Abstract
Building upon research that found that coalitions are more likely to win wars, recent work has sought to differentiate effective from ineffective coalitions. Much of this work focuses on characteristics of member states and not the coalition itself. This paper takes a first step in exploring how the structure of a coalition contributes to its performance. Specifically, coalitions vary in how much control members must transfer to the coalition. Some coalitions form weak command structures with states maintaining primary control while other coalitions form a strong centralized command. The impact of command structure on coalition performance is vital to understanding the success and failure of coalitions. Highly centralized command structures allow states to overcome the problems associated with coalition warfare and achieve victory. Empirical evidence supports this claim; the odds of victory increase as states surrender more control to the coalition. These findings provide direct advice to policy makers considering forming a coalition. The implications of this research extend to the alliance literature and the war bargaining literature, which tend to assume either that wars are dyadic or that states can seamlessly aggregate military capabilities.
In May 1948 a coalition of five Arab armies invaded the new state of Israel. The 1948 War, along with its sequels, provide students of wars with repeated examples of the weakness of coalition warfare. The state of Israel was just coming into existence and had little in the way of an organized military and even less modern equipment (Ilan, 1996). The coalition was the initiator of the war and enjoyed a sizable edge in military power. Despite these advantages, the coalition was outfought and defeated. The 1948 War highlights the puzzle of coalition success. Why do some coalitions fail, even under seemingly ideal conditions, while others achieve victory against the odds?
There have been many recent advances in the study of coalition warfare; however, we do not yet know why some coalitions are successful while others fall apart and are defeated. This paper takes a step in exploring how the structure of a coalition contributes to its performance. Structural characteristics that make it harder to achieve unity of effort weaken a coalition, while characteristics that increase unity of effort lead to synergies and a greater probability of victory (Marshall et al., 1997; Silkett, 1993). Coalition forces must be able to adequately aggregate their military efforts in order to realize the benefits of fighting together. While there are numerous design choices that could play a role in coalition victory (Weitsman, 2013), this paper focuses on the command relationship. There is a great deal of variation in the amount of authority states surrender when joining a coalition. Weak command structures lead to poor aggregation of military effort while strong command structures lead to unity of effort and maximize the aggregation of force applied against a target.
Research into coalitions has proceeded along different lines. One line focuses upon the pre-war phase, exploring the topics of coalition formation (Henke, 2017, 2020; Kreps, 2011; Tago, 2005; Wolford, 2015; Wolford and Ritter, 2016) and different bargaining dynamics (Wolford, 2014b, 2015). A second line explores the behavior and outcomes associated with coalitions during a conflict. Tago (2009), Pilster et al. (2013), Weisiger (2016) and Mello (2020) examine when states abandon coalition partners while Wolford (2014a, 2015) looks at when a coalition faces counterbalancing. This paper adds to the line of research seeking to explain how coalitions perform during war. Morey (2016) established that coalitions help states win wars. However, discrepancies remain in coalition performance; some perform well while others become ineffective. Pilster (2011) argues that democratic coalitions are more successful since democracies provide military leaders with the necessary discretion to conduct successful multinational operations. Opposed to this, Graham et al. (2017) argue that democracies form larger and more powerful coalitions enabling them to win. Weitsman (2013) focuses on the difference between allies at war and ad hoc coalitions, as well as the size of a multinational effort. Larger efforts increased challenges while “straightforward, hierarchical decision-making institutions fostered effectiveness” (Weitsman, 2013: 190). Cranmer and Menninga (2018) push this research further, exploring the connection between how a coalition is constructed and the probability of success. They argue that coalitions vary based on levels of skill, coordination, and legitimacy and that these differences affect success coalition operations. This paper builds upon the idea that coalitions vary in important ways, especially in the area of coordination. However, instead of focusing on the qualities of the member states, this paper focuses on the structural characteristics of a coalition, specifically the command relationship, to understand the success and failure of coalitions.
The argument bears some similarity to command structure theory (Grauer, 2016), which argues that employing the correct command structure, determined by the specifics of the battlespace, allows the effective use of military capacity and increases the odds of victory. Grauer (2016) focuses upon battlefield command and the tactical employment of forces while the argument below is conceived at the operational level. The two approaches are not competitors, they simply approach the concept of military effectiveness at different levels of aggregation. This work also integrates with findings by Weitsman (2013) but incorporates different types of hierarchal command.
When most people think of military organizations, they normally envision strict hierarchical command structures with clear patterns of superior–subordinate relationships. While this view holds at lower levels of military organizations, it becomes more complicated when studying inter-branch relations within the same state and military cooperation between states. In these cases, inter-branch or interstate rivalries make it difficult, if not impossible, to form coherent command arrangements. Coalitions are not created equal and the command relationship is one important difference. Some coalitions overcome interstate squabbling and form an efficient command structure. In other cases, the jealousy or mistrust between member states is too great for high levels of command integration.
