it is clear that states within that group should be considered as high-powered states. However, it was more difficult to distinguish the medium powers from low. Ultimately, I coded the lowest two groups on the histogram, 269 cases, as weak powers and those between the strong and weak groups as medium. While this differentiation may seem somewhat arbitrary, separating the cases using these numbers allowed for a roughly equal distribution of states between medium and high, with a slightly larger base of weak states. In addition, separating the cases based in these criteria would seem to provide an accurate representation of state strength around the world. There are more weak states than medium states and more medium states than strong states. Thus, it seemed that separating the variables according to these numbers would provide the best statistical observations. These distinctions should allow for meaningful statistical observations about each of the three groups.

This study will utilize the data gathered from the Correlates of War Project's database to perform a quantitative analysis of 535 cases of states entering into conflict from the period of 1995 to 2000. The independent variable for this study is whether the state is in an alliance with the originator of the conflict. The dependent variable is whether the state originated the conflict or not. Both the dependent and independent variables are nominal level variables. The control variable is state strength, which has been broken into three discrete categories, making it an ordinal level variable because the responses can be ordered into a scale. Since both the independent and dependent variables are nominal, three chi-square tests with proportional reduction of error measures will be utilized to analyze the data. These tests will determine whether alliance membership makes a difference in the decision of a state that did not originate a conflict to enter into conflict. The proportional reduction of error measures will demonstrate the strength of the relationship, if it exists. This analysis will make it possible to determine what effect defensive alliance membership has on the decision to enter a conflict.

Analysis

The first step in determining whether the two hypotheses are supported by the data is to determine whether alliance membership has any impact on whether a state enters into a conflict that they did not initiate. Thus, it is necessary to run a chi-square test to determine whether the null hypothesis can be rejected. In this case, the null hypothesis would be that alliance membership with a conflict originator has no impact on the decision of a state to enter into a conflict. If, however, the chi-square test results in statistically significant findings, then it is possible to say that alliance membership does have some impact on the decision of a state to enter into a conflict.

Three chi-square tests were run to determine significance; one for low powered states, one for medium powered states, and one for high powered states. Statistically speaking, a p-value that is less than .05 is generally considered to be a significant result. For the low powered states, the p-value was .000, which indicates that the null hypothesis is not accurate in the case of these states. The chi-square test of the medium powered states resulted in a p-value of .012, which is also a significant result. Finally, the p-value of the chi-square test for high powered states was .255 (Appendix 2). This result is not statistically significant. These results are particularly interesting because they would seem to confirm both of the hypotheses put forward in this paper. The low and medium powered states entered into conflicts that they did not initiate based on whether or not they were in an alliance with the conflict originator. However, the test of the high powered states resulted in a statistically insignificant result, which means that in this case, the null hypothesis is supported. The high powered states did not enter into conflicts based on their alliance partnerships with conflict originators.

While this would seem to support the hypotheses on face value, it is also important to examine the proportional reduction of error (PRE) measurement to determine the strength of the association. The Cramer's V statistic gives us a measure of the strength of the association between the two variables for each of the separate groups. For the low powered states group, the Cramer's V was .403, or a strong association. This means that knowing if a state is in an alliance with a conflict originator allows a 40.3% more accurate prediction of whether the state will have entered into a conflict. For the medium powered states group, the Cramer's V was .213, or a moderately strong association. This means that knowing the alliance status of a state increases our prediction power by 21.3%. Since the test for the high powered states group returned results indicating there was no association between the two variables, it is meaningless to measure the strength of the association with the Cramer's V test. The PRE measure indicates that for both the low and medium powered states, there is a strong association between the two variables being examined.

The directionality of the relationship between membership in an alliance with an originator and originating a conflict is also vital in determining whether the hypotheses have been supported. In order to do this, it is important to examine the percentages of cases where a state that did not initiate a conflict was in an alliance with the originator and joined the conflict. For the groups with statistically significant results, low and medium powered states, it is possible to examine the breakdown of cases to determine the directionality of the results. For low powered states, 38.2% of states that were in an alliance with a conflict originator were not originators of a conflict. This is compared to the 6% of states not in an alliance with the originators that did not originate a conflict (Appendix 3). Thus, it is more likely that weak states will join in a conflict that they did not originate when they are in an alliance with the initiator. Medium powered states had a similar breakdown. For these states, 26.5% of alliance members were not originators of the conflict, where 9.4% of non-alliance members did not originate the conflict. Given the significance of the statistical testing, it is clear that medium and low powered states that did not originate a conflict were more likely to join the conflict if they were in an alliance with an originator. However, it is not possible to make the same claims for high powered states, as the p-value was insignificant.

