ABSTRACT: I develop a deductive model of civil-military relations in the American context, arguing that the essence of civil-military relations is deciding three issues under conditions of uncertainty: the civilian decision to delegate some degree of policymaking power to the military, especially on matters relating to the use of force; the civilian decision on how best to monitor that delegation; and the military decision to act strategically to enhance delegation and minimize monitoring. Drawing from the principal-agent framework, I then develop a rationalist theory where the decision at each point is a function of the costs of delegating and the costs of not delegating; since the cost variable is latent, I specify a variety of indicators such as advances in communications technology, the representativeness of the military, and the complexity of the issue at stake. Finally, I conclude with a few suggestions for ways of extending this framework to investigate other civil-military issues, particularly the amount of friction in the civil-military relationship.

How do civilians control the military?1

The question has two meanings. Given that military institutions enjoy an overwhelming advantage in coercive power, how is it that civilian institutions are able to impose their will on their more powerful military agents? And, given that civilians in democracies seem to have figured out how to gain general supremacy over the military, how in fact does the control relationship play out on a day-to-day level? This paper explores the second meaning of this most basic political question, developing a new theory of civil-military relations based on principal-agent framework.

Understanding how civilians exercise control -- identifying the factors that shape civilian control and the trade-offs inherent in the relationship -- has particular salience for policies relating to the use of force. The civil-military challenge is to reconcile a military strong enough to do anything the civilians ask them to with a military subordinate enough to do only what civilians authorize them to do. If the military is generally to be subordinate to civilian political leaders, what in particular is the proper role for the military when force is used? How much autonomy can they enjoy without violating the principle of civilian control? Alternatively, how much control can civilians wield without interfering disastrously in the conduct of the military mission? Civilian leaders decide these normative questions everyday for better or for worse in exercising control over the military.

This longstanding democratic theory problem has resurfaced recently in the form of an alleged crisis in American civil-military relations. Participants and observers of the American political scene have noted that relations between the uniformed military and their civilian masters have reached their stormiest in several decades, if not ever.2

The public vitriol directed at the Commander-in-Chief is only the most obvious and controversial manifestation. Perhaps more serious still is the extent to which senior military leaders will now challenge the civilian role in the making decisions for the use of force -- witness recent policymaking on Bosnia, Somalia and Haiti. While such public debates have many historical precedents, they nevertheless raise troubling questions about the durability and manner of civilian control of the military in the United States.

It is imperative, therefore, that civil-military relations theory adequately explain the nature of the relationship as it plays out in decisions governing the use of force in the post-Cold War era. Samuel Huntington and Morris Janowitz, writing over thirty years ago, have advanced the two most prominent theories of U.S. civil-military relations.3

They have inspired the most important empirical work on U.S. civil-military relations and remain the primary if not dominant influences in the field. Despite their prominence, however, I have argued that they do not adequately explain the problem of civilian control and so are uncertain guides for future study and policymaking.4

They skirt some aspects of the civilian control problem and make claims about other aspects that subsequent empirical research has challenged. In short, an alternative theory of American civil- military relations is called for.

I propose to advance such a theory, drawing upon relatively recent advances in the study of political oversight of the non-security bureaucracy. I argue that the essence of civil-military relations is deciding three issues under conditions of uncertainty: the civilian decision to delegate some degree of policymaking power to the military, especially on matters relating to the use of force; the civilian decision on how best to monitor that delegation; and the military decision to act strategically to enhance delegation and minimize monitoring. Drawing from the principal-agent framework, I then develop a rationalist theory where the decision at each point is a function of the costs of delegating and the costs of not delegating; since the cost variable is latent, I specify a variety of indicators such as advances in communications technology, the representativeness of the military, and the complexity of the issue at stake. Finally, I conclude with a few suggestions for ways of extending this framework to investigate other civil-military issues, particularly the amount of friction in the civil-military relationship.

I. THE DEPENDENT VARIABLE: DELEGATION AND CONTROL

As I have argued elsewhere, the civil-military problematique is a simple paradox: because we fear others we create an institution of violence to protect us, but then we fear the very institution we created for protection.5

Following the parable of the development of the polity out of the state of nature, once individuals band together in community, however, the problem of agency arises. Perhaps for a time it is possible to imagine that everyone in the band fights until the threat is gone at which time everyone returns to normal civilian life. But at some point, and well before the emergence of professional militaries, divisions of labor creep in. As the society modernizes, specialization increases. Just as individuals will come to rely on a few producers to provide the goods society needs to subsist, so they will rely on the designated defenders to provide security for the group. In short, societal protection will be institutionalized and this mythical society will directly encounter the civil-military problematique: how to insure that the military will defend the polity and not subvert it.

The problem of agency is inherent in the specialization that results from communal living and it obtains even in polities where the uniformed military are in control of all political affairs. In polities where political leaders and the fighting groups alike wear uniforms, i.e. they both appear to be the same military organization, it may seem as if the division of labor does not exist. But even here, responsibility is divided between those who do the fighting and those who remain behind to wield political power. Wearing the same uniform does not prevent those who stay behind from worrying about whether the fighters are adequate to defend them or whether the fighters are liable to turn around and unseat them, as the prevalence of coups and counter-coups in military dictatorships attests.

The problem of agency is seen most starkly, however, in democracies where the prerogatives of the protectee are thought to trump the protectors at every turn, where the metaphorical delegation of political authority to agents is enacted at regular intervals through the ballot box. Democratic theory is summed in the epigram that the governed should govern. People may choose political agents to act on their behalf, but that should in no way mean that the people have alienated their political privileges. Most of democratic theory is concerned with devising ways to insure that the people can remain in control even as the business of government is conducted by professionals. Civil-military relations are just a special extreme case, and within that case decisions governing the use of force are of greatest importance.6

Agency and control are central to civil-military relations, but how can this be expressed as a dependent variable? Before turning to the one I select, it is instructive to review the alternatives in the literature today. Samuel Huntington suggested that the focus should be on military subordination to civilian authority and this focus was fruitfully extended in comparative studies where the dependent variable has been the decision to launch a coup or perhaps the success or failure of a coup.7

Yet, I would argue, it is precisely this focus on coups that has stunted the development of political (as opposed to sociological) theory of American civil-military relations. American civil-military relations received considerable attention during the 1950s with several major studies,8

but it has been largely ignored by political scientists ever since and the important exceptions have all, by and large, operated within frameworks established by the early Cold War theorists.9

Even the trauma of the Vietnam War proved insufficient to motivate a thorough reexamination by political scientists.10

From a crass political science point of view, the American case seems uninteresting, occupying with dreary regularity the "stable," "harmonious," or "balanced," cell in whatever 2x2 table the typology generates.11

Consequently, political scientists interested in civil-military problems have been uninterested in American politics and those interested in American politics have been uninterested in civil-military relations.

Yet, as I have argued elsewhere, the history of American civil-military relations has been rich with conflict.12

The relationship could only be characterized as harmonious and stable if measured in terms of the extreme values of battlefield collapse or military coup. The traumas of a world war precipitated by a pathological civil-military relationship or a fragile political system brought down by a coup are, in a word, foreign to the American experience.13

But the politics of American civil-military relations have remained vigorous and tense. The pulling and hauling among interest groups, the authoritative allocation of scarce resources, the exercise of power in pursuit of interests -- all this is readily evident in any historical examination of the American military experience. The American experience thus poses an interesting challenge to civil-military relations theory: to develop an explanatory model that is sufficiently broad as to encompass both extremes of the problematique, but also is sufficiently nuanced so as to detect and explain variations in even the "stable" American civil-military tradition.

If a focus on coups would be unfruitful in the American context, what about adopting the dependent variables used by the sociological school, the degree of integration of the military and civilian spheres and the professionalization of the military experience? Unlike setting out to explain the absence of a coup, these variables have the advantage of actually varying over time. From a political point of view, however, they are less interesting than focusing on agency and control. It is instructive that Janowitz, who founded the sociological school as an explicit alternative to Huntington's model of American civil-military relations, does not in fact advance an alternative explanation of civilian control.14

Janowitz offers a wealth of insights into the nature of the military and its position in society and his descriptions of likely areas of friction are in the main accurate. But, excepting his recognition that even professional and generally subordinate military institutions will seek ways to influence if not manage civilian leaders, he makes remarkably little advance over Huntington on the all-important question of agency and control.15

As I outline below, the sociological concern of the integration of the officer corps can be usefully incorporated as an explanatory variable but, at least from a political science viewpoint, it is less interesting as a dependent variable.

Two other potential alternative dependent variables bear consideration: the degree of civil- military conflict and military compliance with civilian directives. A focus on conflict, as opposed to coups, would certainly generate variation over the American experience. Indeed, the recent renaissance of the study of American civil-military relations has been triggered by the heightened acrimony that has characterized the relationship over the past four years.16

Friction is not a trivial concern. Too much friction could be indications and warnings of a coup in the offing. As a dependent variable, it also has the virtue of being relatively easy to measure since evidence of friction and conflict would likely find their way into the public record. Yet it seems a second order consideration, at least in terms of the central civil-military problematique of agency and control. It seems more a consequence of different patterns of civilian control than it is a civilian control issue itself. It is worth relating different forms of control to the presence of friction (and I do so in the conclusion), but the presence or absence of friction does not directly capture the problem of civilian control.

Military compliance would seem to be an even more attractive focus, fitting nicely the problematique's concern with whether the military is doing what it is supposed to do. It is not synonymous with military subordination since it concerns individual decisions and issues, rather than the general scope of political authority. Certainly it is possible that the American military have been able to prevail over civilian leaders on specific decisions even though there has never been a coup. Michael Desch has offered a comparative theory of civil-military relations that uses the military compliance focus.17

Yet measuring compliance is extremely difficult. Determining who wins and who loses in a complicated policy process is not obvious, especially since apparent submission in the decision phase can be undone in the implementation phase. The variable is more interesting than coups, but may be too difficult to operationalize. Focusing on instances of civil- military conflict, as Desch does, risks understating military compliance; what about all the cases where the military obeyed without even challenging enough to get into the news? Moreover, while it yields more variance than the monotonic "no coup" measure, it is relatively stable. By Desch's own measures, between 1938 and 1992, the military complied in all but two cases (Carter's abortive withdrawal from Korea and Reagan's efforts to shift strategic doctrine to encompass protracted nuclear war). Between 1992 and 1994, there have been six major civil-military conflicts and the military has prevailed in all but one (the decision to invade Haiti).

While these other potential dependent variables touch on issues that are important for civil- military relations in general, none cuts to the heart of the relationship itself. Therefore, I use a different dependent variable: the degree of civilian delegation to the military and the kinds of monitoring that civilians adopt to regulate that delegation.18

Such a focus has the virtue of directly capturing the "what" or the "how" of civilian control. Delegation and monitoring constitute the actual processes of civilian control, as opposed to the consequences or inadequacies of civilian control. Moreover, as I will argue in the conclusion, the delegation/monitoring factor can be turned into an explanatory variable to predict, at least partly, other matters of interest to civil- military relations such as civil-military conflict and even military compliance. Consequently, in the next section I propose an alternative to the traditional models, derived from a deductive analysis of the civil-military problematique.

