Abstract
Scholars have begun to examine the determinants of military coalition participation. Most studies pay closer attention to international (both systemic/dyadic) rather than domestic factors. While admitting the predominance of such international factors, this study claims that the domestic conditions of potential participants also determine if a state actually joins in a coalition. Specifically, domestic conditions work as suppressing factors for coalition participation. First, riots and violent domestic protests lead to a significant reduction in the probability of sending military troops for a coalition operation, because military resources cannot be deployed when they might be needed to maintain domestic order. Second, a state that is experiencing an economic recession cannot easily send its troops to a coalition operation because the domestic audience would prefer to prioritize resource allocations for domestic economic recovery.
Introduction
In March 2011, when multinational forces began a military operation to establish a no-fly-zone over Libyan airspace and naval patrols to enforce a UN arms embargo, Germany and Turkey made contrasting decisions about participation in the coalition. Both faced similar internal political conditions of a forthcoming, important election and domestic opposition to the military operation against Libya. Germany failed to join in the operations (Weiland and Nelles, 2011), but Turkey took part (Akinci, 2011). This is an example of mixed evidence as to how significantly domestic conditions matter in a decision about coalition participation. The case of Germany suggests that domestic politics might prevent a state from joining multinational forces, but the Turkish example implies that domestic political opposition and an election cycle will not prevent a country from joining in a coalition.
Such a mixed example contributes to the recent academic trend of understating the effects of domestic factors on coalition participation. For instance, most large-N studies of the determinants of coalition participation pay close attention to international (both systemic and dyadic) factors and deny the systematic effects of domestic conditions (e.g. Lake, 2009; Tago, 2007; Vucetic, 2011). These studies argue that the major powers, states with large military capabilities, states in the same geographical region of country to which the coalition is deployed, English-speaking countries and states at the high end of the American security/economic hierarchy are more likely to join in a coalition led by the USA. Furthermore, the coalition’s type and purposes significantly affect the number of participants. A coalition with UN Security Council authorization attracts more states, but a domestic-intervention coalition reduces the number of participants. These observations are seemingly robust, and we now know that international, systemic conditions and dyadic attributions increase/decrease the chance of a state joining a coalition.
However, such a consensus among scholars might be due to an unfair selection of the proxies for domestic determinants of coalition participation. For instance, Tago’s article (2007) includes a single variable—“number of government crises”—and concludes that domestic conditions are relatively weak to explain who will join in a coalition military operation. This study, with a better theoretical argument and inclusion of proper proxies for domestic politico-economic conditions, reveals how they can affect the decision to send troops to a coalition-of-the-willing operation. Specifically, the article features the effects of the election cycle, domestic political instability and a recession.
The scope of the study is limited to the coalitions led by the USA after the Second World War, because the existing consensus about the dominance of international motivations is empirically supported by the data analyses of US-led, multinational, military operations after the Second World War. It is possible to include non-US-led coalitions and/or to expand the temporal domain of the empirical research to pre-Second World War periods; however, I leave that for future research. 1
The remainder of this study is organized as follows. The next section surveys the current understanding of the determinants of coalition participation. The third section elaborates how domestic political processes are related to the likelihood of joining a coalition. The fourth section elaborates the research strategy of the paper by explaining the dataset and statistical methods, and the fifthe section looks at the results of the statistical analyses. In addition to reporting estimation results of various regression analyses, the section reveals why contrasting decisions were made by Turkey and Germany at the formation stage of the 2011 Libya no-fly-zone operation. The final section concludes the article by suggesting that a state cannot help others when there are too many “home problems”, that is, when it faces critical political or economic challenges.
The current understanding of coalition participation
Coalitions are created in times of interstate wars (Powell, 1999; Starr, 1972; Yamamoto and Bremer, 1980) and serve as tests of alliance reliability (Gartzke and Gleditsch, 2004; Tago, 2009). They are also formed for intervention into domestic turmoil (e.g. coups) and for peacekeeping after prolonged conflicts (Tago, 2007). They involve short-term, ad-hoc and purpose-oriented international security cooperation. Coalitions are thus different from formal alliances (e.g. the North Atlantic Treaty Organization and US–Japan Security Treaty) and security organizations (e.g. the United Nations).