The argument in this paper is straightforward – the greater the level of centralized command within a coalition, the greater the odds of victory. At the base, coalitions must aggregate the military effort of multiple states. While a great deal of literature assumes states can automatically aggregate power, the process is complicated and involves many political and technical decisions (see Bensahel, 1999). It also requires guidance and constant refinement. Absent a command structure able to provide this guidance, even powerful coalitions can operate inefficiently and be defeated by an inferior foe.
The findings in this paper support the claim that strong command is essential for coalition victory. The more power member states surrender to a coalition, the better the odds of victory. That said, the findings also indicate that to truly gain the benefits of fighting together, states need to surrender a great deal of power. While there are marginal benefits from increasing the command authority of the coalition, the true benefit comes from the highest level of command integration. Policy-wise this means that states should only consider forming coalitions when they will be able to overcome the obstacles to forming a unified command.
This research contributes to the growing literature on coalition warfare and coalition success. The results also provide direct policy advice for states wishing to form coalitions to deal with future threats. The results also speak broadly to interstate cooperation in security affairs and the issue of capability aggregation prominent within the alliance literature. The implications from this study can also extend to war bargaining models with a new source of incomplete information or information asymmetry that could complicate negotiations between a potential target and a coalition.
Initiation, power advantage, and coalitions
There are three broadly accepted explanations for victory in war: initiation, power advantage, and democracy. Initiators, stronger states, and democracies have all been found to have an increased probability of winning. However, when seeking to explain coalition victory, initiation and power advantage do not perform to the same level as in bilateral wars. 1 Coalitions function differently than states fighting alone against a single opponent and require new theories to explain when they win. The answer in this paper is that the advantages from initiating or having a power advantage are not realized without a high degree of centralized coordination in a multilateral conflict.
Despite the strength of past findings and the general belief that initiating a conflict and holding a power advantage increase the odds of victory (Wang and Ray, 1994; Gartner and Siverson, 1996), neither initiation nor power advantage seem to play a strong role in explaining coalition victories. Looking at all wars since 1816 (Sarkees and Wayman, 2010), we see that the initiator wins approximately 59.6% of the time. Coalitions perform almost identically, winning 58.8% when they initiate. However, coalitions win 71.4% of the time when they are the target of an attack. Looking at all wars, the target only wins 32.4% of the time. On average, coalitions are more likely to win when they do not initiate and they are over twice as likely to win compared with all targets. In the end, it is not that initiating a war is a large disadvantage for coalitions so much as it does not appear to provide the advantage we find in bilateral wars. Coalitions can be very successful as the initiator or the target. Something other than the first mover advantage explains coalition success.
Looking at capability advantage, we also find that it does not play a simple role in explaining coalition victory. Figure 1 displays the win percentage for wars divided between bilateral (one state on each side) and multilateral (multiple states on one side) 2 over different levels of power advantage or disadvantage. 3 In the first column are all cases where a warring side was at a power disadvantage; they faced an opponent that was militarily stronger. Moving right across the table the cases shift to warring sides having a low power advantage up to a high power advantage. 4

War winning percentage and power advantage.
The results in Figure 1 make one thing clear: power advantages (disadvantages) do not play the same role in multilateral wars as they do in bilateral wars. Looking at wars fought under a power disadvantage, it is clear that in multilateral wars a warring side does better when compared with a state fighting alone. Fighting at a power disadvantage still reduces the odds of victory. However, the key point is that in a multilateral war a warring side at a power disadvantage is approximately 10% more likely to win than a state fighting alone under a disadvantage. The power disadvantage matters, just not to the same degree. In the low advantage column, we see the same pattern, only the relationship is stronger. Looking at cases of moderate or high advantage we see that instead of continuing to outperform states fighting alone, states fighting multilaterally now see a decline in the winning percentage relative to states fighting alone. Under a moderate advantage, states fighting multilaterally win approximately 55% of the time. This is less than when fighting multilaterally under a low advantage and more than 20% behind states fighting alone with a moderate advantage. Even with a large advantage, multilateral efforts are only successful in six of 10 cases. 5 This is approximately the same winning percentage as multilateral efforts with a low advantage. Further, states fighting alone with a high advantage win approximately 90% of the time, meaning that multilateral wars with a high advantage are 30% off pace. Surprisingly, the odds remain slightly better than a coin flip and approximately equal to the win percentage if the war was fought with a low advantage, while the odds of victory for a state fighting alone increase dramatically under either a moderate or high advantage.
From this discussion one thing is evident, power preponderance does not work the same way in multilateral wars as it does when a state fights alone. Multilateral efforts tend to outperform states fighting alone under a power disadvantage or a slight advantage; however, the relationship flips to multilateral efforts underperforming with moderate or large advantages. The monotonic relationship between power preponderance and victory found in bilateral wars does not hold in the multilateral context.
Why do initiation and power advantages serve as poor predictors for victory in multilateral wars? The problem is that most works take for granted the ability to coordinate military efforts in order to reap the rewards of initiation and holding a power advantage. This is at least partially driven by the fact that most works explicitly or implicitly assume war is bilateral, which would negate coordination problems.