II. CIVIL-MILITARY RELATIONS IN A PRINCIPAL-AGENT FRAMEWORK

Civilians invent the military to protect them from enemies,19

but how can civilians assure that the military will behave as intended? This is a classic principal-agent problem akin to the models developed by micro-economists to explain the challenges confronting an employer setting out to hire workers.20

The employer (principal) would like to hire a diligent worker (agent), and, once hired, would like to be certain that the employee is doing what he is supposed to be doing and not shirking. The employee, of course, would like the job and so has an incentive to appear more diligent during the interview than he really is. Moreover, once hired, the employee may not want to work quite so hard as the employer would wish, all the while sending information back to the employer that suggests he is performing at an acceptable level. The principal-agent framework, then, is all about ways the principal can shape the relationship so as to insure that he has hired employees who will carry out his wishes.

The most fruitful political applications of principal-agency have been in examinations of the way Congress and/or the President interact with the bureaucracy.21

In particular, the principal-agent framework has been used to explore whether the act of delegating policy-implementing and, in some cases, policymaking power to bureaus and agencies is tantamount to an abdication of control by political authorities.22

Overall, as we shall see, the principal/agent perspective cues the analyst to look for evidence of principal control in non-obvious places. The primary claim of political principal-agency is that delegation need not be an abdication of responsibility. A debate continues as to whether it is Congress or the President that is most influential in exercising political control, but the general consensus is that political control of the bureaucracy is greater than the problems of delegation would indicate. Politicians, either as President or as members of Congress, have remained dominant in policymaking even in arenas where they have apparently delegated responsibility to strong bureaucracies, as in the case with environmental regulation. The reason, the principal-agent framework suggests, is that principals are able to exert control in non- obvious ways. Even the absence of the more visible manifestations of oversight -- intrusive audits and close monitoring of the agent's day-to-day behavior -- does not mean the absence of control.

In the civil-military context, the civilian principal seeks to ensure that the military agent does what the civilians want them to do while minimizing the dangers associated with a delegation of power and authority to the military agent.23

The story of the polity establishing the military as a protector force could thus be told in the following terms. Civilians recognize the need for the instruments of violence so they establish the military institution and contract with it the mission of using force on behalf of society.24

Once established, however, civilians must decide how much of the force management assignment to delegate to the military. On the one hand, the more that is delegated to the military the less the burden on civilians. On the other hand, delegation carries with it risks that the will of civilians will be flouted. Once a level of delegation is identified, civilians must decide what ancillary mechanisms they will establish to make sure that the delegation is not tantamount to a total abdication of control. These ancillary control mechanisms have associated costs similar to those involved in the delegation decision itself. Thus, civilians could delegate authority with one hand but then institute such restrictive controls with the other so as to undo the delegation. These decisions by the civilian are a function of the associated costs; costs which are themselves a function of the likely response by the military, who have a stake of their own in the relationship. The military may share the civilians' desire not to lose on the battlefield, but they would also prefer not to be subject to meddling interference by civilian authorities. Moreover, as we shall see, they may not share identical preferences with the civilians on all policy questions and so may seek to manipulate the relationship so as to maximize the likelihood that their will can prevail in policy disputes with their civilian superiors. In short, the military has the ability and the incentive to respond strategically to civilian delegation and control decisions.

A complete model of the civil-military problematique, then, should fully specify the conditions governing each of these stages. This suggests breaking up delegation/control into three different dependent variables reflecting the three decisions: the civilian decision to delegate, the civilian decision to select among a variety of post-delegation control mechanisms, and the agents decision to respond to civilian action.25

A formal game theory model of the American problem, with multiple civilian principals (the public, Congress, the President, and senior executive branch personal), multiple agents (the four services), and multiple decision points, could in theory produce those conditions but would in practice be intractable. Analyzing the civil-military problem in terms of the principal-agent relationship does suggest, however, plausible hypotheses for many parts of the problem, enough to serve as a basis for ongoing analysis. Before identifying those hypotheses, however, it is necessary to examine two sets of factors that affect the decisionmaking of both civilian principals and military agents: one set common to all principal- agent relationships and one set particular to the problem of controlling the use of force.

Principal-agent relationships are characterized by information asymmetries. Both sides share common information. In the civil-military context, they know who the domestic players are, the size of the defense budget, the general identity and nature of their enemies. They share a common history and political memory. But the military agent also has private information that the civilian principal can only discern dimly. The military agent's status as an expert on the management of violence confers significant informational advantages over civilians on matters ranging from tactics to logistics to operational art. Likewise, crucial aspects of military behavior and even predilections may be unknown to civilians. For instance, the civilians cannot know for certain whether the military is inclined to coup nor can it know if the spirit of its orders are being carried out. Moreover, as the operation of the military moves closer to combat, civilians are at an even greater informational disadvantage. It is hard enough to monitor the activities of the military when they are bivouacked near the capitol. When they are deployed on a distant battlefield in the fog of war, communications difficulties could render monitoring impossible for even the most attentive civilian leader.

Two problems emerge from these informational asymmetries: adverse selection and moral hazard. Adverse selection refers to the moment of hiring in the employer metaphor. Has the employer hired someone who is naturally a hard worker or has he been deceived by the interview and hired a lout? Because of hidden information, however, the principal is more likely than not to hire a lout: it is hard to verify the true type and the lout has a great incentive to appear even more attractive than a normal worker. Indeed, because the employer offers a wage that is pegged to attract someone with the statistically average set of qualifications needed for the job, the job will be especially attractive to louts who will know that the offered wages are higher than the employee's true worth; considerably more diligent employees are likely to find any given job less appealing because they will know that the average wage understates their true value. More generally, adverse selection can extend beyond the hiring phase to include all those situations where the agent presents himself, or some proposal, to the principal for approval or decision. For instance, it means that because of their informational advantage over superiors, subordinates tend to propose policies that benefit their own interests rather than the interests of the superiors.26

In a civil-military context, the adverse selection problem shows up both in the accession of personnel into the military, in the promotion of individuals up the chain of command, and in the ongoing give-and-take of military policy. First, it is at least plausible that the peculiar mission of the military -- to kill people and blow things up -- attracts a special kind of person, one that may make the principal-agent relationship particularly problematic. It is reasonable that the demanding mission would attract people with a sense of adventure, a tolerance for hardship, a commitment to order and discipline, and so on. It is possible, however, that these same qualities (or others correlated with them) accentuate the difference between civilians and the military and so lead the military to be especially distrustful of civilian leadership. One of the major concerns of traditional civil-military relations theory is precisely the great divergence of viewpoint between what Huntington calls the liberal civilian ideology and the military mind.27

Adverse selection could suggest that the civilians are liable to pick either people like themselves and so poor warriors, or good warriors who are likely to resent their authority. The decision to "hire" the military is revisited at regular intervals in decisions selectively to promote and/or to fire certain individuals. Of course, as the personnel proceed through the ranks this problem should ease somewhat since civilians have more and more information on which to base decisions because candidates for the higher ranks have already served for many years in many different capacities.

Adverse selection also crops up in the uncertainty civilian leaders would have in evaluating proposals originating from military organizations. Is this budget request necessary to accomplish the mission or is it padding to serve the military organization's interests? Again, because the military has an information advantage it can advance artfully drawn proposals that appear to meet civilian needs but in reality are tailored to their own interests. In the extreme, adverse selection might lead civilians to adopt policies they think are increasing the military's ability to protect society but in fact are increasing the ability or even propensity of the military to take over society.

Moral hazard refers to the behavior of the employee once hired.28

Shirking cannot be easily monitored so performance is usually measured by proxies that substitute, with some loss of validity, for the true goal. Once established, however, the workers have an incentive to optimize on the indicator rather than on the true behavior desired. The gap between the indicator or rule and the desired behavior can be so great that perverse subordinates can bring organizations grinding to a halt simply by strictly observing the official rules, and only the official rules.29

Moral hazard creeps into the civil-military relationship via the peacetime-wartime gap. Militaries exist to fight the country's wars, but they must be maintained during peacetime. How can civilians ensure that the peacetime behavior they are able to monitor correctly indicates that the military will perform as directed during wartime?

Given the problems of adverse selection and moral hazard, the civil-military relationship could degrade into any number of sub-optimal arrangements. The worst and most obvious are the central concerns of the civil-military problematique: collapse on the battlefield or military overthrow of the regime. Other, less dramatic, but nevertheless unsavory options are possible. The military could drag civilian leaders into an unwanted war, which, even if successful, would impose costs that civilians might otherwise choose not to pay. The military could, on a sustained basis, extract higher resources from society than society "needed" to pay for the desired level of protection. The military could successfully resist being used by the civilians according to civilian dictates, perhaps by claiming that a command is impossible (or too costly) or simply by failing to carry out the order.

Most political applications of the principal-agent framework have concerned domestic issues. The special concern at the heart of the civil-military relationship -- controlling the use of deadly force -- introduces peculiar twists that affect the decisionmaking calculus of both principals and agents. The first and most obvious distinctive feature is that the stakes are much higher. If an elected representative turns up a cropper, the damage to the polity (even with the highest elected office) is bounded. Lousy political agents can commit many sins of omission and commission but they are hard-pressed to bring down the republic. Failure to get the military agency problem right can result in one of two grave disasters: the military agent might turn on society and rob it of its political freedom or the military agent may fail on the battlefield and leave the polity vulnerable to conquest by external enemies. In almost all cases, human lives will be lost in the process. The cost of failure, the price of trial and error, is paralyzingly prohibitive. The very existence of the state is at stake and a catastrophic failure can result in the extermination of the polity that initiated the agency relationship. Of course, the fate of the republic does not hang on every military issue and there are many mundane matters where the stakes seem small. In general, however, the consequences of failure are profound and this will cast a shadow over the actors' decisionmaking.

It is plausible that the basic information asymmetry problem inherent in any principal-agent relationship is particularly acute in the civil-military relationship. Traditional theory emphasizes the unique expertise of the military officer: the management of violence. Certainly civilians can gain expertise on a wide variety of defense policy issues, but civilians, by definition, leave combat for the military and combat is the distinctive mark of military expertise.30

Like many other complex policy issues, questions about technical competence and specialized knowledge exacerbate the basic informational challenges facing civilian principals.

Another distinctive feature is that the military is perhaps the only profession that never really gets a chance to practice. Much like a high-priced lawyer on permanent retainer, the military is available for the emergency that may not arise. Unlike the lawyer, however, the activities of the military in the meantime are only very poor indicators of how good they will be when the emergency does arise. To be sure, the military can train and rehearse, but the essence of combat -- wielding force against a determined enemy -- cannot be simulated reliably. The unreliability of performance indicators is a common problem in principal-agent relationships, but the military condition adds an additional layer of uncertainty. Neither side, neither civilian nor military, can be sure about the military's type, at least with respect to performance on the battlefield. In conventional analyses, the expert agent generally understands the consequences of his actions even if the political principal may be uncertain. The military expert can claim that his expertise gives him an advantage and narrows somewhat the uncertainty boundary, but the advantage is not so great as to make him fully aware of whether he will fail catastrophically in battle. The military will have private knowledge about their inclinations to coup, and will certainly know whether their day-to-day activities track with the performance indicators established by the civilian -- neither of which the civilian will know with certainty -- but neither the military nor the civilian will know for sure what all this means for the ultimate purpose of the military: the probability that the military can protect society when challenged in a war.31

The information problem is further exacerbated by the secrecy restrictions that accompany military actions. A common obstacle in principal-agent relationships is the tendency of the agent to withhold information that reflects unfavorably on the behavior of the agent. The classification system vastly eases the task of an agent who wishes to keep inconvenient information from disseminating. While in theory the civilian principal may be entitled to know everything, in practice the costs of doing so are great in any principal-agent relationship. The costs become prohibitive when secrecy laws, with strict punishments for release of classified information, reinforce the agent's natural tendencies. Secrecy laws enjoy legitimacy because they are thought to contribute to the general protection of society. By association, this confers at least a certain amount of legitimacy on military efforts to keep principal civilians in the dark. Moreover, since so many of the principal-agent control mechanisms are essentially efforts to open the agent's hidden behavior to public scrutiny and thereby to alert the principal to improper agency, they are much less appropriate in contexts where even the principal agrees that the agent's behavior ought to avoid general public scrutiny.32

Losing this vehicle, however, makes it easier for the agent to abuse the system, as was evidenced by Oliver North's off-budget covert operations.