These results would generally support the two hypotheses. The low and medium powered states are statistically more likely to enter into a conflict originated by an alliance partner than they would be if they were not in an alliance with an originator. In addition, the low states have a stronger association than the medium states. This seems to support the logic that the weaker a state is, the more likely they are to act to prevent a loss of strength from an alliance partner and therefore preserve the credibility of the alliance's deterrent threat. However, there was not a statistically significant association between alliance membership and originating a conflict for the high powered states. This would indicate that for high powered states, alliance membership with a conflict originator has no impact on the decision of a state to enter into a conflict. This data would seem to directly support the hypothesis that strong states or great powers do not intervene in conflicts based on their alliance membership because they are less concerned about a potential reduction in the strength of their alliance partners.

Conclusion

This paper has attempted to demonstrate the importance of state strength in determining alliance behavior. Rational deterrence theorists have long argued that each state within the international system makes decisions based on a rational cost-benefit analysis that aims to maximize the marginal utility of the state's decision to act. Therefore, states can prevent potential aggressors from attacking by increasing the cost of attacking and potentially losing in comparison to the potential benefits of expanded territory or wealth. Following this logic, defensive alliances should increase the security of the state because they allow states to pool resources and send a stronger deterrent signal to potential adversaries than they would be able to alone. This argument has been well documented in the literature on rational deterrence. However, this paper attempts to add a corollary to this argument. Since the strength of the defensive alliance depends on the strength of all of its alliance members, any loss of strength for an alliance member represents a decrease in the credibility of the deterrent signal sent to other states. For strong states or great powers, this is not of great concern. Even without the backing of the alliance, the strong military capacity of these states allows them to deter potential attackers. However, weak or medium powers depend more strongly on the alliance as a deterrent to aggressors. Any decrease in the power of an alliance partner decreases the security of the state itself. This means that weak or medium powers are more invested in maintaining the strength of an alliance and will therefore intervene in conflicts that they did not originate in order to ensure that their alliance members do not lose.

This argument has been supported by the data gathered from the Correlates of War Project. The data showed that for both weak and medium powered states, alliance membership with a conflict originator had a significant impact on the decision of the state to enter into a conflict that it did not initiate. However, the same level of significance was not found for high-powered states. Thus, the first hypothesis, that a great power or strong state will not necessarily feel compelled to join in conflicts initiated by alliance partners because the state is stable enough that changes in the strength of alliance member do not impact the security of the state, has been supported by this study. The second hypothesis, that a weak or medium power state will be compelled to join in conflicts initiated by alliance partners, even when the conflict may go against the state's desires or the mandate of the alliance has also been supported by the data. This indicates that the explanation of alliance behavior backed by the logic of rational deterrence is more accurate than the explanations provided by the grievance model, which predicts that this joining behavior would not occur across all three groups, or the Rubicon model, which predicts that it would occur uniformly across the groups.

However, there were some challenges associated with this study that may have an impact on the results. Due to time constraints, the tests of the hypotheses were only carried out on conflicts occurring within a five-year period. More accurate results might be obtained with further study and an expansion of the dataset to include more cases. In addition, it was impossible to include data for each of the states that were in an alliance with an originator but did not decide to enter into the conflict. The dataset that this study was based on differentiated originators from joiners in conflicts, but did not identify states that did not participate in the conflict at all. It is fairly easy to understand why this would be the case. The number of cases would expand exponentially if every single state were included for each conflict record. However, this means that the data used for this study cannot take into account the decision of states not to join and can only test whether the state is more likely to join when they are in an alliance with the originator or not. A more comprehensive study would have to include states in an alliance with an originator that decided not to join at all.