A second problem is that when research expressly takes a multilateral view it often assumes states can nearly seamlessly aggregate their power. This is especially true in the alliance literature where it is common to assume that the strength of an alliance is the sum of the power of individual members. The process, and problems, of combining the power of individual states is largely ignored (see Bensahel, 1999: 15). Studying coalitions and coalition structure provides a remedy for these problems.
In order to resolve the disconnect between the theoretical view of warfare prominent in the literature and the actual historical record regarding coalition succes, we must realize that the level of coordination in multilateral wars is a variable, not a constant. Strong coordination leads to coalitions maximizing their potential as is assumed in the current literature. However, poor coordination leads to coalitions squandering their potential.
What is command?
The first step in understanding the role of command structure in coalition success is to understand what is meant by command. Within the military, command can be divided between strategic and operational control. While it is true that command implies the ability to give orders, the issue is complex and is about how far a commander has authority to issue those orders. Can the commander only set overall unit objectives (strategic) or can they micromanage by issuing orders regarding the strategies and tactics employed to reach the objectives (operational) (for more see “Multinational Operations”, 2013; “Command and Control for Joint Land Operations”, 2014). There are also issues regarding how much authority a commander has to subdivide national units or replace personnel holding command positions within coalition forces. Solving these politically dicey, but militarily necessary, issues is the first step towards coalition success.
States rarely, if ever, hand over complete control of their forces to a commander from another state. The military represents the physical manifestation of a state’s ability to protect itself and handing over complete control would leave a state vulnerable. Beyond the need for defense, political leaders are unlikely to surrender control over the military even under dire circumstances. Leaders do not want to be held accountable for the actions of their forces and for battlefield losses without some ability to control actions. Leaders want to maintain some control over what missions national forces conduct and the timing of those missions, and keep their forces from being divided (Durrell-Young, 1997: 30).
At a minimum, states maintain control over the administration of the army regardless of the coalition command structure. Regulations, staffing, and discipline are not transferred out of national hands. As an example, during the Second World War the UK and the United States created a single command structure and possibly the closest coalition relationship in history. However, throughout the war there were two distinct armies, one British and one American. At the area of theater or area commanders there was a great deal of coordination but each army maintained control over most aspects of their own personnel.
Strategic command of coalition forces is primarily the area of politics between coalition members. The objectives and goals, even large parts of the means employed, are determined by negotiations between members. Big picture issues of the coalition’s goals always remain with the body of coalition members. As with the general administration of forces, strategic control usually remains an issue between the member nations and is not surrendered when joining a coalition.
In a coalition, command refers to control over the employment of troops in combat. Once the goals are set and the troops committed, who controls the military mission? The issue of command is plagued by multiple meanings across militaries and international bodies. For example, NATO divides command authority into two categories with four total levels. Operational and tactical authority comprise the major levels. Operational authority concerns the overall mission and deals with assigning specific objectives to units within the coalition. Operational authority is subdivided between Operational Command (OPCOM) and Operational Control (OPCON). There are several nuances between OPCOM and OPCON, the important difference being the ability to “task organize” (Durrell-Young, 1997: 23). The authority to task organize means that a commander (OPCOM) can divide subordinate units to meet mission needs. The coalition commander has greater flexibility to employ forces in the most efficient manner. Tactical command covers control over forces within an assigned mission and is generally for a shorter duration. Similar to Weitsman (2013), the focus of coalition command is conceived at the operational level.
Levels of command
While military leaders recognize the need for a strict hierarchy of command, frequently this logic is not applied to coalitions. The establishment of the chain of command is often a highly political and divisive process. First, state leaders often push to have the overall commander come from their own nation. This creates conflict before the coalition is created which often cannot be overcome and members do not agree upon a single chain of command. Second, state leaders are often reluctant to hand over control of their military forces to an officer from another state. For both of these reasons, historically, coalitions have had many different command arrangements.
At one of end of the scale states transfer very little, if any, control over the state’s forces. This is the case in the Arab coalitions against Israel. While there is often an overall commander in name, the control remains with each state. The overall commander cannot issue orders directly to field commanders. Instead, orders originate from the home capital. I call this command structure an Independent Command. The members of the coalition form a joint plan of battle; however, the execution is left to individual members. Often, each state is assigned its own geographic region within which to conduct operations. States within the coalition work together in that they have a joint plan of attack but the integration and coordination are minimal as national command structures dominate operational decisions. Often the full force of the coalition is not brought to bear on the target. One state will launch an attack in one region while other states do not launch attacks in their zones. Thus, the target does not feel the full force of the coalition simultaneously and shifts forces to meet threats as they arise one at a time.
The next type of command is a Joint Command. 6 In a Joint Command there is one command authority at the top of the coalition; however, instead of one commander there is a committee. The command committee makes all of the military decisions and issues orders to the forces under their command. The size of the committee varies by coalition. In some cases, every state in the coalition has representation. This was the case with the coalition fighting against China in the Boxer Rebellion. In other cases, only a subset of states have representation. Usually this group includes the commanders from the most powerful states in the coalition. In rare cases only a group of two or three people comprise the committee. In the Gulf War there was a two-man committee. General Schwarzkopf shared command authority with Saudi General Prince Khaled bin Sultan (Schwarzkopf, 1992). The two determined military policy and planning for the coalition. After the plan was agreed to, each commander passed instructions to units under their control.