Still another distinctive feature of this relationship concerns the problem of competence. The principal-agent framework was first developed as part of microeconomic's theory of the firm and in the original version the principal was assumed to be as expert as the agent in all matters; the agent is hired merely to ease the work burden on the principal. A senior manager in a business has presumably worked his way up, learning the ropes of the junior positions, and hires workers merely for the sake of efficiency. Political science applications have tended to emphasize the difference in expertise between the principal and the agent, where the former may not be technically competent to do the things asked of the agent. Democratic theory further blurs the issue with the notion of political competence discussed earlier -- the agent is not competent to judge risks even if he has special, even unique, technical competence. The special situation of the civil-military relationship adds one further wrinkle: the military agent is asked to put his life on the line because the civilian principal is unwilling to do so. By virtue of his willingness to sacrifice, the military agent may be thought of as possessing a special moral competence, balancing somewhat the political competence of the civilian principal. This moral competence serves to muddy the lines of authority between civilian and military, particularly when the civilian is directing the military to put themselves in harms way. The de jure hierarchy may be unchanged, but the moral ambiguity of the relationship bolsters the hand of a military agent should he choose to resist civilian direction.

The combination of these novel features has a profound effect on the civil-military relationship. Because the stakes are so high, the civilian has an incentive to revisit and retinker agreements and conditions they may have found acceptable before. Because the military cannot claim exclusive knowledge about at least one specific value at stake -- the probability of success on the battlefield -- the civilian is justified in this meddling. Because of classification restrictions, the military have an extra advantage in withholding information thereby frustrating civilian oversight and control. Because the military is prepared to do something that the civilians are not willing to do, the military has an added moral arsenal with which to fight unwanted civilian interference. In short, the expectation is that this principal-agent relationship should be particularly characterized by distrust and friction, and any equilibria of delegation and control are unlikely to endure, ever giving way instead to new arrangements as the costs and benefits of each arrangement shift.

The bulk of the political principal-agent literature is devoted to challenging the proposition that delegation amounts to abdication. Remarkably little attention is paid, in comparison, to the determinants of the principal's choice.33

In general terms, the principal faces two choices: (1) how much to delegate, i.e. over what range and scope of decisionmaking; (2) having delegated, what kind of control mechanism or sets of mechanisms to use to supervise the agents behavior.34

As the preceding section demonstrated, some degree of agency is inevitable in modern civil-military relations: everyone cannot go to the battlefield to fight. But in theory any amount of delegation short of that basic distinction is possible. Force management can be broken down into three broad categories: strategy, structure, and operations.35

These broad categories can be further broken down into still smaller discrete steps. Consider the hypothetical case of a single use of force, say the decision to bomb a Serbian artillery site outside Sarajevo. This might consist of: the decision to flip the switch that drops the bomb; the decision to use a manned aircraft to deliver the ordinance; the decision to target that particular artillery piece; the decision to target an artillery site; the decision to use force against the Serbs; the decision to establish an artillery exclusion zone; the decision to defend the Muslim enclave at Sarajevo; the decision to get involved in the former Yugoslavia; the decision to commit to security and stability in Central Europe; the decision to commit to NATO; the decision that Europe is a vital national interest; and finally, the decision to be engaged in world affairs rather than isolationist. One could imagine another parallel or interwoven series of decisions that would walk back from the ordinance and aircraft, through the pilot's training regimen, through the aircraft production, back to line-items in the defense budget.

Civilians could devise the strategy, deduce operations and battle plans therefrom, specify tactics to achieve those aims, outline logistics and equipment needed to accomplish all this, and direct the provisioning of the forces -- in essence giving complete marching orders to the military. At the other extreme, civilians could simply tell the military "deliver us from our enemies," and let the military decide all the rest. The former would be tremendously costly -- not to mention risky, if the civilians are incompetent. In the extreme, over meddling could so jeopardize the lives of the military, or the fate of the mission, that the military would turn in revolt. Over delegation, would be the least burdensome and would avoid a de jure coup, but it would amount to a de facto coup: the military would be deciding policy and making decisions that by rights belong to the civilian political masters.36

It would require the greatest trust in the military and would leave civilians vulnerable to the agency problems discussed above. How, then, do civilians decide how much to delegate?

Traditional civil-military relations theory has treated this as a normative question: how much ought civilians delegate? Huntington's answer is the ideal division of labor presumed in the social contract: the military will do military things and the civilians will do civilian things.37

The division of labor derives from Clausewitz' theory about the interrelationship of war and politics summarized in the epigram that war is the continuation of politics by other means.38

Deciding what exactly is "civilian" and what is "military" was difficult when Clausewitz wrote; it is decidedly more so today.39

To understand what civilians will delegate under this uncertainty, as opposed to what they ought to delegate, it is necessary to have some theory of civilian motivation. Most principal-agent treatments assume that civilian principals are cost sensitive and I adopt this as my point of departure. Traditional treatments assume that political principals are primarily electorally motivated, and then measure cost in terms of whether the activity diverts from reelection efforts. It is not that political principals are lazy but rather that time spent doing the work that could be done by agents is time not spent out on the election trail. This, however, probably understates costs in the civil-military context (and, indeed, foreign policy generally). If political principals were only electorally motivated -- they had no other interest in the policy, neither for reasons selfish (historical legacy) nor noble (they believe in using political power for good) -- they would probably not devote much time to civil-military relations at all, since its direct electoral impact is marginal.40

That political principals do concern themselves, however, does not make them cost-insensitive or even electorally unmotivated. Rather, it suggests using a richer understanding of costs, to include both electoral costs (time and effort) and what may be called policy costs: the disutility they attach to a divergence between their preferred policy and the actual policy outcome. Since the model can capture both effects, it is an empirical question whether political principals are in general more sensitive to electoral costs or policy costs when considering civil-military relations.

Therefore, I will use a rationalist formulation and treat the delegation decision as a form

of expected utility calculation.41

In this way, the degree of delegation is a function of the associated costs: the costs of not delegating and the costs of delegating. This may be thought of as a make-or-buy decision -- should the civilians "make" the decision or should they "buy" it from the military -- as when a firm decides whether to make a component part internally or to contract for it in the market. Economists have shown that such a decision is a function of the transaction costs of managing production (not delegating) vs. the transaction costs of purchase relationships (delegating).42

In the simplest form, the theory produces the following general hypotheses:

CD_I: The greater the costs of not delegating, the more civilians will delegate.

CD_II: The greater the costs of delegating, the less civilians will delegate.

The challenge for this formulation is in measuring the associated costs. In theory, many factors could affect delegation costs. As a first cut at the problem, I propose the following indicators of the underlying latent cost variable.

The costs of not delegating include: (1) the possibility that the less-expert civilians will make poor policy choices on complicated matters; (2) the cost of civilian time diverted from other matters. The possibility of poor policy choices will vary with the type of issue; the more the issue requires specialized military knowledge the greater the cost of not delegating. For matters touching most closely on military expertise, civilians will have more confidence that the military will produce better policy and less confidence that they can do it better.43

Since the costs of not delegating go up with increasing military expertise, other things equal:

CD1: The more the issue depends on military expertise, the greater the civilian delegation.

The cost of civilian time will also vary with issue type; the less the public salience of the issue, the less benefit there is for electorally minded politicians to devote time to it.44

Other things equal:

CD2: The lower the salience of the issue, the greater the civilian delegation.

The cost of civilian time will also vary with what may be called the complexity or volume factor; the larger the military establishment or the more complex the issue, other things equal, the more time it would take to manage the same range of decisions. Other things equal:

CD3: The larger the military establishment, the greater the civilian delegation.

CD4: The more complex the issue is, the greater the civilian delegation.

Finally, the cost of civilian time will vary with technical feasibility; when rudimentary technology made communication with a far-flung military extremely difficult, the costs of trying increased, whereas when communication technology improved, the costs of trying to direct the military's affairs reduced. Other things equal:

CD5: The more primitive the communications technology, the greater the civilian delegation.

The costs of delegation include: (1) the costs associated with military disobedience on this particular issue, discounted by some subjective probability estimate (i.e. the civilians estimate of how likely it is that the military will disobey); and (2) the costs of general military subversion (which are assumed to be extremely large) discounted by some subjective probability estimate that a particular act of disobedience will lead to general subversion. The costs of particular military disobedience vary with the issue type; the greater the stakes involved, the greater the cost of military disobedience. Obviously, civilians will assess the costs of a nuclear mishap to be much greater than an equally probable conventional mishap -- the stakes are simply much higher in nuclear matters. Likewise, when the external threat facing the country is high, the stakes of any given decision are higher. Other things equal:

CD6: The greater the stakes, the less the civilian delegation. Operationally, the greater the external threat, the less the civilian delegation.45

The civilian's estimate of the probability of particular military disobedience is partly a function of how unified the civilians are on a question. If, for instance, the President is strongly opposed by powerful members of Congress, the military has more opportunity to resist presidential control, hiding, as it were, behind a legislative-executive branch squabble. A divided civilian government, then, increases the likelihood that the military will disobey and so, other things equal:

CD7: The less unified the civilian political leadership, the less the civilian delegation.

The subjective probability estimate also varies with the divergence of viewpoints; the greater the divergence of viewpoint between civilian and military, the less confidence the civilian will have that the military's autonomous decision would be the same as if a fully informed expert civilian had made it. The divergence in viewpoints can be measured directly, as in opinion polls about military attitudes to the use of force.46

In the absence of such data, the divergence of viewpoints can be estimated because it is likely to vary inversely with the degree of representativeness of the military; the more representative the military is, i.e. the more like civilian society in sociological makeup, the less the divergence of viewpoints and the less likely (i.e. the lower the subjective probability estimate) it is that the military decision will diverge from the civilian one.47

Since civilian estimates of the likelihood of the military delivering preferred policy affect the costs of delegating, other things equal:

CD8: The greater the divergence of viewpoints between civilian and military, the less the civilian delegation. The less representative the military is, the less the civilian delegation.

The civilian's subjective probability estimate also varies with the civilian's history of interaction with the military; the longer the experience of military subordination, the less civilians will expect the military to disobey in the next encounter.48

Other things equal:

CD9: The longer the time since a major civil-military conflict, the greater the civilian delegation.

The civilian's subjective probability estimate that a particular act of disobedience will lead to general subversion varies with civilian confidence in their monitoring mechanisms (which are explained more fully in the next section); will the civilian be able to monitor the military or will the military be able to foil those efforts? The effectiveness of monitoring mechanisms varies with many of the factors already mentioned, the size of the military, their historical pattern of obedience, the divergence in viewpoints, and so on. Significantly, the ability of the military to foil these efforts varies with the unity of civilians; the more civilian principals there are in a given civil-military relationship, the easier it is for military to frustrate monitoring mechanisms.49

This reinforces hypothesis CD7: the more divided the civilian principals, the less the civilian delegation.