Numerous scholarly efforts have used case studies to identify the determinants of coalition formation and participation (Bennett et al., 1999; Davidson, 2011; Lake, 1999; McDermott and Skjelsbaek, 1991; Pierce et al., 2002; Saideman and Auerswald, 2012). For instance, Bennett et al. (1999) examined the 1990–1991 Persian Gulf War coalition and argued that a coalition was formed primarily for the purpose of burden sharing. They concluded that the coalition partners were motivated by international factors, such as dependence on a US alliance and the perception that the Iraqi occupation of Kuwait was threatening to them. They believed that domestic factors could only affect the nature of a state’s contribution by sending military troops or providing financial support. David Lake (1999) similarly concludes that division of labor is a key mechanism of coalition formation and that international determinants are more prevalent than domestic ones. Furthermore, by studying the cases of the Kosovo, Afghanistan and Iraq coalitions, Davidson (2011: 22) argues that coalition participation is mainly determined by international factors such as alliance value, the level of target threat to national interest and the state’s prestige. Public opinion may matter only when the international factors are split (i.e. they lead to opposite policy implications) and public opinion has an electoral relevancy and opposes coalition participation. Saideman and Auerswald argue that the domestic, institutional difference between coalition governments and presidential/majoritarian parliamentary governments better explains the variance in national restrictions on NATO’s mission in Afghanistan than other factors, such as threat-balancing, public opinion and the national culture about the use of force abroad.
In addition to this case-study-oriented research, there are a number of large-N data analyses of coalitions. Tago (2007) is one of the first of these studies, but the scope of the research is limited to coalitions formed by the USA after the Second World War. He created a dataset of participants of US-led coalitions from 1950 to 1999 (i.e. from the Korean War to the Kosovo crisis) and conducted a statistical analysis of which states were likely to join a coalition. The study found that factors at both the system level (geography and major power status) and dyad level (alliance relationships and cultural ties with the USA) are significant determinants of who enters a coalition. An important finding of the study is that a coalition’s legitimacy (its purpose and external authorization by the United Nations) increases the number of states in a coalition.
David Lake (2009) uses the dataset by Tago (2007) to support his theory of hierarchy in international relations. His analysis reveals that location at a high level in the security and economic hierarchies drives states to join US-led coalitions. The finding suggests that the level of subordination to the leader state matters. Srdjan Vucetic (2011) expands the same dataset to support his argument about “Anglosphere” politics and finds that Anglo-Saxon states are more likely to join in a coalition led by the USA. Although it is not a large-N study in a strict sense, Anders Wivel and Oest Kajsa Ji Noe (2010) focus on the coalition (non)participation of 11 Pacific microstates and identify factors that motivate a microstate to join a US-led coalition. After supporting the basic findings of other studies, they add the perspective that small states are motivated by the lessons of past security challenges—another international-level factor.
By including broader cases of coalition military operations, Pilster (2011) conducts multidimensional analyses of coalition military operations by focusing on their formation and duration (i.e. termination). His research provides empirical evidence for the theoretical expectation that democratic allies make more attractive coalition partners (and that nondemocratic states do not necessarily attract other nondemocratic partners). The focus on regime type is a very important and unique contribution of Pilster’s study; however, it fails to include domestic, politico-economic variables other than regime type to explain coalition formation/duration. Bove and Elia’s (2011) study about participation in peacekeeping operations focuses on determinants other than domestic, politico-economic conditions; indeed, the unemployment rate (the only domestic politico-economic variable in their troop participation estimations) fails to bear statistical significance. Although state-level factors, such as a comparative advantage of military manpower, the number of multiple missions and the level of tolerance for casualties, significantly affect the probability of sending troops, Bove and Elia also fail to reveal how domestic politics and the economic situation of each state would affect the likelihood that it would contribute troops to a coalition operation.
The existing studies of coalition military operations concur on the importance of international-level determinants; systematic and dyadic conditions basically predict who will be a member of a military coalition. In contrast, domestic factors are believed to be insignificant. However, previous studies have failed to study thoroughly the effects of domestic conditions. For example, Tago (2007) includes only one domestic, politico-economic variable: Government Crises—the level of domestic turmoil in a state when a coalition is formed. 2 Thus, it is only fair to conclude from the Tago study that one domestic factor—a situation that threatens the downfall of an existing political regime—does not significantly change the probability of coalition participation. Needless to say, there are more domestic political/economic conditions that could affect the decision-making process about coalition participation, and such factors need to be fully included in a statistical analysis. Otherwise, it is impossible to conclude that only international (system level/dyadic level) determinants are powerful predictors of coalition membership.