While there is variation in the committee, the key marker for a Joint Command is the group at the top. This group is more coordinated than a series of individual commands but still suffers from many weaknesses, including the need to form consensus which at the very least takes time. The coalition in the Boxer Rebellion would halt operations after most major events to determine the next steps. This led to internal debates leaving the coalition idle for long periods, allowing Chinese forces to withdraw and reorganize. Instead of quickly following a defeated foe, the coalition would halt and then resume operations after an agreement was reached, only to fight a prepared enemy (see Keown-Boyd, 1991). Beyond the time involved, the decisions committees reach are often sub-optimal. To maintain cohesion, the members must form a unanimous decision, not a simple majority. The coalition is bound by the least aggressive member, slowing progress and leading to a series of half measures instead of an aggressive push. In the end, a Joint Command presents a mixed picture. There is more coordination brought about from a single command structure; however, the negotiation and log rolling necessary to make decisions introduce new complications. Overall, the effect should be improved efficiency but not an overall efficient operation.
Unified Command represents the last command style. This is the Eisenhower model with all power flowing through single individual responsible for determining and coordinating the plan of the coalition. To be a Unified Command each member state must surrender a great deal of authority to the coalition commander. The commander must be able to send orders directly to all coalition units and have them executed. States do not have to surrender full control of their forces, but they must come under operational command of the overall commander. In some cases, states can limit the types of missions their forces can participate in. This was common in the Kosovo War. Large numbers of NATO troops and fighter planes were placed under General Clark’s operational command for the campaign against Serbia. However, many European states placed limits on the types of missions their planes would fly. 7 A major concern was avoiding civil casualties; many leaders would not allow their air forces to conduct missions with a high risk of collateral damage. There were also deep political concerns limiting national mission profiles. Many states had long-standing relations with ethnic groups that occupied parts of Serbia but were not part of the Serbian majority. In order to avoid harming these relationships, states would not allow their forces to operate in certain parts of Serbia (Clark, 2002; Lambeth, 2001). Outside these limitations, General Clark was free to manage the forces under his control.
As we move up the structural hierarchy of coalitions the level of centralization increases. Independent Commands have low levels of integration once the planning phase is complete. The states work together but there is no central force ensuring a focused effort. Shifting to a Joint Command mitigates some of these problems. One central body determines policy and works to coordinate coalition activities. While higher centralization is achieved it does come at the cost of greater deliberation and slower reaction times. At the top, the Unified Command provides the greatest level of coordination. One commander with operational control over the entire coalition provides a high degree of coordination and the maximum amount of focused effort.
What role does command play in victory?
While a strong command structure cannot solve all problems associated with coalition warfare, centralizing control can mitigate many of those problems. While there is no magic formula to ensure victory, it will become clear that we should expect coalitions to perform better, all else equal, the stronger their command structure.
The major hurdle for coalitions is to coordinate military actions to such a degree that the force brought to bear against the enemy is maximized. The ability to harmonize efforts across different coalition members allows for unity of effort (“Command and Control for Joint Land Operations”, 2014; “Multinational Operations”, 2013). Unity of effort means that separate parts of the same force (national or multinational) are working towards the same overall strategic objectives. Beyond the basic concept of working together, unity of effort implies task specialization. Disparate units are not seeking the same objective at the same time, which leads to confusion or different units working at cross purposes. Instead, unity of effort divides tasks among the various members and ensures the proper timing and coordination of operations in order to bring about the maximum effect and greatest probability of success. It is important to stress that these advantages can lead to better performance on the battlefield when coalition members are fighting together. However, the advantages go beyond the battlefield to the overall strategic operation of the war. Well-coordinated efforts place an opponent at a strategic disadvantage as they face attacks on multiple fronts and must divide their forces (Morey, 2016: 538). This leads to coalitions facing less resistance on a specific battlefield. The true advantage, however, comes in battles not fought as the target pulls their forces back to avoid fighting at a disadvantage or moves to terminate the fighting once the strategic situation becomes clear.
While the biggest advantage of stronger command is synergistic use of coalition forces, strengthening the chain of command has other benefits. First, one of the major criticisms of coalitions is decision by committee. As previously discussed, this approach to warfare can lead to delays in decision making as members negotiate and muddled plans of action form from compromises between states. However, this process of constant negotiation and compromise is eliminated, or at least greatly reduced, as a coalition forms a clear centralized command structure. The political goals of the coalitions may still be the product of intra-coalition debate, but the military policy to reach those goals becomes more rationalized as command authority is increasingly centralized. To achieve the maximum benefit, command would ideally be concentrated into one commander with the power to guide all military forces. Centralizing command increases the speed of military decisions and the coordination of the military effort. Thus, the full weight of the coalition is effectively concentrated, allowing maximum advantage. This is especially important when attempting to fight on multiple fronts.