Finally, there is also a general cost to revisiting the overall delegation relationship. Changing the relationship requires effort and, as noted earlier, it is unlikely that the civilian principal will make changes willy-nilly. Civilians will make changes, however, when the prevailing pattern is considered inadequate. This is likely to obtain when civilian principals have changed, for instance because of an election, or when civilians and the military must confront a new issue for which there is no obvious precedent of how to structure the relationship. Accordingly, other things equal:

CD10: Elections that replace civilian political leaders lead to changes in the amount of delegation.

CD11: The emergence of a new defense policy issue leads to changes in the amount of delegation.

The eleven specific hypotheses outlined above suggest ways of linking indicators of the latent "cost" variable to the delegation decision. For the empirical testing phase of the project, it might be desirable to group the separate indicators into composite indexes. For instance, I could construct one index for costs of delegating and another for costs of not delegating, which would permit an assessment of the relative importance of different cost factors. At this stage, however, it is sufficient to list the indicators.

Once the degree of delegation has been chosen, how will the civilian monitor the military agent's behavior? A central premise of political applications of the principal-agent framework is that political control does not end with the delegation decision. Civilians still have means available with which to direct the military and thereby mitigate the adverse selection and moral hazard problems inherent in delegation. Selecting among those means, however, is difficult. Some mechanisms are costly in terms of time; in the extreme, they could involve so much direct monitoring that they essentially undo the act of delegation. Mechanisms are also costly in terms of distracting the military from their primary mission and so potentially ruining military effectiveness; I relax Huntington's assumption that any post-delegation meddling produces sub- optimal outcomes on the battlefield, but the potential for inexpert civilians to produce bad policy is real and one that civilians must factor into their calculation of how best to monitor the military. The principal-agent literature has identified a wide range of monitoring tools available for political principals. The civilian principal's selection of particular tools is the dependent variable for this stage of the analysis of civil-military relations. Before examining the determinants of this selection, I examine the range of possible monitoring mechanisms and discuss their analogs in the civil-military context.

The principal-agent framework cues us to look for operational control measures in non- operational contexts. For instance, control over the budget and doctrine could be surrogates for control over the performance of the military on the battlefield--an arena the principal cannot oversee because of information asymmetries. These monitoring measures arrange into four broad categories. First, principals can design contracts that give agents an economic incentive to perform in ways that the principal wants; giving an agent a financial interest in the firm's residual, i.e. profit sharing, is a classic example. Political scientists have had difficulties applying this solution to political situations of agency since there is no obvious "profit" to be distributed. Profit-sharing is straightforward when the organization's basic output is easily measured (number of widgets) and the main goal of the principal is greater efficiency (most widgets at least cost). In political applications, the output (policy) itself can be in dispute or only uncertainly measured and economic efficiency (most policy at least cost) is not the only desideratum. Political actors are sensitive to costs, however, and so a potential political analog is something called "slack," the difference between the actual budget appropriation and the minimum cost of providing the service. Slack can be used to buy things that the agent (bureaucrat) wants, like new equipment, perquisites, and so on, but does not actually need to provide the service.50

In this way, the agent has an incentive to be efficient in providing the desired service since he can spend the slack on things he values. Budget autonomy, giving bureaus a fixed amount and allowing them to allocate it as desired, is the most common example of slack in political organizations. The problem with using slack as a control mechanism, is two-fold. It requires that the principal consistently overpay for the service. More seriously, slack does not solve the problem of ensuring that the policy output in fact accords with the principal's desires and is not simply coming in at a lower cost.

The principal-agent perspective suggests that contractual incentives should be at the heart of the control relationship between civilians and military. In the absence of an obvious profit to be shared, however, and given the limitations of slack as a surrogate, what are plausible incentives? Part of the problem of incentives is addressed through screening mechanisms (discussed below) that serve to populate the military with people who share, so much as possible, civilian preferences over outcomes. There are also historical examples of crass versions of profit- sharing as a means of ensuring civilian control; the Romans, for instance, essentially bribed the capitol garrison to keep it out of politics.51

Wages and benefits, and the implied sanction of withholding them, may be modern day equivalents. A particularly intriguing incentive, however, can be found in traditional organization theory's premise that organizations (agents) prize autonomy. Autonomy is slack without a monetary denomination. Since monitoring mechanisms vary in their degree of intrusiveness, and assuming that the military prefer less intrusive means, civilians have a powerful incentive with which to influence military behavior: offer to use less intrusive means to monitor military agents. Indeed, this is how traditional civil-military relations theory treats autonomy. It is the centerpiece of Huntington's ideal-type objective control and is even supported by fierce critics like S.E. Finer.52

Claude Welch describes it as a "noli-me-tangere" approach; civilians promise autonomy to the military in matters of lesser import as an incentive for military acceptance of the ethic of subordination.53

Traditional theory, however, does not distinguish between the initial delegation of authority for a military matter and the subsequent decision on how to monitor that delegation of authority. I am suggesting that civilians have the opportunity to take back with their monitoring hand what they have given with their delegating hand. The incentive they can offer to the military thus derives from the implied threat of intrusive monitoring.

The second general control strategy involves using screening and selection mechanisms to ensure that only the right sort of agent enters into the contractual relationship. Education requirements, skill tests, problem-solving exercises, all represent attempts by employers to screen undesirables out of the candidate pool. Traditional principal-agent treatments discount this control mechanism, however, because the adverse selection problem suggests that undesirable agents will always try to foil entrance screens. Since the principal is inherently limited in his ability to determine the true type of the agent, traditional analyses have focussed on things the principal can do, such as administering punishments and rewards, that enhance control even if screening mechanisms fail. This emphasis seems justified by analyses of iterated games that find leader attributes and the reward structure (which, in principal-agent situations, can be shaped by principals) critical to the initiation and sustainment of cooperation.54

More recent applications of the principal-agent framework, however, have focused on the agent's type and its role in determining whether the agent would comply with the wishes of principals. John Brehm and Scott Gates found that the predispositions of subordinates are far more effective in generating compliance than are the credible threats of supervisors.55

This finding implies that screening mechanisms should be important in determining the overall resolution of the principal-agent relationship.

Accession policy, determining who can join the military and how, is the military version of screening and selection mechanisms. Of course, the military uses elaborate physical, emotional and mental tests to weed out poorly qualified applicants; this directly addresses the civil-military problematique by ensuring that the military will have the physical capacity to defend the country as needed. Accession policy can also mitigate civil-military problems by selecting or promoting personnel who will share civilian preferences. For instance, the emphasis on loyalty in military accession screens is an effort to select applicants who will share as many of the basic civilian preferences as possible. This can be seen in the traditional emphasis on loyalty and its triple meaning in the military context: loyalty as the opposite of betrayal (the military shares the civilian preference of protection from external enemies), loyalty as fealty (the military shares the civilian desire for civilian control), and loyalty as fidelity (the military shares the civilian desire that it honor its contract). Likewise, Janowitz found that the lower the rank of the military officer, the more frustrated they were with civilian viewpoints. This suggests that promotion and selection procedures either winnow out people who do not like civilian control, or at least that their views are shaped by the organization.56

The ability to shape viewpoints points to yet another way that screening can serve to monitor delegation, i.e. through organizational culture. David Kreps and Gary Miller, offer formal analyses of cooperation that stress the role played by organizational culture in enforcing compliance. Organizational culture serves to provide common conceptions of mutual behavior, allowing actors in a political game to have shared expectations of what the other will do.57

In empirical tests of this model, Brehm and Gates found that these cultural factors, which they variously called "cohesion," "solidarity," and "professionalism," had powerful explanatory power in determining when subordinates would work or shirk.58

In the civil-military context, an organizational norm that stresses obedience gives both civilian and military a common expectation that the military will be subordinate. In this one respect, the principal-agent framework shares a finding of traditional civil-military relations theory: that a common culture of subordination -- Huntington's "professionalism," Welch's "cult of obedience," Smith's "norm of civilian control," Hendrickson's "ethic" -- is a crucial component civilian control.59

The third set of controls identified by other applications of the principal-agent framework consists of monitoring and reporting mechanisms designed to track key output variables that indicate whether the agent is complying with the principal's directives. Monitoring and reporting procedures are efforts to overcome the hidden information/hidden behavior problem by making public the activities of the agent. Mathew McCubbins and Thomas Schwartz' influential study of congressional oversight identified two basic types of monitoring systems, dubbed police patrols and fire alarms.60

Police patrol refers to regularized audits and intrusive reporting requirements designed to turn up evidence of agent wrong-doing and, through regularized inspection, to deter moral hazard problems; fire alarm refers to monitoring by a third party who has a vested interest in the actions of the agent, for example, interest groups and affected constituents, and whose objections to agent behavior will set off an "alarm" to alert the principal whenever the agent misbehaves. Reporting and monitoring have obvious political applications, particularly in the Congress-bureaucracy relationship. Public investigative hearings and specific mandated reports are staples of congressional oversight and represent one of the more visible avenues of political control. Similar mechanisms operate within the executive branch to facilitate hierarchical control and some, like the PPBS system in the Department of Defense, directly concern security issues. Congressional investigations have at times featured prominently in the history of American civil- military relations, particularly the Committee on the Conduct of the War during the Civil War and the Senate Permanent Investigations Subcommittee under Senator Joseph McCarthy. Likewise, reports and audits are ubiquitous in the politics of the defense budget. In the civil-military context, an important indicator of police patrol monitoring is the size of the civilian secretariat of the Office of the Secretary of Defense and the service secretariats. Here it is important to distinguish between the confirmable senior political appointees, which I treat as institutional checks (see below), and the larger number of civilian bureaucrats over whom Congress does not have direct influence. The latter are the extension of the executive branch principals, the police patrolmen, if you will, who are in place to monitor closely and directly the activities of their military counterparts. Accordingly, large numbers of nonconfirmable civilian officials is evidence of a police patrol monitoring mechanism; relatively large numbers, relative to the number of military officers working the same issue, are even stronger evidence of police patrol monitoring.

Rules of engagement also act as police patrol monitoring by being, in effect, reporting requirements concerning the use of force. By restricting military autonomy and proscribing certain behavior, rules of engagement require that the military inform civilian principals about battlefield operations whenever developments indicate (to battlefield commanders) that the rules need to be changed. Rules of engagement, then, are both a leash on the military and an information source for senior leaders, civilian or military. So long as the military operators do not "pull" on the leash, the senior commanders know that the pace of the military operation is less than the bounds set by the ROE. The more restrictive the criteria the more closely can senior commanders monitor the military operation.61

Similarly, communications links from national command authorities to operational commanders constitute tangible monitoring and auditing channels. Like traditional principal-agent oversight mechanisms, these measures are costly in terms of civilian attention (not to mention dollars) but can mitigate somewhat the informational asymmetries in the civil-military relationship.

In some respects, the news media functions as a fire alarm in monitoring the military. The media act as independent third agents, self-appointed public watchdogs. They uncover abuses of delegated authority, from weapons acquisition scandals to training scandals to battlefield incompetence, and so trigger more intrusive oversight by civilian officials.