Among various, politico-economic, domestic factors in this study, systematic effects are featured, including election cycle, domestic political vulnerability and conditions of national economy. Scholars of international relations have extensively studied their connection to international conflict behaviors through the mechanism of diversionary effects (Chiozza and Goemans, 2011; Davies, 2002; Fordham, 2002; Huth and Allee, 2002; Kisangani and Pickering, 2009; Leeds and Davis, 1997; Sirin, 2011). The next section shows the theories and testable hypotheses about domestic determinants of coalition participation.
Deciding coalition participation and domestic political economy
Although this study focuses on the same three domestic, politico-economic conditions that diversionary theory does, the existing arguments on diversion cannot be used to explain the connections between coalition participation and domestic conditions. Joining a coalition led by some other states and starting a militarized interstate dispute to defend vital interests are quite different political processes.
On the one hand, international conflict behaviors (i.e. militarized interstate disputes), which are the central focus of diversionary theory, are often conducted unilaterally (Tago, 2005). There is always a clear enemy (a target state) and the military actions are justified as a legitimate act of protecting vital national interests. On the other hand, coalition participation, by definition, is always part of a multilateral military operation. It is usually requested and invited by a leading state (or multiple coalition leaders) that is more interested in the mission’s accomplishments and thus commands the multinational operation. 3 In contrast to the leading state(s), for every coalition partner, there is an incentive to be a free-rider (to attempt to send as few troops as possible when deciding to take part in the coalition) because the operation may not be in the vital interests of the partner state (Bennett et al., 1999).
Moreover, coalition participation is usually motivated and justified as an activity for fulfilling a state’s responsibility as a member of the United Nations/a regional organization/a defense alliance treaty and maintaining international (regional) peace and security (Tago, 2007). It would thus rarely generate a rally-around-the-flag effect for a political leader. For instance, combat casualties among the national contingent in a coalition operation could draw significant domestic attention (and thus “divert” public attention from domestic political and economic challenges to the coalition’s use of force). Yet, it would rather reduce public support for a political leader who has decided to send troops to a dangerous mission (Davidson, 2011). Successful coalition participation without significant incidents would, in contrast, not be big news to the domestic audience, because the vital national interest is not at stake and the coalition action is often far from home. Thus it is less likely to gather as much media attention as a militarized interstate dispute that was started for the defense of core interests and often occurs in contiguous areas. It is less likely that a coalition’s use of force will generate rally effects for political leaders of the participating states. 4
Table 1 summarizes the key differences between the two types of use of force. This is not an exhaustive list, but it shows how the two types are distinctive and why the implications of the existing studies on use of force are not applicable to a study of coalition military operations. A new theoretical framework and a set of hypotheses about domestic determinants of coalition participation are required.
Types of use of force and key features.
Note: Use of force to protect the core national interest may not always be justified as “self-defense”.
In this study, I argue that domestic, politico-economic factors rarely become motivating forces for coalition participation but rather function to negate the positive inducements of international conditions to encourage coalition contribution. Conditions of domestic politics and economy “cancel out” the motives for sending troops for a coalition. Here, I make two arguments based on (a) the information asymmetry on coalition participation between a political leader and the general public (related to an election cycle) and (b) the domestic challenges that limit the resources for coalition contribution (related to violent domestic unrest and economic recession).
Domestic dissatisfaction with sending troops for a coalition
A military coalition is usually led by and formed for a leader state. 5 The leader state pays the largest portion of the cost of collective military actions and calls for other states to join in a coordinated action. There are explanations for why the leader state needs coalition partners (Tago, 2005). Burden sharing is one of the well-known motives for the leader to ask others to send troops. Collective legitimization is another motive to form a coalition.