Further, while transferring power to a single commander will not bridge the political differences between coalition members, a strong central command can mitigate some political infighting. During the Second World War the UK wanted to invade the Balkans with the goal of denying Soviet control over the area post war. The United States objected to the plan, arguing it was not directly aimed at harming Axis war-making capabilities. A large part of what ended the debate was Eisenhower’s refusal to allow any landing craft to be diverted from the invasion of Europe (Stoler, 2005). Eisenhower’s strong stance, made possible by his position as supreme allied commander, ended the debate. 8
The overall argument is that if missions are not sufficiently coordinated, the opponent of a coalition will not feel the full weight of the coalition simultaneously and will stand a better chance of shifting forces to deal with separate attacks, essentially defeating a coalition piecemeal. In essence, without effective command a coalition begins to function more like a group of states fighting a war in parallel (see Morey, 2016) with the advantages of the coalition under-utilized or wasted. However, as coalitions form stronger centralized command structures they should become more effective in wielding their power and achieve greater military success by concentrating and coordinating efforts. Stronger command structures help mitigate, at least in part, many of the weaknesses associated with coalition warfare.
None of the above should be taken to imply that centralizing command will be easy. While it may be difficult for states to form strong centralized commands, the strength of the command structure can be vital in determining coalition success. Coalitions with weak command forgo most of the advantages of fighting as a coalition; however, if states form a centralized command, coalitions can be an effective tool of warfare.
Centralization is not a panacea, distancing decision making from front line leaders can reduce effectiveness as front line leaders will often know the best way to employ their forces (see Chandler, 1962). This could be the case in an international coalition where a commander might not fully understand the strengths of forces from other countries. While these potential downsides are important, the theory presented here assumes that on average coalitions will perform better the stronger the command structure is. Further, this argument focuses on how states can build the most efficient coalition; however, it may often be the case that states intentionally select a lower level of coordination because of political considerations (see Joiner, 2019, on the security–politics trade-off). 9
Methods and data
The population of cases for this analysis is all Correlates of War (COW) interstate wars (Sarkees and Wayman, 2010). The unit of analysis is warring side; in multilateral wars states fighting on the same side are aggregated to form one observation. Thus, each war has two observations, side A and side B. There are two reasons for this design. First, using a dyadic or state unit of analysis exaggerates the influence of coalitions because each coalition will be represented multiple times. If a coalition composed of three states wins a war there will be three observations of coalition victory, when in fact there was only one; the same will happen for every coalition loss. This would magnify the impact of any coalition beyond its true importance and this bias would increase with coalition size. The design adopted here avoids this by scoring only one victory (defeat), meaning that a coalition can only win or lose once per war. Second, Poast (2010) provides evidence that breaking multilateral events into dyadic observations leads to biased estimates. The bias comes from measurement error in the explanatory variables since they represent the relationship between states A and B instead of the relationship between the coalition and state B. In the case of measuring the effect of power, the dyadic design tests the ratio of state A’s power to that of state B; however, in the coalition context the correct measure should be the ratio of the coalition to state B. Using warring side allows both problems to be avoided and unbiased estimation.
Following past studies (Reiter and Stam, 1998) I divide several large wars. This division helps accurately reflect the nature of coalitions in different theaters. In the First World War Germany fought alone on the western front, but in coalition with Austria in the east. The same pattern is visible during the Second World War, where Italian involvement focused on the Balkans and North Africa. Along with dividing the world wars and the Vietnam War, as done by Reiter and Stam (1998, 2002), I also divide the Korean War into two conflicts, pre- and post-intervention by China. With the entry of China the war became a contest between two coalitions. Further, the UN forces expanded post-intervention, altering the basic nature of the coalition.
In order to measure the outcome of each war I code the variable Victory. Victory is a dichotomous variable coded 1 if a side wins the war and 0 in all other cases. Data for this variable comes from the COW Interstate War Dataset (Sarkees and Wayman, 2010) and Reiter and Stam (1998, 2002). For the Korean War I code the first part as an allied victory and a draw for the conflict post-intervention.
The first step in testing the effectiveness of coalition command structures is to determine the population of cases. To identify cases of coalitions in international wars I rely upon Morey (2016). In order to count as a coalition states must coordinate their military activity during a war; fighting against the same target at the same time is not enough. This data relies upon the COW Interstate War Dataset to identify interstate wars.