Finally, the fourth set of controls are institutional checks to balance one agent against another. Here the principal has delegated to one agent, but then set up another agent who is in a position to veto or check the first. Institutional checks work best when the interests of the two agents are in conflict, either because they face different contract incentives (one is paid for cutting costs the other for boosting production); otherwise, the two agents could collude and the principal would be back facing the moral hazard problem. Institutional checks can be effective, but there are high costs associated therewith: they make it harder for the agent to do bad things but they also make it harder for the agent to do good things.62

There are numerous political analogs to this set of controls. Indeed, the essence of Madison's genius is in the checks and balances built into the U.S. constitution, power checking power, ambition checking ambition. The classic institutional check for the use of force is the Framers' decision to divide decisionmaking authority on the use of force between the executive and the legislative branch is a classic institutional check; in that instance, the Framers treated the electorate as the principal and government as the agent. Treating civilian government as the principal, pitting individual services against each other can be a means of checking military influence, thus interservice rivalry is an institutional check.63

Moreover, viewed from the perspective of Congressional attempts to control the military, an important institutional check is the presence of senior civilian officials in the Department of Defense over which Congress, through the Senate confirmation process, has some control. Likewise, the civilian staffs of congressional committees are another important indicator of institutional checks.

What factors will determine how civilians will select from among this array of monitoring mechanisms? Surprisingly, the principal-agent literature is largely silent on the question of how actors choose among monitoring mechanisms.64

At one level, one might argue that civilians will select all of them and monitor delegation as rigorously as possible. Perhaps the civil-military relationship is simply too important to allow for any delegation problems.65

It is more likely, however, that civilians will be sensitive to costs at this point just as they were in the delegation phase. After all, if the issue requires such strong monitoring, would the civilians delegate it in the first place? Hence, I will use again a rationalist cost-based theory. Civilians are trading off the costs of policy slack with the costs of reducing it:

CM_I: Principals will only adopt a monitoring mechanism if the benefits exceed the costs.

The question becomes, as before, what are appropriate indicators for the underlying latent variables of cost and benefit? The most obvious is technical feasibility. Certain principal-agent relationships naturally lend themselves towards particular monitoring arrangements and make others unfeasible. For instance, because of the lifetime appointment provision, university principals have very little control over professor agents after tenure. Consequently, selection and screening mechanisms predominate in that relationship. Likewise, fire alarms only work when there is an affected third party;66

in the civil-military context, the news media fire alarm is handicapped by classification restrictions and so is comparatively less useful the more militarily sensitive the issue. As noted earlier, communications technology is a prime determinant of technical feasibility in the civil-military context. The more primitive the technology the less feasible are intrusive monitoring devices like restrictive rules of engagement and large civilian secretariats. Principals are likely to prefer more permissive rules of engagement when communications technology is primitive because, lacking the ability to get the information back from the battlefield in time to respond, restrictive ROE will merely tie the hands of the military. Likewise, large civilian secretariats, which are manifestations of intrusive monitoring, should be less feasible when communications technology deny them access to military operators. Other things equal:

CM1: The more primitive the communications technology, the more permissive the rules of engagement and the smaller the civilian secretariat.

If principals are cost sensitive they will choose the least expensive method that provides control. The easiest control measures to compare are the police patrol and fire alarm forms of monitoring; properly designed, fire alarms can be as effective (if not more so) and they are considerably less expensive in terms of time and effort. Hence the most prominent hypothesis in the literature is: principals will prefer fire alarms to police patrol mechanisms whenever possible.67

This hypothesis assumes, however, that the principal has no reason to doubt the veracity of the fire alarm. But if the fire alarm has selfish interests that vary from the interests of the principal, the principal may not be sure that the fire alarm is accurately reporting on problems caused by the agent. Lupia and McCubbins, noting this problem, make the adoption of a fire alarm a function of the probability of learning from the fire alarm. The more likely the principal is to learn something from the fire alarm -- either because there are penalties for lying, outside verifiers or the principal and the fire alarm share preferences -- the more likely it is the principal will rely on it.68

In the civil-military context, the fire alarm is news media investigation, hence the more likely civilians are to learn something from the press the more they will trust it over the audits and direct investigations of police patrol monitoring. Drawing from Lupia and McCubbins model, one can deduce two factors that might influence the likelihood that the press would serve as a reliable fire alarm. First, the freer the news media (in terms of government restrictions and competition from multiple news sources) the greater the number of outside verifiers and so the greater the likelihood that "false fire alarms" will be exposed (perhaps by a competitor reporter who rebuts the story) -- in short, the more reliable the news media. Second, the more highly classified the subject matter, the less likely the press will have access to the information and so the less reliable they are as a fire alarm. Other things equal:

CM2: The freer the press the less civilians will rely on the civilian secretariat. Under these conditions, most civilian audits and investigations will be triggered by news media reports rather than precede them.

CM3: The more classified the military issue, the larger the relative size of the civilian secretariat and the larger the proportion of government audits and investigations are held in advance of media reports.

Principals will select more costly measures like institutional checks or police patrol monitoring only when the downsides of inappropriate agent behavior are very great, i.e. when the salience of the issue is particularly high. Of course, if the salience is so high, the principal may wish not to delegate in the first place. But delegation may be unavoidable, for instance because the stakes measured in terms of the external threat is also high. Under these conditions, one would expect delegation but one constrained with more costly monitoring measures. Other things equal:

CM4: Under conditions of high external threat, civilians will use more costly monitoring mechanisms like a large civilian secretariat, confirmable and nonconfirmable, for high salience issues.

The relative strength of principals vis-a-vis agents should also influence the kind of costs principals are willing to incur. For instance, some mechanisms like police patrol monitoring, require that principals make a regular and direct challenge to the agents' autonomy. Other mechanisms, like institution checks, constitute a more indirect challenge, i.e. by a third party. When the principal is relatively strong vis-a-vis the agent, the direct challenge of police patrol monitoring is trivial. When the principal is relatively weak, however, the indirect challenge will seem more attractive. What makes a principal relatively weak vis-a-vis the agent? Whether the principal is unified or divided, for an agent can play divided principals off each other. Accordingly, one would expect that the more divided the principal the more they opt for institutional checks, other things equal. In civil-military context, one would expect that divided civilians (through divided government, for instance) will prefer institutional checks like divided military (through interservice rivalry) while a unified civilian government would rely less on interservice rivalry and other institutional checks. Other things equal:

CM5: Countries with divided civilian governments will have more interservice rivalry than will countries with unified civilian governments.

The principals choice of monitoring mechanisms depends in part on their expectation of what the agent is going to do. This dynamic will be explored more fully in a moment, but at this point it is possible to deduce that the greater the principal's expectation that the agent will disobey, the greater the incentive to monitor closely. This same factor influences the decision to delegate and, as argued above, one would expect less delegation when civilians had a greater expectation of military disobedience. Some delegation is inevitable, however, and so under conditions of inevitable delegation and high expectation of disobedience, civilians are likely to use more intrusive monitoring mechanisms. The divergence of viewpoints between civilian and military is one indicator of a probability of disobedience, and divergence itself varies inversely with the representativeness of the military, so one would expect the more lower the representativeness of the military the greater the reliance on intrusive monitoring like police patrol and institutional checks. Other things equal:

CM6: The greater the divergence of views, the larger the civilian secretariat. The less representative the military is, the larger the civilian secretariat.

The foregoing suggests an auxiliary hypothesis: the divergence of viewpoints between the military and the civilian is also likely to increase when the civilian principal is replaced, say by an election.69

If screening mechanisms are used to bring the preferences of the military in line with the preferences of the civilian, then a change in civilians (if it also amounts to a change in civilian preferences) may make the preferences of the military inappropriate again. In other words, the terms under which they were screened no longer obtain because civilian preferences have changed.70

The arrival of new civilian principals should therefore cause an increase in all monitoring mechanisms, and in non-screening ones in particular until such time as the new civilians have weeded out the newly undesirable agents. Other things equal:

CM7: After an election there will be an increase in intrusive monitoring.

A critique of early applications of the principal-agent framework to political problems is that they overemphasized the perspective of the principal. The literature was itself a reaction against claims that principals are powerless after delegation. But in exploring the power of political principals, the literature has overstated its case, and has politicians wielding more control over bureaucratic agents than they in fact wield. The debate between congressional versus presidential dominance obscures the fact that agents may be foiling them both, or playing them off each other.71

Likewise, a simple-minded translation of the economic principal-agent relationship to the political realm misses an important aspect of the agent's motivations. Whereas economic agents may generally prefer shirking to working, political agents are likely to be motivated (at least in part) by a substantive interest in the policy itself. In short, there is a reason why an agent ends up working for the Department of Defense rather than Department of Health and Human Services and this has important implications for how the agent reacts to the principal.

Accordingly it is necessary to analyze the relationship from the point of view of the agent, to make the agent's behavior a dependent variable. The military agent is assumed to have two sets of preferences. First, the military has preferences over policy outcomes. Unlike an economic agent who might not intrinsically care how many widgets he produces, the military agent cares about policy and has a general idea about what should be done. Ideally, the military agent would like to be told to pursue the policy he wants to pursue. Second, the military has preferences over monitoring relationships. Regardless of what the military agent is asked to do, he would like to do it with the minimum of civilian interference and oversight. As in traditional organization theory, the military agent prizes autonomy -- policy autonomy, the ability to decide what to do, and implementation autonomy, the ability to decide how to do it.

Knowing that the civilian principal is attempting to structure the relationship so as to maximize his control (defined as maximizing the likelihood that the outcome will match his preferences), what can the military agent do to foil this effort and retain maximum autonomy? The first and most obvious strategy is to exploit his informational advantage, overpricing civilian requests and underpricing his own preferred policies. As experts, the military know better than civilians what is needed to produce a given policy outcome and can manipulate the apparent cost to the point where civilians will decide it is prohibitive; in formal jargon, the agent has private information about his production function and can inflate the costs to the principal's reservation point. This is the old "Washington monument" tactic, where a bureau responds to a proposed budget cut by offering up obviously unacceptable cost-saving measures like closing the Washington monument.72

The military can similarly resist a civilian direction to use force by claiming that mission cannot be accomplished using limited means and instead requires a massive commitment of force -- a commitment much greater than civilians are willing to contemplate.73

From a military perspective, this is a cheap way to resist civilian policy proposals without being openly insubordinate and so one would expect it to be a relatively widespread phenomenon especially when military and civilian viewpoints diverge (which itself is more likely the less representative the military is). Other things equal:

SA1: The less representative the military is, the more likely they will inflate estimates of the costs associated with civilian proposals.

The military preference for implementation autonomy translates into a general preference for the less intrusive categories of monitoring (contract incentives and screening) as against the

more intrusive categories (oversight and institutional checks).74

One would expect the military to resist measures like police patrol and institutional checks and to propose as alternatives stronger contracts and screening mechanisms. Other things equal:

SA2: Military proposals for monitoring will likely favor contract and screening mechanisms. The military will oppose the growth of the civilian secretariat.

Using the rationalist formulation developed in the analysis of the two previous sections, it is possible to deduce further hypotheses about the agent's behavior. Thus, the military is also cost sensitive, although it is assumed that costs are all in terms of policy outcomes, i.e. the disutility associated with the difference between the expected policy outcome and the military's preferred outcome. While the latent cost variable may be unobservable, it is possible to identify plausible indicators. The disutility itself is a function of how large the discrepancy is expected to be and how important that particular discrepancy is. The discrepancy will be smaller the greater the expectation of civilian deference to military expertise. If the military expects the civilian principal to accept their policy proposals, they will have less reason to act strategically.75

This military expectation is partly a function of the divergence of interests between the civilian and the military (itself a function of the representativeness of the military or whether there has recently been an election) and partly a function of the previous experience. Have civilian principals ignored military advice in the past? Has civilian meddling produced disastrous results, from the military perspective? If the relationship has been characterized by these problems, the military will act strategically. Other things equal:

SA3: The less representative the military the greater the tendency to act strategically.