Whatever motives exist for a leader state, coalition partners also need motives for joining a military action. For instance, a nonvital but common desire with the leader state to achieve the military goal of the coalition is the most intuitive motive for joining a coalition. Indeed, existing large-N studies insert variables such as geographical proximity and major power status to reflect the degree of shared interest with a coalition leader state for sending troops (Lake, 2009; Tago, 2007; Vucetic, 2011). A fear of punishment by a leader state might also be a driving force for a potential contributor to actually send troops (Lake, 2009; Tago, 2008). If surrounded by enemies and with a coalition leader state as its only close ally, a state would follow the leader’s request for coalitional troop deployment. Reward of military/economic aid may be important as well (Lake, 2009). Although Tago’s (2008) data analysis rejects the generalizability of the argument, there are many anecdotes about aid-for-coalitional-contribution (e.g. Bennett et al., 1999; Woodward, 2004). For instance, US aid to South Korea during the Vietnam War is widely known. Arab countries obtained various rewards for their participation in the coalition war against Iraq in 1991. Furthermore, coalition participation might result from a desire for legitimization by the United Nations and other international organizations (Tago, 2007). Contributing to an authorized operation would help a state to build a good reputation as a responsible member of international society. To avoid being denounced as an irresponsible free-rider, a state would be more willing to take part in a coalition with collective authorization.
Information about how troop contribution to a coalition would benefit a state is complex and, in most cases, available only to the members of the decision-making circle. The political leaders know the existence of common interest in a coordinated military action; they recognize how serious a possible punishment by a coalition leader state would be if they fail to contribute troops. They also know how much military/economic aid they can obtain in exchange for troop contributions if they obtain such a promise from a leader state. Furthermore, they are deeply concerned about national prestige outside the country and are eager to participate in a coalition that has been legitimized by international organizations. In sum, they can access plenty of information and thus understand the importance of sending troops to a coalition.
In contrast, the domestic audience does not fully understand why a state must send troops to a coalition. They would recognize the need of sending troops if a coalition were formed for their vital national interest (e.g. for defense of their territory); however, on most occasions, a coalition is formed to defend the nonvital interests of contributors. The domestic audience would not see the common interest in allocating resources, would not know how punitive a leader state could be and would not know that a state could obtain aid from a coalition leader state in exchange for sending troops. Thus, they prefer to free-ride and reduce the amount of defense resources they provide for the interests of another country. For instance, the USA initiated the “many-flag program” during the Vietnam War and asked allies and friendly countries to send military and economic assistance to South Vietnam. Diplomatic cables at the time reveal that potential coalition partners found it difficult to commit to the program because they could not easily persuade domestic audiences to send resources (especially military) to defend their nonvital interests. 6
Information asymmetry inevitably exists between a political leader, who decides on coalition participation, and the domestic audience, who can decide if the incumbent leader should remain in the office, based on his/her performance record. In most cases, political leaders are aware of the benefits of troop contribution and are inclined to join in a coalition simply because they know what they can get from doing so. However, domestic audiences never fully understand the value of the allocation of defense resources to a seemingly unimportant multinational operation led by and for other states.
Political leaders could use their time and resources to explain to the public the benefits of sending troops; however, it is not so easy to obtain public support to send troops for nonvital interests and to risk soldiers’ lives for another country. In many cases, promises of aid, which might be the key information necessary to persuade a reluctant domestic audience of potential contributor state, cannot be publicly revealed because a coalition leader state could be accused of “buying troops”. Thus, the information on potential aid in exchange for troops remains secret in most cases. Moreover, the domestic audience might be wary of the government’s explanation that a lack of contribution to a coalition might result in a harsh punishment from a coalition leader state and cause its bilateral relationship to deteriorate. In short, the government cannot effectively explain why troops must be sent abroad for another country’s military action.
I argue that the information asymmetry about coalition participation would be especially problematic during national elections. Elections serve as the best opportunity for people to express a (negative) evaluation of the policies of the current administration. After the government decides to send troops to a coalition operation, the voters can express their disagreement with the government’s decision by casting a negative vote for a leader or a leader’s party. Moreover, the opposition parties can take advantage of the information asymmetry and criticize the government for the decision to participate in a coalition. The opposition’s attack on a policy to contribute to a coalition would be sharper before an election because of the desire to appeal to and obtain the votes of a large domestic audience. I can thus hypothesize that a political leader will be extra cautious and avoid a decision to participate in a coalition at the time of a forthcoming election.
In a similar vein, it can also be hypothesized that leaders of countries with possible troop contributions will feel relatively comfortable sending military forces to a coalition when the decision is made right after a national election and when there will be considerable time until the next election. A lack of an election means a lack of a factor to prevent coalition participation. Those two hypotheses must be tested respectively.