In order to test Hypothesis 1 it is necessary to distinguish different types of coalitions, specifically the strength of command. In order to measure the effect of command structure on coalition victory I create the variable Coalition Type. This variable takes the value of 0 if there is no coalition for a warring side, 1 for Independent Command, 2 for Joint Command and 3 for Unified Command. To code this variable I consulted historical works on each coalition. Consistent with the discussion above, I mapped the body, or person, who had effective operational control over coalition forces. If command stayed with the national command staff the case is coded Independent, if command was shared by a committee it is coded as Joint, if a single person held operational control it is coded as Unified. In most cases, the command relationship was clear and easily coded. However, in a few wars many sources were necessary to understand the command structure. Most of these wars are smaller, relatively unknown conflicts without a great deal of documentation. Table A1 in the Online Appendix provides information regarding the type of each coalition. 10
In order to control for other important factors in war outcomes several variables are included in the model. First, initiating a war can provide an advantage (Wang and Ray, 1994). Along with arguments that initiators carefully select targets, states that initiate are able to attack on ground favorable to their forces and to strike before the other side is prepared. To control for these advantages, the variable Initiate is coded 1 if the party to a conflict is the one that first begins hostilities according to the COW dataset. In cases of multilateral wars, the side with the initiating state is considered the initiator.
To ensure that Coalition Type is not capturing the influence of enhanced military strength, especially in cases where a coalition fights a weak opponent, I measure the ratio of power between each side. Capability Ratio equals the power of side A divided by the power of its opponent. 11 To measure power I use the COW Composite Indicator of National Capabilities (CINC) score from the last year prior to the war. In the case of multilateral wars, I sum the CINC scores for each side. 12 Higher values of Capability Ratio should increase the probability of victory.
Given the war-fighting abilities of democratic states, it is important to control for the number of democracies on each side (Pilster, 2011; Reiter and Stam, 2002). To capture this influence the variable Democracy is a count of democracies on one side of a conflict. Past research finds that a larger number of states fighting together increases coordination problems, reducing efficiency (Weitsman, 2013); States is a count of the number of countries on one side.
Last, coalitions that have a pre-existing structure or experience to draw upon should have an advantage in working together (Cranmer and Menninga, 2018). To capture this potential influence, two variables are added to the model: Alliance and Coalition Experience. Alliance measures if there is a pre-war alliance between coalition members. Data for this variable is drawn from the ATOP Dataset (Leeds et al., 2002). Coalition Experience captures if any of the states in a coalition have fought as a coalition in the past. Coalition Experience is coded 1 if the states have been members of a coalition prior to the current coalition and 0 otherwise.
Methods
Because Victory is dichotomous I employ Probit regression. All estimates are reported using errors clustered by war. Given the nature of this study, endogeneity is a concern. Past studies (Morey, 2016) found that endogeneity did not impact the relationship between coalition and war outcome. In the case of command structures, it appears states agree to surrender more authority to coalitions only in dire situations (Allies in the First World War being the classic example). This would indicate that a high probability of defeat encourages states to form stronger command structures, which works against finding support for H1. Since the nature of any relationship between Victory and Coalition Type is uncertain, I test for endogeneity and use instrumental variable models when endogeneity is indicated.
Results
The results in Table 1 provide strong, if qualified, support for the central claim of this paper: increasing command integration increases the odds of victory. Since the proper sample for testing is not obvious, Table 1 presents results using all wars (column 1), multilateral wars (columns 2 and 3), and coalition wars (column 4). 13 In all cases, Coalition Type is positive and significant (P < 0.05), meaning that as coalitions adopt stronger command structures the probability of victory increases.
Coalition command and victory.
Dependent variable is Victory.
P < 0.01; **P < 0.05; *P < 0.1.
Errors clustered by war. P-values reported in parentheses.
To understand the results better, Figure 2 plots the marginal effect for each command type. The graph shows how the probability of victory increases for three states as they increase their level of coordination using the results in column 4. 14 The probability increases from near 0 (0.069) for Independent Command to 0.906 for Unified Command holding all other variables constant. The magnitude of this increase demonstrates the importance that command structure plays in coalition victory. Given the gravity of war outcomes, this increase is significant.

Marginal effects.
Figure 2 also contains a note of caution. Because of the relative scarcity of coalition wars, there is a fair degree of uncertainty. Incorporating the error, it is difficult to delineate the impact of one command structure from its neighbor. The wide confidence interval for Joint Command means that we cannot say for certain that its effect is distinct from that of Independent Command and this holds true regardless of sample. When looking only at coalition wars (column 4) Unified Command is distinct; however, this does not hold for all samples. Therefore, while the overall trend in the probability of victory is strong, the degree of difference between Independent Command and Joint Command and Joint Command and Unified Command must be treated with caution. However, based on all of the results we can say with a high degree of certainty that Unified Command clearly outperforms Independent Command.
To explore this further, I disaggregate Coalition Type into a series of indicator variables. The results of this disaggregated test, while not conclusive, point toward an interesting conclusion. First, in all tests Independent Command fails to reach standard levels of significance. There is no evidence that forming a coalition with this command structure performs better than fighting autonomously. The level of coordination appears to be too low to generate the benefits of fighting as a coalition. The results for Joint Command are not much better. In only one model does Joint Command approach statistical significance (P < 0.1). Overall, the evidence is weak that forming a Joint Command enhances the probability of victory.