SA4: After an election, the military will have a greater tendency to act strategically.

SA5: The less the previous autonomy enjoyed by the military, the greater the tendency to act strategically.

The same degree of discrepancy will "cost" more to the military the more strongly they believe that the issue is by rights theirs to decide -- i.e. the more the issue falls in the military bailiwick of the traditional division of labor or the closer the issue is to the military's organizational essence. In other words, the more divergent the civilian and military viewpoints, the greater the incentive for the military to act strategically and seek autonomy. Likewise, the more "military" the issue the greater the military desire for autonomy. Other things equal:

SA6: The closer the issue is to combat operations or to the organizational essence of the military, the greater the tendency to act strategically.

The foregoing outlines conditions under which the military agent will want to act strategically. Of course, civilian oversight complicates the ability of the military to act strategically. The military will have greater ability to influence the outcome the more powerful they are, relative to civilians. A divided principal weakens the civilian and absence of interservice rivalry strengthens the military. Also, the greater the importance of military expertise (i.e. the more technical the issue) the greater the relative power of the military. Other things equal:

SA7: The more divided the civilian government, the greater the tendency of the military to act strategically.

SA8: The more unified the military, the greater the tendency to act strategically.

III. CONCLUSION

The next step, left for further research, is to test the model's hypotheses for each decision point against the empirical record.76

Unlike normative treatments of civil-military relations theory, this model is concerned less with how civilians ought to control the military and more with how civilians are likely to control the military. The causal and dependent variables are all observable in principle, although problems of measurement would have to be overcome before making a robust test of each hypothesis in practice. Moreover, while in theory the causal conditions could change in tandem creating strong expectations of direct effects, in practice, the conditions probably often move in contrary directions. Some factors will be pushing for more delegation while other are pushing for less; some pushing for less intrusive oversight while other pushing for less; some pushing for the military to act strategically while others indicating that civilian oversight will be strict.

The possibility for contrary movements greatly complicates the task of systematically testing the model as specified, but it does suggest an extension of the argument to explain yet another dependent variable: the degree of civil-military friction. Arguably, contrary movements should be associated with open civil-military conflict. Contrary movements in the values of the various indicators signify that the civilian principal is facing cross-cutting pressures on how much to delegate; he is pushed to delegate a lot while at the same time pushed to withhold delegation. Similarly, he is facing incentives to use intrusive monitoring and contradictory incentives to use unintrusive monitoring. All of this should bring decisions to the forefront, exposing the tensions inherent in the civil-military problematique. In other words, open civil-military conflict -- for instance contentious public debates between civilian and military leaders over policy issues -- may reflect an underlying difficulty in making the decision to delegate, the decision to monitor, and/or the decision to respond. Countries with the greatest number of contrary indicators should have the stormiest civil-military relationships. Even countries with generally harmonious relationships should see incidences of conflict increase when the various factors work in contrary directions.

Arguably, this explains the current "crisis" in American civil-military relations which might be viewed as the confluence of two cross-cutting trends: first, strong values on most of the factors predicting that the military would act strategically; and second, contradictory values on the factors predicting whether civilians should delegate. Briefly reviewing the hypotheses on strategic agency, one finds that the military is increasingly less representative, especially measured against the civilian political elite (SA1 and SA3).77

The civilian secretariat of the OSD has remained large, and new offices for humanitarian affairs and environmental security were created or upgraded by the Clinton administration (SA2). Civil-military conflict did intensify upon the heels of the election (SA4) and most observers agree that the military has reacted to civilian micromanagement in Vietnam by acting more strategically to resist unwanted civilian policies (SA5).78

Those who see a crisis generally give prominent mention to civil-military disputes over two issues: Clinton's proposal to lift the ban on homosexuals serving openly in the military and proposals from civilians in both the Bush and the Clinton administrations for a tougher line in Bosnia.79

The debate over homosexuals turns on whether openly gay personnel would be prejudicial to good order, discipline, and morale. The debate over what to do in Bosnia turned on whether a limited military response was worse than no response.80

Both of these issues cut across the different competencies, political, technical, and moral, of civilians and the military and both are at the heart of the military's organizational essence (SA6). Civilians have been divided in general and especially on both issues, while the military, since Goldwater-Nichols strengthened the joint staff, have been much more unified (SA7 and SA8). Even without the end of the Cold War, one would expect much more strategic action by today's military, given these conditions.

Civilian confusion over how to delegate and monitor the military has compounded the problem. On the one hand, several factors push civilians to delegate, especially on the contentious issues cited as most provoking civil-military conflict. The issues are technical (CD1), the issues are complex (CD4), the threat faced by the United States is low (CD6), and there has been a relatively long time since the intense civil-military conflict of the Vietnam war (CD9). On the other hand, the issues are highly salient (CD2), the military is shrinking in size (CD3), communications technology is improving markedly (CD5)81

, civilians are divided (CD7), there is a strong divergence of opinion between civilians and military in general (CD8), there was an election (CD10), and the post-Cold War environment has raised novel issues of America's role in the world (CD11). In other words, the current crisis may be a function of the military having a strong incentive to act strategically and civilians having a difficult time determining an appropriate level of delegation.

Such a hasty overview does not explain every aspect of the current tension in American civil-military relations, nor does it constitute a critical test for the model. It does suggest, however, that the framework can be used to illuminate current problems in American civil-military relations and it points to other uses of my model. The theoretical framework developed in this paper focuses on what may be called the fulcrum of the civil-military problematique: the kinds of measures civilians adopt to regulate their relationship with the military. As I have suggested above, this can be used to explain at least one other potential dependent variable, the degree of civil-military conflict. In follow on work, it would be interesting to turn those delegation/monitoring decisions into explanatory variables affecting still other phenomena, especially the two horns of the civil-military dilemma: concern for battlefield failure and concern for military subordination. Perhaps different kinds of delegation systems are more conducive to military success or continued military subordination. Such an extension depends, however, on the integrity of the original framework and its validity in capturing what I argue is the essence of civil- military relations, the delegation of authority to use force to the military organization and the monitoring of that delegation.

The monopoly on the legitimate use of force is what distinguishes government from other institutions. Understanding how this monopoly is delegated and controlled is therefore central to the enterprise of political science. While traditional approaches to the study of civil-military relations serve as a useful point of departure, I have argued that a new, potentially richer theory is needed. The model presented here draws upon insights gleaned from the study of other domestic institutions, but is tailored to the political challenges peculiar to the civil-military problematique. The model seeks to explain changes in patterns of civil-military relations over time and in response to at least potentially observable features of particular civil-military relationships. The model uses the standard principal-agent framework to identify a wide range of control mechanisms available to principals. The model goes beyond many previous applications of principal-agency, however, in offering hypotheses for how principals might choose how much to delegate and how to monitor that delegation, in light of the agents' incentives to act strategically.

CD_I: The greater the costs of delegating, the less civilians will delegate.

CD_II: The greater the costs of not delegating, the more civilians will delegate.

CD1: The more the issue depends on military expertise, the greater the civilian delegation.

CD2: The lower the salience of the issue, the greater the civilian delegation.

CD3: The larger the military establishment, the greater the civilian delegation.

CD4: The more complex the issue is, the greater the civilian delegation.

CD5: The more primitive the communications technology, the greater the civilian delegation.

CD6: The greater the stakes, the less the civilian delegation. Operationally, the greater the external threat, the less the civilian delegation.

CD7: The less unified the civilian political leadership, the less the civilian delegation.

CD8: The greater the divergence of viewpoints between civilian and military, the less the civilian delegation. The less representative the military is, the less the civilian delegation.

CD9: The longer the time since a major civil-military conflict, the greater the civilian delegation.

CD10: Elections that replace civilian political leaders are likely to lead to changes in the amount of delegation.

CD11: The emergence of a new defense policy issue is likely to lead to changes in the amount of delegation.

CM_I: Principals will only adopt a monitoring mechanism if the benefits exceed the costs.

CM1: The more primitive the communications technology, the more permissive the rules of engagement and the smaller the civilian secretariat.

CM2: The freer the press the less civilians will rely on the civilian secretariat. Under these conditions, most civilian audits and investigations will be triggered by news media reports rather than precede them.

CM3: The more classified the military issue, the larger the relative size of the civilian secretariat and the larger the proportion of government audits and investigations are held in advance of media reports.

CM4: Under conditions of high external threat, civilians will use more costly monitoring mechanisms like a large civilian secretariat, confirmable and nonconfirmable, for high salience issues.

CM5: Countries with divided civilian governments will have more interservice rivalry than will countries with unified civilian governments.

CM6: The greater the divergence of views the larger the civilian secretariat. The less representative the military is, the larger the civilian secretariat.

CM7: After an election there will be an increase in intrusive monitoring.

SA1: The less representative the military is, the more likely they will inflate estimates of the costs associated with civilian proposals.

SA2: Military proposals for monitoring will likely favor contract and screening mechanisms. The military will oppose the growth of the civilian secretariat.

SA3: The less representative the military the greater the tendency to act strategically.

SA4: After an election, the military will have a greater tendency to act strategically.

SA5: The less the previous autonomy enjoyed by the military, the greater the tendency to act strategically.

SA6: The closer the issue is to combat operations or to the organizational essence of the military, the greater the tendency to act strategically.

SA7: The more divided the civilian government, the greater the tendency of the military to act strategically.

SA8: The more unified the military, the greater the tendency to act strategically.

Alchian, Armen and Harold Demsetz. 1972. "Production, Information Costs, and Economic Organization. American Economic Review 62.

Allard, C. Kenneth. 1990. Command, Control and the Common Defense. New Haven: Yale University Press.

Avant, Deborah D. 1993. "The Institutional Sources of Military Doctrine: Hegemons in Peripheral Wars," International Studies Quarterly 37, No. 4, pp. 409-430.

Avant, Deborah D. 1994. Political Institutions and Military Change: Lessons from Peripheral Wars. Ithaca: Cornell University Press.

Bawn, Kathleen. 1995. "Political Control Versus Expertise: Congressional Choice and Administrative Procedures," American Political Science Review 89, No. 1, pp. 62-73.

Betts, Richard K. 1991. Soldiers, Statesmen, and Cold War Crises. New York: Columbia University Press, Morningside Edition.

Boene, Bernard. 1990. "How 'Unique' Should the Military Be? A Review of Representative Literature and Outline of a Synthetic Formulation," European Journal of Sociology 31, No. 1, pp. 3-59.

Bouchard, Joseph F. 1991. Command in Crisis: Four Case Studies. New York: Columbia University Press.

Brodie, Bernard. 1973. War and Politics. New York: McMillan Company.

Brehm, John and Scott Gates. 1992a. "Supervision and Compliance," Papers in American Politics, Working Paper 159, Duke University Program in Political Economy.

Brehm, John and Scott Gates, 1992b. "Policing Police Brutality: Evaluation of Principal-Agent Models of Noncooperative Behavior," paper presented at the Midwest Political Science Association meetings, April 1992.

Brehm, John and Scott Gates. 1993. "Donut Shops and Speed Traps: Evaluating Models of Supervision on Police Behavior." American Journal of Political Science 37, No. 2.

Brehm, John and Scott Gates. forthcoming. Working, Shirking and Sabotage: Bureaucratic Response to a Democratic Public. Book manuscript.

Bianco, William T. and Robert H. Bates. 1990. "Cooperation by Design: Leadership, Structure, and Collective Dilemmas." American Political Science Review 84, No. 1, pp. 133-149.