A lack of resources owing to problems at home
Coalition participation requires the unexpected allocation of military resources to join a multinational mission. Unlike the UN peacekeeping operations in which developing countries can obtain a subsidy from the UN peacekeeping budgets for their contributions (thus causing “the mercenarization” of the troops) (Bove and Elia, 2011), coalition participation does not come with compensation for the cost of troop deployment. A coalition leader state or other wealthy nation in support of the coalition might partially cover the costs of troop contribution (e.g. as during the Persian Gulf War), but this is uncommon (Tago, 2008). A coalition partner must bring its own resources to join a coalition.
There are two kinds of resources that must be prepared for coalition participation: military manpower and a special budget for sending and maintaining troops abroad. To join a coalition operation, the military personnel must meet international standards to work with other states’ national contingents. For non-English-speaking countries, to join a US-led coalition, at least the officers must be fluent in English to smoothly communicate and successfully coordinate with the counterpart US forces. Moreover, states prefer not too send troops with limited capabilities to avoid being blamed by other states for failing to function well in the coalition forces. This means that the military manpower allocated to a coalition must include elites in the military with good training and equipment.
Preparing the military manpower is obvious, but there must also be a special national defense budget to enable the military forces to travel to the coalition theater zone and participate in an operation for a certain period of time. Because coalition participation is not predicted unless it is a continuation from the previous year, a state must find a new funding resource in the middle of a fiscal year for the military operation. A government may have an emergency or reserve fund for such an operation, but it may not be large enough. If funds are limited and the government must pass a supplementary budget (particularly in democratic states), leaders have to convince parliamentarians and the general public that the operation is necessary and military resources must be expended. Even if the state is an autocratic regime and the political leader has a relatively free hand over budgetary processes, the money for a coalition contribution still has to be prepared by a budget-transfer from somewhere in the national budget. This might cause intergovernmental struggles over which department gives up a portion of its budget for the year to pay the costs. 7
This study argues that both resources might be scarce during periods of domestic political instability and economic recession. Political instability makes it difficult to prepare military manpower for a mission, and economic recession prevents the government from allocating enough money for the mission.
Domestic political instability, especially during violent disruptions of law and order, keeps the government from sending troops abroad. At such a time, a political leader needs the military to maintain domestic order. The military is a critical asset for the government to prevent riots from spreading. The national police can usually contain most violent movements, but in many countries, the military serves as the last defender of domestic tranquility. For instance, the US government recognizes this, and provides military aid for self-defense and to contain internal disorder. Even among developed nations, violent riots are more commonly deterred by the utilization of military forces. For example, the British political leadership was seriously considering use of the military to counter nationwide riots in 2011. This is more likely to occur in authoritarian and military regimes, but the use of the military inside a country is not uncommon, even among democracies, to restore peace during severe domestic unrest. Therefore, regardless of regime type, military forces cannot easily be moved outside of the state when there are systematic, violent movements against the government.
A violent disorder is different from a government crisis, which was inserted into the regression model of previous large-N studies and was supposed to be unrelated to the probability of coalition contribution. In a government crisis, the country faces a rapidly developing situation that threatens to bring the downfall of the political regime. Thus, it might occur in democratic countries at times of key elections without any violent, domestic unrest or danger of losing law and order. 8 A violent domestic disorder, in contrast, would be better included in a statistical estimation of coalition participation, because it covers public unhappiness with the government and public use of violent protest—a situation in which the military might have to be used to maintain/recover domestic law and order.
Economic conditions cannot be ignored because they affect how easily a government can prepare funding for a coalition contribution. Faced with a booming economy, a government can be more optimistic about preparing a military budget to enable coalition participation. It can rely on increasing revenue from taxes, and there is no need for stimulus policies; the government can use the money for extra activities such as a coalition contribution. However, faced with a tough economic situation, it would be difficult for a government to send troops abroad to help other states. The first priority for a political leader is to reverse a recession, and a declining budget must be used for economic stimulus policies. Coalition participation would be a very low priority to boost the economy, because most of the funding would be used for payments for military personnel, transportation and equipment. There are better targets for the allocation of government resources to recover economic growth than sending troops to foreign soil.