Looking at the results for Unified Command, the results reverse. Unified Command is positive and reaches standard levels of significance (P < 0.05). These results suggest that the real benefits of coalition warfare emanate only at the highest level of integration. Instead of states having a potential menu of command options, each increasing the odds of victory, coalitions may be more of an all or nothing investment. Partial efforts do not appear to be enough to overcome the obstacles of combining military forces. This is not to say that coalitions that do not form a Unified Command will never win; instead, these coalitions do not appear to enjoy the actual benefits of fighting as a team.
Evidence for endogeneity between Coalition Type and Victory increases as the proportion of coalitions increases within the sample. When using all wars rho fails to reach standard levels of significance. Narrowing the sample to multilateral wars increases endogeneity to a marginal level (P = 0.171). 15 When we narrow the sample to only coalitions there is clear endogeneity (P = 0.046). However, regardless of the level of endogeneity, the relationship between Coalition Type and Victory remains positive and significant. As seen when comparing models 2 and 3, correcting for endogeneity generally improves the relationship between command strength and victory (ratio between beta and the errors). In this case, endogeneity plays a role and works against finding a significant role for command structures.
A potential conflating factor is the role of the United States. Four wars (Gulf War, Kosovo, Invasion of Iraq, and Invasion of Afghanistan) are US-led missions where there was a clear advantage from US military superiority. Further, three of the wars were fought with Unified Commands and one (Gulf War) was a Joint Command. I re-ran the analyses removing each war and then all four simultaneously. Across all of these tests the primary finding holds with only slight changes in the estimated coefficients. The overwhelming superiority of the United States in these recent wars does not drive the findings (see Online Appendix).
A final issue to consider is the presence of an outside factor driving both Coalition Type and Victory. One possible factor is the presence of a disproportionately strong state within a coalition. Holding a position of influence within the coalition allows the state to demand greater integration as well as provide resources to increase the chance of victory. However, the results show that the presence of disproportionately powerful state does not lead to increased coordination and appears to decrease the chances of victory. Further, most variables associated with victory appear to play little role in determining coalition type. The results from the instrumental variable regressions also indicate that any potential influences currently excluded from the model are not driving the reported findings regarding coalition command structures (see Online Appendix).
Looking at Initiate and Capability Ratio, we see the results highlighted at the beginning of this paper. Both are positive and significant when the sample is all wars. However, once we restrict our attention to multilateral and coalition wars, Capability Ratio and Initiate both fail to reach standard levels of significance. This would indicate that coalitions do not win by overwhelming weak opponents, at least not as a general rule. They also do not appear to gain an advantage from initiating a conflict. While these standard understandings of victory hold true in the bilateral context, there is a clear need to theorize regarding their dynamics in multilateral wars.
While not the primary focus of this paper, the results for the size of a coalition and the number of democratic states are interesting. In most of the samples, as coalition size increases, it becomes a significant impediment to victory (States is negative and significant). This fits arguments that increasing the number of states increases coordination problems and expands the political differences within a coalition (Weitsman, 2013). The case reverses for democratic states with Democracy positive and significant across most models. This would support findings (Choi, 2004) that democratic states do better managing the stresses of coalitions. However, when the sample is limited to coalitions, these results both disappear, with States and Democracy failing to reach standard levels of significance. Care must be taken since the coalition only sample is small and this could hinder inferences. However, the results seem to indicate that once we account for the level of coalition integration the potential negative impact of coalition size and the advantage of democratic partners disappear. 16
Neither pre-war alliances nor previous coalition experience appear to aid in coalition victory. This is surprising given that the essential problem with coalition warfare is coordination. Both working within an alliance and having fought together previously should allow states to work together within a coalition. Future research should explore if there are certain aspects of alliances or experiences from past operations that allow states to work better together in the future. It may be the case that there are nuances of these experiences that matter.
Conclusion
The findings from this paper have important implications for international relations research and policy. First, it is possible for states to combine forces to defeat an attempt at hegemony, thus confirming the central component of classic Balance of Power Theory. However, the results here indicate that low levels of integration will not be enough to aid victory. Instead, states must form a strong hierarchical command structure to achieve the benefits of fighting as a coalition. This means that building effective coalitions may be more difficult than current theory realizes, especially in great power-balancing coalitions, because of overlapping rivalries. If the major powers are too divided, it could mean that an effective counter balancing coalition is impossible even when the necessary military power is available.
In a similar vein, the results provide support for the concept of capability aggregation within the alliance literature. However, it is qualified support; aggregating power does not happen automatically and is a complicated process. This adds another layer to the question of alliance reliability. Moving beyond the question of whether allies will fight together (Leeds et al., 2000), we need to understand whether the states can effectively work together. Alliance agreements that pre-designate a military command structure or create an actual command structure that activates in times of conflict should be more capable than those that do not. Further, the tighter the command integration, the stronger potential targets should rate the alliance.