Burk, James. 1994. "Major trends in Civil-Military Relations," paper presented for conference on "Sociology and War," sponsored by the Triangle Universities Security Seminar, Chapel Hill, North Carolina, 18 November 1994.

Burk, James. 1993. "Morris Janowitz and the Origins of Sociological Research on Armed Forces and Society." Armed Forces and Society 19, No. 2, pp. 167-186.

Clausewitz, Karl Von. 1976. On War, edited and translated by Michael Howard and Peter Paret. Princeton: Princeton University Press.

Coase, Ronald. 1988. The Firm, the Market and the Law. Chicago: University of Chicago Press.

Colton, Timothy. 1979. Commissars, Commanders and Civilian Authority: The Structure of Soviet Military Politics. Cambridge: Harvard University Press.

Colton, Timothy. 1990. Soldiers and the Soviet State: Civil-Military Relations from Brezhnev to Gorbachev. Princeton: Princeton University Press.

Cohen, Eliot. 1985. Citizens and Soldiers: The Dilemmas of Military Service. Ithaca: Cornell University Press.

Cushman, John (USA-Ret). 1994. "Bungling the Informal Contract," Proceedings of the Naval Institute, 120, No. 1, pp. 10-13.

Dahl, Robert. 1985. Controlling Nuclear Weapons. Syracuse: Syracuse University Press.

Desch, Michael. 1995. "Losing Control? The End of the Cold War and Changing U.S. Civil- Military Relations," paper presented at the 1995 Annual Meeting of the American Political Science Association, September 1995.

Dunlap, Charles. 1992/3. "The Origins of the American Military Coup of 2012," Parameters 22, No. 4, pp. 2-20.

Dunlap, Charles. 1994. "Welcome to the Junta: The Erosion of Civilian Control of the U.S. Military," Wake Forest Law Review 29, No. 2, pp. 341-392.

Eagleton, Thomas T. 1974. War and Presidential Power: A Chronicle of Congressional Surrender. New York: Liverright.

Ekirch, Arthur A. Jr. 1956. The Civilian and the Military. New York: Oxford University Press.

Fearon, James G. 1995. "Rationalist Explanations for War," International Organization, 49, no. 3, pp. 379-414.

Feaver, Peter D. 1992. Guarding the Guardians: Civilian Control of Nuclear Weapons in the United States. Ithaca: Cornell University Press.

Feaver, Peter D. 1993. "Discords and Divisions of Labor: The Evolution of Civil-Military Conflict in the United States." Paper presented at the 1993 Annual Meeting of the American Political Science Association, September 1993.

Feaver, Peter D. 1995. "Civil-Military Conflict and the Use of Force," in Donald Snider and Miranda A. Carlton-Carew, eds. U.S. Civil-Military Relations: In Crisis or Transition?

Feaver, Peter D. 1996a. "An American Crisis in Civilian Control and Civil-Military Relations? Historical and Conceptual Roots," The Tocqueville Review 17, No. 1, pp. 159-184.

Feaver, Peter D. 1996b (forthcoming). "The Civil-Military Problematique: Huntington, Janowitz, and the Question of Civilian Control," Armed Forces and Society.

Fenno, Richard F. 1973. Congressmen in Committees. Boston: Little, Brown.

Ferejohn, John and Charles Shipan. 1990. "Congressional Influence on Bureaucracy," Journal of Law, Economics, and Organization 6, pp. 1-43.

Finer, S.E. 1962. The Man on Horseback: The Role of the Military in Politics. London: Pall Mall Press.

Gacek, Christopher. 1994. The Logic of Force: The Dilemma of Limited War in American Foreign Policy. New York: Columbia University Press.

Goldstein, Judith and Robert Keohane. 1993. Ideas & Foreign Policy: Beliefs, Institutions and Political Change. Ithaca: Cornell University Press.

Goodpaster, Andrew J. 1977. "Educational Aspects of Civil-Military Relations," in Andrew J. Goodpaster and Samuel P. Huntington, Civil-Military Relations. American Enterprise Institute: Washington, pp. 29-54.

Gordon, Michael R. 1992. "Powell Delivers a Resounding No on Using Limited Force in Bosnia," New York Times, 28 September 1992, pp. A1, A5.

Halperin, Morton. 1974. Bureaucratic Politics and Foreign Policy. Washington: Brookings.

Hamilton, James T. and Christopher H. Schroeder. 1994. "Strategic Regulators and The Choice of Rulemaking Procedures: The Selection of Formal vs. Informal Rules in Regulating Hazardous Waste," Law and Contemporary Problems 57, Nos. 1&2.

Hammond, Thomas H., Jeffrey S. Hill, and Gary J. Miller. 1986. "Presidents, Congress, and the 'Congressional Control of Administration' Hypothesis," paper presented at APSA.

Hendrickson, David. 1988. Reforming Defense: The State of American Civil-military Relations. Baltimore: The Johns Hopkins University Press.

Henkin, Louis. 1972. Foreign Affairs and the Constitution. Mineola, NY: Foundation Press.

Huntington, Samuel P. 1957. The Soldier and the State: The Theory and Politics of Civil-Military Relations. Cambridge: Harvard University Press.

Huntington, Samuel P. 1961. The Common Defense: Strategic Programs in National Politics. New York: Columbia University Press.

Janowitz, Morris. 1971. The Professional Soldier: A Social and Political Portrait. New York: The Free Press.

Javits, Jacob K. 1973. Who Makes War: The President vs. Congress. New York: Morrow.

Jervis, Robert. 1976. Perception and Misperception in International Politics Princeton: Princeton University Press.

Kerwin, Jerome G. 1948. Civil-Military Relationships in American Life. Chicago: University of Chicago Press.

Kiewet, D. Roderick and Mathew D. McCubbins. 1991. The Logic of Delegation: Congressional Parties and the Appropriations Process. Chicago: Chicago University Press.

Kohn, Richard. 1994. "Out of Control," National Interest, 35, pp. 3-17.

Kreps, David M. 1990. "Corporate Culture and Economic Theory," in Perspectives on Positive Political Economy, ed. James E. Alt and Kenneth A. Shepsle. New York: Cambridge University Press.

Lasswell, Harold D. 1950. National Security and Individual Freedom. New York: McGraw Hill.

Luttwak, Edward. 1994. "Washington's Biggest Scandal," Commentary, May.

Lyons, Gene M. 1961. "The New Civil-Military Relations." American Political Science Review, 55.

Lindsay, James M. 1994. "Congress, Foreign Policy, and the New Institutionalism," International Studies Quarterly 38, No. 2, pp. 281-304.

Lupia, Arthur and Mathew D. McCubbins. 1994. "Designing Bureaucratic Accountability," Law and Contemporary Problems 57, Nos. 1&2, pp. 91-126.

March, James and Johan Olsen. 1984. "The New Institutionalism." American Political Science Review, 78:3.

McCubbins, Mathew D. and Thomas Schwartz. 1984. "Congressional Oversight Overlooked: Police Patrols Versus Fire Alarms." American Journal of Political Science 2, pp. 165-179.

McNollgast. 1987. "Administrative Procedures as Instruments of Political Control." Journal of Law, Economics and Organization 6, pp. 243-277.

McNollgast. 1989. "Structure and Process, Politics and Policy: Administrative Arrangements and the Political Control of Agencies." Virginia Law Review 75, pp. 431-482.

McNollgast. 1990a. "Slack, Public Interest, and Structure-Induced Equilibrium." Journal of Law, Economics and Organization 6, pp. 203-212.

McNollgast. 1990b. "Positive and Normative Models of Procedural Rights: An Integrative Approach to Administrative Procedures." Journal of Law, Economics and Organization 6, pp. 307-332.

Miller, Gary J. 1992. Managerial Dilemmas: The Political Economy of Hierarchy New York: Cambridge University Press.

Millis, Walter. 1956. Arms and Men: A Study of American Military History. New York: Putnam.

Millis, Walter, Mansfield, H. and Stein, Harold. 1958. Arms and The State: Civil-Military Elements in National Policy. New York: Twentieth Century Fund.

Mitnick, Barry M. 1994. "The Hazards of Agency." Paper presented at the 1994 Annual Meeting of the American Political Science Association.

Mitnick, Barry M. 1975. "The Theory of Agency: The Policing "Paradox" and Regulatory Behavior." Public Choice 25, pp. 27-42.

Moe, Terry M. 1984. "The New Economics of Organization," American Journal of Political Science 28, No. 4.

Moe, Terry M. 1990. "The Politics of Structural Choice: Toward a Theory of Public Bureaucracy." In Organizational Theory: From Chester Barnard to the Present and Beyond, Oliver Williamson, ed. New York: Oxford University Press.

Moskos, Charles C. 1970. The American Enlisted Man: The Rank and File in Today's Military. New York: Russel Sage Foundation.

Moskos, Charles C. and Frank R. Wood. eds. 1988. The Military: More Than Just a Job? Washington: Pergamon-Brassey's.

Niskanen, William. 1971. Bureaucracy and Representative Government. Chicago: Aldine-Atherton.

Nordlinger, Eric. 1977. Soldiers in Politics: Military Coups and Governments. New York: Prentice-Hall.

Perlmutter, Amos. 1977. The Military and Politics in Modern times: On Professionals, Praetorians, and Revolutionary Soldiers. New Haven: Yale University Press.

Peterson, V. Spike. ed. 1992. Gendered States: Feminist (Re)Visions of International Relations Theory. Boulder: Lynne Rienner Publishers.

Powell, Colin, John Lehman, William Odom, Samuel Huntington and Richard Kohn. 1994. "Exchange on Civil-Military Relations." National Interest, 36, 23-31.

Posen, Barry R. 1984. The Sources of Military Doctrine: France, Britain, and Germany Between the World War. Ithaca: Cornell University Press.

Putnam, Robert. 1995. "Bowling Alone: America's Declining Social Capital." Journal of Democracy 6, No. 1.

Rosen, Steven. 1973. Testing the Theory of the Military-Industrial Complex. Lexington, MA: Lexington Books.

Ross, Stephen A. 1973. "The Economic Theory of Agency: The Principal's Problem." American Economic Review 62, No. 2, pp. 134-139.

Rouquie, Alain. 1982. The Military and the State in Latin America, translated by Paul E. Sigmund. Berkeley: University of California Press.

Sagan, Scott D. 1991. "Rules of Engagement." Security Studies, 1:1.

Sagan, Scott D. 1993. The Limits of Safety: Organizations, Accidents, and Nuclear Weapons. Princeton: Princeton University Press.

Sagan, Scott D. 1994. "The Perils of Proliferation: Organization Theory, Deterrence Theory, and the Spread of Nuclear Weapons." International Security 18, No. 4.

Sarkesian, Samuel C., John Allen Williams, and Fred B. Bryant. 1992. "Civilian Graduate Education and U.S. Military Professionalism," paper presented at the Annual American Political Science Association Convention.

Sarkesian, Samuel C. 1975. The Professional Army Officer in a Changing Society (Chicago: Nelson-Hall Publishers.

Schiff, Rebecca. 1992. "Civil-Military Relations: A Theory of Concordance." Paper presented at the 1992 Annual Meeting of the American Political Science Association.

Segal, David R. 1994. "National Security and Democracy in the United States," Armed Forces & Society 20.

Smith, Louis. 1951. American Democracy and Military Power. Chicago: University of Chicago Press.

Snyder, Jack. 1984. The Ideology of the Offensive: Military Decisionmaking and the Disasters of 1914. Ithaca: Cornell University Press.