Moreover, even if the political leader knew that troop deployment would not really harm economic policy because the operation was limited in scale and duration, the domestic audience would not be likely to come to the same conclusion as the political leader. At the very least, the opposition parties in a democratic regime would use this to criticize the incumbent official. It would be more difficult for democratic leaders to pass a supplementary budget to send troops to a coalition in the face of such a strong attack from the opposition parties. Even in autocratic regimes, although the budgetary process is dominated by the political leadership and may not need parliamentary approval, the government must satisfy a winning coalition by allocating resources to them. During an economic recession, resources are more likely to be scarce. This suggests that autocracies would also like to avoid a coalition use of force, especially during a flagging economy since they lack economic resources for maintaining the system through patronage politics.
Data and research design
This study utilizes the frequently replicated dataset by Tago (2007), which covers post-Second World War coalitions led by the USA. 10 In the original dataset, there are 2100 observations; however, owing to limited availability of domestic indicators, it was necessary to drop the coalition operations before 1960. The data cover eight coalitions (7.2 coalition operations on average for a state) from the 1982 Lebanon multinational forces to the 1999 Kosovo Crisis (a NATO operation), and there are 1156 observations (161 states at maximum). The unit of analysis is country-coalition.
This study creates the following equation to estimate:
where Y*ic is a latent continuous response (coalition participation) for each individual state i in US-led coalition c. In addition to three types of domestic condition variables (election, political disorder and economic conditions), the vector Xic represents coalition-varying characteristics measured at the individual state level (control variables, such as regime type, major power status, military capability and alliance status), and the vector Zc captures coalition-fixed-effects, such as a status of collective legitimization; ξic is a residual.
There are two election variables in this analysis. The variable Election within 1 year corresponds to hypothesis 1a and is coded a value of 1 if there is a national election within a year after the coalition was formed and started its operation; otherwise coded 0. 11 For hypothesis 1b, the variable Right after the Election is created. A value of 1 is coded if a national election is conducted within a year before a coalition is formed; otherwise it is coded 0. Election data were provided by the “Institutions and Elections Project” (Regan et al., 2009).
To test hypothesis 2, the variable Riots, which counts the number of instances of violent civil unrest annually, is included as a proxy for the level of domestic political disorder. Based on the widely used data compiled by Arthur Banks’s Cross-National Time-Series Data Archive, 12 the number of violent demonstrations/clashes of more than 100 citizens involving physical force per year is coded. Tago (2007) uses the Banks’s Govcrises variable, which counts the number of government crises, as a proxy for domestic political challenge. However, the variable would not be the best proxy for measuring political conditions where the military forces must be mobilized to contain a violent internal unrest. Indeed, the correlation of the variables Riots and Govcrises is quite low (0.11).
As a variable of economic conditions and to test hypothesis 3, the variable Recession is inserted into the regression. Based on the annual gross domestic product (GDP) growth rate recorded in the World Bank Development Indicators 2001 database, a state-year with negative growth rate is identified. The variable Recession obtains the value 1 if the growth rate is smaller than −2.0%; otherwise, the value 0 is assigned. 13
For control variables, there are eight indicators. The variables Major Power Status, Military Expenditure, Geographical Proximity, Core Anglosphere States, Democracies and Alliance Tie with the USA are coalition-varying characteristics measured at the individual state level (Xic). I follow Tago (2007) for coding the status of major power. Besides the Union of Soviet Socialist Republics and China during the Cold war, a state that was recognized as a major power in a given coalition case is coded 1; otherwise, it is coded 0. Military Expenditure represents military capability of a state in the year when a coalition is formed. Annual military expenditure data in the Correlates of War national capability dataset are used. 14 Geographical Proximity is a dichotomous variable that records geographical contiguity. A state that shares the Correlates of War region code with the coalition’s operation zone is coded 1. Core Anglosphere States is a variable to control for Vucetic’s (2011) use of the four non-US members of the Anglosphere “core”. 15 A value of 1 is assigned for Canada, the UK, Australia and New Zealand; otherwise a value of 0 is given. As a proxy for regime type, I include the variable Democracies, which is measured using the Polity IV democracy score (a value of 7 or higher is considered a democracy) for the year coalition forces are formed. A democratic state is coded 1; otherwise, it is coded 0. 16 Finally, to code Alliance Ties with the USA, I used the Correlates of War Formal Alliance dataset. A value of 1 is coded for a state that was a formal ally of the USA by signing a defense pact when coalition forces were formed. 17
The indicators Coalition for Domestic Intervention and UN Collective Legitimization are binary variables that capture the coalition-fixed-effect (Zc) of a coalition’s purpose and the status of legitimization of the military operation. 18 If a coalition is created for a domestic intervention, the variable is coded 1; otherwise 0. Collective legitimization must be the formal decision of the UN Security Council. The resolution or equivalent statement of the council that provides the USA with the legal and/or political foundation for military operations must be adopted before the formation of a coalition.