The findings in this paper also provide more nuance to our current understanding of the role of initiation and power advantage in explaining war outcomes. In the opening, I noted the puzzle that initiation and power advantage did not work the same way for coalitions. The statistical results support the conclusion that increased command integration increases the odds of victory; however, the relationship back to the initiation/power advantage puzzle is harder to see. Exploring the data, it becomes clear that centralizing command does help explain this puzzle. Coalitions with Unified Commands that initiate a conflict win 71% of the time, Joint Commands win 57% of the time, and Independent Commands win 33% of the time. The small number of cases means that the difference between Unified and Joint Commands should be treated with caution; however, the trend is toward coalitions with a more centralized command structure being able to exploit the first mover advantage. Exploring the relationship between capability ratios, command centralization, and victory appears to show even stronger support that coordination plays a central role. First, coalitions fighting with a power disadvantage but with a Unified Command win half of the time. The ability to coordinate what power the coalition does have gives a weak coalition a chance. The more telling results come from looking at outcomes when coalitions have a high power advantage. Unified Commands have never lost when they have a high advantage, 17 while Joint Commands have a record of 2 for 3, and Independent Commands are 1 for 3. Having an overwhelming power advantage is an insufficient guarantee of coalition victory. The inability to coordinate coalition forces and effectively use the coalition’s power leads to defeat by a weaker foe. Nowhere is this clearer than with Independent Commands who have lost four of five wars when fighting with a moderate or high advantage. Clearly, the standard understanding of the relationships between initiation and power advantage to war outcome needs more detail when it comes to multilateral wars.
The need to integrate at a high level will also form a counter balance to the desire to create large coalitions. Large coalitions should be more powerful and generally seen as more legitimate. However, as the number of states increases, the challenges to creating a Unified Command structure should also increase. A small number of like-minded states would have an advantage in forming a functional command agreement. The coalition formed between Australia, the UK, and the United States to invade Iraq in 2003 is a perfect example of a small yet highly functional coalition. This can be contrasted with the larger Arab Coalitions that fought Israel. The Arab forces enjoyed substantial military advantages but the political divisions kept them from forming an effective coalition (Herzog, 1982). Future coalitions will have to balance the political desire for diverse membership with the military reality that comes with broad membership. There appears to be a clear trade-off between legitimacy and efficiency.
The wide variance in the results of Joint Commands is puzzling and warrants further investigation. It is clear that some Joint Commands function smoothly and provide coordination advantages similar to those for Unified Commands while others suffer levels of dysfunction that rival or exceed many Independent Commands. One obvious avenue to explore is the number of commanders. The coalition in the Gulf War functioned smoothly and had only two people sharing command authority while the Boxer Rebellion had representatives from most of the states in the coalition and often did not function effectively. However, the problem will most likely run much deeper and require an understanding of why the states in the coalition could not agree upon a Unified Command in the first place. The rivalries and distinct goals of the states who have been forced together most likely cause a high level of dysfunction with a failure to form a stronger command structure just one part of a larger problem. A more nuanced study focusing on Joint Commands should provide greater understanding of their success and failures and provide advice for how to form a successful Joint Command.
Moving forward, research should consider why, given the advantages associated with Unified Commands, coalitions do not immediately adopt a strong command arrangement. Given the findings states should only be forming highly structured coalitions. What factors explain why states elect to remain loosely affiliated when facing a common threat? Above I suggest that the size and political diversity of the coalition may play a part. However, other factors, such as the nature of the threat, could also be important determinants of coalition cohesiveness. Work in this line will provide a greater understanding of interstate cooperation in security affairs as well as provide advice to policy makers seeking to construct functional coalitions.
Another important direction is to consider the signaling value of coalition structure. Facing a challenge from a coalition, a target must consider more than just the level of military strength. The target must take into account the ability of the coalition to coordinate operations. Prior to the first shot, the coalition command structure will provide a strong signal regarding how the coalition will function and if it will hold together. If the target believes that the coalition cannot work together, they will increase their estimate for victory. This will have important implications for pre-war bargaining and the chance for war (Reed et al., 2008; Wolford, 2014b). Through the adoption of strong command structures, coalitions can send clear signals regarding the level of commitment within the coalition. State leaders will not pay the costs associated with subordinating their troops under foreign command except when commitment is high. In the end, the strength of the command structure, and the process by which the members arrived at it, could serve as important signals regarding the true commitment and strength of a coalition.
Moving beyond the presence or absence of a coalition opens many research opportunities. This paper focused upon one structural aspect, command structure, and found that it plays an important role in determining the probability of coalition success. Coalitions are not the millstones they are often thought to be; however, structure clearly matters. As research continues to examine the characteristics of successful coalitions, we will gain further insight into this important form of international cooperation and be able to offer practical policy advice to leaders contemplating creating a coalition to deal with emerging international threats.
Supplemental Material
Coalition_Command_Results_Appendix_Final – Supplemental material for Centralized command and coalition victory
Supplemental material, Coalition_Command_Results_Appendix_Final for Centralized command and coalition victory by Daniel S Morey in Conflict Management and Peace Science
Footnotes
Acknowledgements
I would like to thank Frederick Boehmke, Doug Gibler, Brian Lai, Stephen Long, Jonathan Markowitz, Dan Reiter, Alex Weisiger, and two anonymous reviewers for comments and suggestions.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Supplemental material
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