Despite these drawbacks, the findings of this study allow for an interesting discussion of deterrence and behavior within alliances. Deterrence is supposed to prevent conflict by allowing states to signal the potential costs of an attack to an aggressor. Defensive alliances are supposed to give states a larger base of power with which to warn a potential aggressor. However, if states view the power of each state within an alliance as vital to their ability to send a credible signal, then states may engage in behavior that actually makes them more conflict prone than they would be without the alliance. Since it is clear that low and medium strength states are more likely to enter into conflicts that alliance partners originate than states that are not in a conflict, the result is that weaker states are involved in more conflicts than strong states. Ironically, their involvement in these conflicts most likely contributes to their weakness in terms of military strength. Thus, it would seem that weaker states would be more secure if they were not in defensive alliances at all.

In addition, these findings have broad implications for future United States policy and the understanding of deterrence as a whole. This paper began with a puzzle. The United States intervened in Libya seemingly solely at the behest of its alliance partners, but refused to intervene in Rwanda, despite pressure from some of those same partners. The findings of this paper are able to explain this conundrum. Strong states or great powers like the United States decide to enter into conflict based on factors other than alliance pressure or the potential of an alliance member to weaken the alliance by 'losing' a conflict that they initiated. This means that the United States should not be concerned that defensive alliance partners will always be able to force the United States into action. Defensive alliances still act to increase the security of the United States overall because they allow access to additional support or resources. These alliances are statistically unlikely to force the US into any conflict that it does not wish to be involved in, despite recent events. It also means that the United States should be cautious when intervening in weak or medium strength states that have defensive alliance partners. These partners are forced to view any potential loss of strength as a threat and will act to maintain their security by joining in to any conflict in which alliance partners are involved.

References

Achen, Christopher H. and Duncan Snidal. "Rational Deterrence Theory and Comparative Case Studies." World Politics. Vol. 41, No. 2. (1989), p. 143-169.

Chen, H. C., and W. S. Neilson. "The Reliability of Alliances and Extended Deterrence: A Game-Theoretical Analysis." Issues & Studies 43, no. 2 (2007): 1-37.

Corbetta, R. "Determinants of Third Parties' Intervention and Alignment Choices in Ongoing Conflicts, 1946-2001." Foreign Policy Analysis 6, no. 1 (2010): 61-85.

Ghosn, Faten, Glenn Palmer, and Stuart Bremer. 2004. "The MID3 Data Set, 1993-2001: Procedures, Coding Rules, and Description." Conflict Management and Peace Science 21:133-154.

Gibler, Douglas M., and Meredith Sarkees. 2004. "Measuring Alliances: The Correlates of War Formal Interstate Alliance Data set, 1816-2000." Journal of Peace Research 41(2): 211-222.

Gurr, Ted Robert. "Why Minorities Rebel: A Global Analysis of Communal Mobilization and Conflict Since 1945" International Political Science Review. Vol. 14, No. 2 (April 1993), p. 161-201.

Johnson, Dominic D.P. and Dominic Tierney. "The Rubicon Theory of War: How the Path to Conflict Reaches the Point of No Return." International Security Vol. 36, no. 1 (2011): 7 40.

Johnson, J. C., and B. A. Leeds. "Defense Pacts: A Prescription for Peace?" Foreign Policy Analysis 7, no. 1 (2011): 45-65.

Lichbach, Mark I. "What Makes Rational Peasants Revolutionary?: Dilemma, Paradox, and, Irony in Peasant Collective Action." World Politics Vol. 43, no. 3 (1994): 383-418.

Singer, J. David, Stuart Bremer, and John Stuckey. (1972). "Capability Distribution, Uncertainty, and Major Power War, 1820-1965." in Bruce Russett (ed) Peace, War, and Numbers, Beverly Hills: Sage, 19-48.

Singer, J. David. (1987). "Reconstructing the Correlates of War Dataset on Material Capabilities of States, 1816-1985" International Interactions, 14: 115-32.

Taha, Hagar. The Failure to Protect, Again: A Comparative Study of International and Regional Reactions Towards Humanitarian Disasters in Rwanda and Darfur. Saarbrücken, Germany: Lambert Academic Publishing, 2011.

United Nations Security Council. "Security Council Approves 'No-Fly Zone' over Libya, Authorizing 'All Necessary Measures' to Protect Civilians." SC/10200. 2011. http://www.un.org/News/Press/docs/2011/sc10200.doc.htm

Appendix

Appendix 1: Histogram of State Strength

Appendix 2: Chi-Square Test Results

Appendix 3: Contingency Table