In selecting estimation models, I follow what Lake (2009) has done in his replication of Tago’s (2007) coalition participation dataset and thus use (1) a random-effect, logistic model—(Generalized Linear Latent and Mixed Models: GLLAMM), 19 (2) a conditional (coalition fixed effect) logistic model, and (3) a logit model with coalition case dummy variables.
Empirical results 20
Table 2 shows three different estimation results depending on which regression model is used. Table 3 reports substantive estimates—the predicted values of change in selected independent variables of model III, based on simulation using the Clarify software (King et al., 2000; Tomz et al., 2003). Since all of the hypotheses are directed, the one-tailed test is used here.
Estimation results.
Note: *** statistically significant at 0.025 level; ** statistically significant at 0.05 level; * statistically significant at 0.10 level (one-tailed test). STATA12.0 is used to estimate.
For model III, all estimates for the case dummies are not reported.
Simulated predicted values of selected independent variables (model III).
Note: Clarify (simqi, fd(pr)) is used for the calculation.
Baseline probability for coalition participation is 0.03 [0.02, 0.04].
The estimates support hypotheses 2 and 3, but they reject hypotheses 1a and 1b. 21 The domestic resource suppression argument is supported, but the information asymmetry explanation fails to be supported. Domestic political disorder significantly reduces the probability of coalition participation. Results show 14% lower odds of sending troops for multinational military operation per one riot incident [exp(−0.146) – 1 = −0.14]. In a similar vein, the likelihood of coalition participation declines when there is an economic recession, with 55% lower odds [exp(−0.788) − 1 = −0.55]. In contrast, no matter how the election variable is coded, it fails to bear statistical significance. Surprisingly, a state within a year from a national election is more likely to send troops (positive coefficients); this is contrary to what is theoretically expected.
It must be emphasized that inserting different domestic political challenges variables, such as the number of government crises, peaceful antigovernment protests and general strikes does not affect the results. It is only violent domestic unrest—an increase in domestic demand of the military for recovering law and order—that leads to a significant decrease in the probability of coalition participation.
A further discussion is needed to explain why the election cycle does not have a significant effect in the model. Logically, there are at least two possibilities. First, the election variables in this analysis may not be the best proxies. It is possible that only some (subset) elections matter (and we should have inserted a differently coded independent variable to observe a significant result). However, this is less convincing because alternative estimations with the varieties of election variables (different coding rules, interaction terms, etc.) were conducted to check the robustness of the analysis, and no election variable bears statistical significance. There is no way to be exhaustive, but it is very unlikely that the election cycle systematically changes the probability of coalition participation as theoretically expected. Second, coalition participation may not often be on the agenda for an election among political parties. Although there is evidence that a strategic position-taking at national elections can be observed at troop withdrawal phases from a coalition operation with substantial cost for a state (Tago, 2009), the same mechanism may not work at the coalition participation stage. The domestic audience is not concerned about sending troops when the coalition starts, but they care a great deal if there are (actual/a high risk of) combat casualties and prefer to cease contributing to the coalition. Coalition participation policy is thus more likely to be politicized over troop withdrawal at national elections.
Table 3 shows that the substantive impacts of domestic variables are relatively small compared with international determinants. In this regard, previous empirical studies are accurate in emphasizing the international systemic and dyadic conditions that change more of the odds of coalition participation. However, this does not mean that we can forget about domestic conditions. Among the negative determinants of coalition participation, the recession variable has slightly less but comparable impact to the domestic intervention variable.
Table 3 shows that the power of the UN’s collective legitimization is superb. If an authorization of the multinational use of force is given by the UN Security Council, the probability of coalition participation increases by 16%. A lack of this factor could not be made up even if we combined two of the other powerful international variables. The effect of the collective legitimization of the United Nations would be overwhelming only if the following three conditions were met simultaneously: the state had major power state status; it was in a region that was geographically close to the theater of coalition operation; and the state was one of the core Anglosphere nations. The UK is the only country that fills this profile when a coalition that is led by the USA is formed in Europe (i.e. within the region). Collective authorization by the UN Security Council is very important for the US government if it wants to have more coalition partner states. A domestic variable (recession condition) contributes at most one-eighth of the substantive impact of the collective legitimization variable.
Finally, I would like to make an argument about the difference between the decisions made by Turkey and Germany in March 2011, the case mentioned in the Introduction. In short, an election, which was mentioned by the media as a reason for Germany’s failure to join the coalition, was not an important factor, as the large-N analysis suggests. Given the estimation results, I would argue that economic conditions must be the key to explaining the difference in the decisions of Germany and Turkey. Figure 1 shows the predicted probability with different values of the related key independent variables (i.e. power status, UN collective legitimization and economic condition), based on model I. Table 4 reports the predicted values of specific conditions of quasi-Turkey and quasi-Germany at the time of the 2011 Libya non-fly-zone coalition formation. 22
Predicted probability for Turkey and Germany in March 2011 (model III).
Note: Germany’s GDP growth rate was not negative by the International Monetary Fund’s prediction for the first quarter of 2011. However, according to its World Economic Outlook Projections, Germany lost 1% of the score in 2011 (prediction) from 2010, whereas France and the UK gained 0.1–0.4% in the same data. Clarify (simqi, fd(pr)) is used for the calculation.
Box plot of predicted probability of selected independent variables (model I).
If each state were facing counterfactual conditions in March 2011, the two countries could have acted in quite opposite directions. That is, during an economic recession, Turkey could not have sent troops with only 9% of the predicted probability, and under good economic conditions, Germany with 40% of the same probability could well have joined a coalition. 23 However, in reality, as Table 3 shows, in March 2011, Turkey and Germany were under conditions that led to almost the same predicted probability (21–24%) of coalition participation. Because Germany was experiencing economic difficulties resulting from the European sovereign debt crisis (in contrast to France and the UK, Germany was the only major European power with a negative economic outlook by the International Monetary Fund at the time), it could not send troops to Libya. It may not have been an election but a flagging economy that led to Germany’s nonparticipation decision. Turkey was enjoying a booming economy with more than 9% of GDP annual growth rate (Bryant, 2011), so the political leadership could send troops even though they knew there was domestic opposition to a military contribution.
Conclusion
Decisions about coalition participation are not independent of domestic political and economic processes. This may not be a wholely new finding and, in fact, it has been repeatedly pointed out by case studies; however, it is a novel contribution of this article to support it by statistical analyses and to sort out which domestic determinants matter and why. Domestic violent disorder and economic difficulties prevent a state from sending troops to a coalition military operation. Election cycles and peaceful domestic protests do not significantly affect the chances of coalition contribution. When a state is facing a significant challenge—in either domestic law and order or domestic economy—the state will hesitate to deploy its military forces for a coalition operation. When there are enough political and economic problems at home, a state cannot really help someone else, even if the help is requested enthusiastically.
Previous statistical analyses of military coalition participation have neglected domestic factors. This is somewhat understandable since the marginal effects of the international conditions are so strong. However, this article shows how domestic politics and economic conditions do affect the likelihood of coalition contribution. The marginal effect of the variable is substantive and cannot be ignored. The two domestic conditions—domestic unrest and economic recession—cancel out the international incentives for sending troops to a coalition, as Germany’s case in the Libyan no-fly-zone mission in March 2011 suggests.
Further research is needed to fully unveil how domestic, politico-economic conditions systematically affect the likelihood of coalition participation. This study could not test all of the possible hypotheses of domestic conditions. For instance, the number of veto players could affect a decision to use force abroad, or a domestic political institution could affect significantly the likelihood of coalition participation (Saideman and Auerswald, 2012). Many follow-up studies will be needed to find such a systematic connection between domestic factors and coalition participation. Moreover, it is important to expand the data to include coalitions led by states other than the USA and to cover time periods before the Second World War.
Footnotes
Funding
This study is supported by the JSPS KAKENHI Grant Number: 24683004.