Sofaer, Abraham. 1976. War, Foreign Affairs, and Constitutional Power: The Origins. Cambridge: Ballinger.

Spence, David B. No Date. "Administrative Law and Agency Policymaking: Rethinking the Positive Theory of Political Control."

Stepan, Alred. 1971. The Military in Politics: Changing Patterns in Brazil. Princeton: Princeton University Press.

Vagts, Alfred. 1937. A History of Militarism: A Romance and Realities of a Profession. New York: W.W. Norton & Co..

Van Evera, Stephen. 1984. "The Cult of the Offensive and the Origins of the First World War." International Security 9, No. 1.

Weingast, Barry R. and Mark J. Moran. 1983. "Bureaucratic Discretion or Congressional Control? Regulatory Policymaking by the Federal Trade Commission." Journal of Political Economy 91, No. 5, pp. 765-800.

Welch, Claude, ed. 1976. Civilian Control of the Military. New York: State University of New York Press.

Welch, Claude. 1987. No Farewell to Arms? Military Disengagement from Politics in Africa and Latin America. Boulder: Westview Press.

Weigley, Russel F. 1993. "The American Military and the Principle of Civilian Control from McClellan to Powell." The Journal of Military History 57, No. 5, pp. 27-58.

Yarmolinsky, Adam. 1971. The Military Establishment: Its Impacts on American Society. New York: Harper and Row.

Zisk, Kimberley M. 1993. Engaging the Enemy: Organization Theory and Soviet Military Innovation, 1955-1991. Princeton: Princeton University Press.

---

1

I thank Deborah Avant, Stephen Biddle, John Brehm, Kurt Dassel, Michael Desch, John Duffield, Colin Elman, Hein Goemans, Jay Hamilton, Jeff Legro, Mariano Magalhaes, Jim Miller, Clifton Morgan, Brian Taylor, and the participants in the National Security Seminar of the Olin Institute for Strategic Studies for their helpful comments on this project. Earlier versions of this paper were presented at the 1995 Annual Meeting of the APSA and the 1995 Annual Meeting of the Interuniversity Seminar on Armed Forces and Society.

2

Cushman (1994); Dunlap (1993); Dunlap (1994); Kohn (1994); Luttwak (1994); Powell, et al., (1994); and Weigley (1993).

3

Huntington (1957) and Janowitz (1971).

4

Feaver (1996b).

5

The section that follows is drawn from Feaver (1996b).

6

Because my focus is on the United States democracy, I bracket for the moment an important problem that emerges in a comparative context: when the military sees itself as the agent of the disembodied state or society rather than of the government/regime and so does not view the government as even in a position to delegate (and not to delegate) authority. While the question "who is the military an agent of?" is largely settled in the U.S. context, it remains a major issue for comparative civil-military relations, of course. The framework I develop here would have to be modified to incorporate this problem for comparative studies. The motivating thesis of my argument is that the problem of delegation and control remain interesting even when the prior question of who is the principal and who is the agent has been settled, as in the U.S. case. Of course, the model I develop here should work best in other democracies (such as the United Kingdom) where the basic principal-agent question has likewise been settled. I have benefitted greatly from discussions with Mariano Magalhaes on this point.

7

Huntington (1957); Nordlinger (1977); Perlmutter (1977); Rouquie (1982); Stepan (1971); Welch (1976); and Welch (1987).

8

Lasswell (1950); Smith (1951); Ekirch (1956); Huntington (1957); Millis (1956); Millis, Mansfield and Stein (1958).

9

Betts (1991); Brodie (1973); Cohen (1985); Hendrickson (1988). One could argue, of course, that there is a vast empirical literature empirical literature on matters touching on civil-military relations, including defense organizational reform, defense procurement, defense policymaking, and so on. My point is that there has been very limited theoretical development, particularly as it pertains to the mechanisms of how civilian institutions control military ones in the United States.

10

Two important exceptions are the body of scholarship examining the role of the President in authorizing the use of force and the literature on the military-industrial complex. The former, however, is largely juridical in perspective, analyzing the issue in terms of what is or is not allowed under the Constitution. See, for instance: Eagleton (1974); Henkin (1972); Javits (1973); and Sofaer (1976). The military-industrial complex literature, in contrast, adopted a political science institutional focus, although its impact has been mixed. See Rosen (1973) and Yarmolinsky (1971).

11

For example, Finer (1962), especially pp. 88-89. Or Welch (1976).

12

Feaver (1996a).

13

So the influential work on civil-military relations and the origins of war of some ten years ago essentially ignored the U.S. case. Posen, at least, addresses the U.S. case. Posen (1984), pp. 220-244. Snyder (1984); Van Evera (1984).

14

Feaver (1996b).

15

The sociological school has virtually kept the flame of American civil-military relations alive. Their findings on the changes in military career tracks, on the challenges faced by military families, and on the nuances of military professionalism are of great sociological import but seem less relevant to political scientists concerned with the exercise of power on the major issues of war and peace. Moskos (1970); and Moskos (1988); and Sarkesian (1975). For a review of this literature see: Boene (1990); Burk (1994); Burk (1993). The principal journal for this branch of the study of civil-military relations is Armed Forces & Society.

16

Cushman (1994); Dunlap (1993); Dunlap (1994); Kohn (1994); Luttwak (1994); Powell, et al., (1994); and Weigley (1993).

17

Desch (1995).

18

As I explain in the next section, this dependent variable will be further broken down into the decision over how much to delegate, the decision over how to monitor delegation, and the decision by the agent to respond to the delegation.

19

Peterson (1992) reverses the causal arrow and has the military inventing outside enemies to create the civilian state. Regardless of the plausibility (or implausibility) of this argument as a general explanation of military politics, it does not amount to a qualitative change in the civil-military problematique, although it might intensify the difficulties associated therewith.

20

Alchian and Demsetz (1972); Niksanen (1971); Ross (1973);

21

For a good literature review see: Hammond, Hill, and Miller (1986) and Spence (no date). For an introduction to the now-standard applications see: Ferejohn and Shipan (1990); Kiewet and McCubbins (1991); McNollgast (1987); McNollgast (1989); McNollgast (1990a); McNollgast (1990b); McCubbins and Schwartz (1984); Mitnick (1994); Mitnick (1975); Moe (1988); Moe (1984); and Weingast and Moran (1993). Note that "McNollgast" is a pseudonym for Mathew D. McCubbins, Roger G. Noll, and Barry R. Weingast.

22

I treat principal-agency as a framework rather than a theory since it is not a system of statements of cause and effect. It is most fruitfully used to formulate question but, as I argue in the text, it requires additional theories to explain changes in the relationship over time. Moreover, as I also argue in the text, the literature is remarkably thin on such important questions as how principals choose to delegate and how they choose among various monitoring mechanisms.

23

Deborah Avant was the first to apply the new institutionalism argument to civil-military relations theory, using it to explain different propensities for innovation across different military organizations. Avant (1993) and Avant (1994). Robert Jervis (1976, pp. 332-342) uses a similar metaphor to understand the difficulty of reliable crisis management. The only other application in a security setting of which I am aware is an examination of Congress' role in foreign policymaking, Lindsay (1994).

24

The contract is ritualized in the officer's oath of allegiance, and reinscribed through a myriad of cultural symbols like the privileged place assigned to the military in celebrations of national holidays.

25

The assertive-delegative typology I used in Guarding the Guardians combined the two civilian decisions, how much to delegate, and how to monitor delegation. Thus, civilians adopted assertive control if they delegated little and monitored a lot. Delegative control involved higher degrees of delegation and lower (or less costly) forms of monitoring. Feaver (1992).

26

Moe (1984), pp. 754-5. Mitnick (1994) traces the origin of the term "adverse selection" to the economics of insurance literature which found that insured populations had a higher propensity to experience the adverse consequences insured against than did the uninsured populations. In other words, unhealthy (or reckless) people are more likely to want insurance and also more likely to need insurance. From the point of view of a profit-maximizing insurance company, this is adverse selection .

27

Huntington (1957), pp. 59-79. Of course, this does not mean that all military officers think alike. As we shall see, manipulating differences of opinion within the military is an important method of civilian control.

28

Mitnick (1994) again traces the origin of this concept to the economics of insurance literature. Once insured, individuals have less incentive to take care of their property since any losses will be covered by the insurance.

29

"Work-to-rule" is a popular form of workers' revolt in large complex bureaucracies like post-offices and police forces.

30

Andrew Goodpaster states the idea particularly forcefully, "It is not that all wisdom resides in the military; far from it. Civilian students of strategy, operations, procedures, intelligence methods, procurement practices, educational methods, and training techniques have offered much in the past and continue to do so, though such contributions, be it noted, have not come without effort or intensive application to the facts. But it is the test of combat, or the perceived probable results of the test of combat -- the unique domain of the military professional -- that ultimately and fundamentally establishes the validity of military posture and action." Goodpaster (1977, p. 32). I would not press this point too

far, however, and I would guess that the gap between civilian and military is more narrow than Goodpaster allows. Civilian expertise grew considerably during the Cold War, whereas combat experience, except for a very small number of officers, did not.

31

Hence this is an even harsher form of agency than Mitnick's "closed agency," where both the information and the action is hidden to the principal. In the civil-military relationship, some information is hidden from the military agent as well. Mitnick (1994).

32

Lindsay notes that this hampers Congressional oversight of foreign policy generally and so limits the claims of congressional dominance made by principal-agent theorists. Lindsay (1994), p. 283.

33

Among the few exceptions are Bawn (1995), Brehm and Gates (1992a and 1992b), Hamilton and Schroeder (1994), Lupia and McCubbins (1994), and Spence (no date).

34

One might be tempted to add a third choice: whether to decide again the question of delegation. It is rather idealized to imagine civilians revisiting the civil-military problematique and deciding anew the question of delegation on every single security issue. In practice, the costs of reinventing the wheel probably make bureaucratic inertia (continuing the current pattern) an attractive decision. Deborah Avant (1994, p. 12) stresses this phenomenon, calls it "institutional bias," and argues that it helps explain the proclivities of different military organizations to accept doctrinal innovations pressed on them by civilian superiors. Civilians still have the option, however, of changing the relationship on every single issue; that they sometimes choose not to just underscores that their choices are sensitive to costs, as my model captures. On the other hand, given the peculiar nature of the use-of-force agency relationship (as discussed earlier), there is reason to believe that the delegation decision will be reconsidered more often than other non-military delegation relationships, even if the review simply reinforces the earlier decision. In any case, this third choice collapses into the first two and can be treated accordingly.

35

Feaver (1992), p. ix. Huntington (1957, pp. 3-4) treats operations as a subset of strategy which he distinguishes from structure.

36

This is precisely the fear of Kohn (1994) and Weigley (1993) with respect to post-Cold War United States.

37

Huntington (1957), pp. 70-74.

38

Clausewitz (1976, pp. 605-6), outlines what has become the classic division of labor: civilians decide political-strategic questions and the military decides operational-tactical questions. On the one hand:




The assertion that a major military development, or the plan for one, should be a matter for purely military opinion is unacceptable and can be damaging, It is in any case a matter of common experience that despite the great variety and development of modern war its major lines are still laid down by governments: in other words, if we are to be technical about it, by a purely political and not a military body. This is as it should be.




On the other hand: "Policy [political considerations] will not extend its influence to the operational details. Political considerations do not determine the posting of guards or the employment of patrols."

39

Feaver (1993).

40

Of course, if the policies prove disastrous or deviate widely from the views of constituents then there will be an electoral connection. Lindsay (1994) found, however, that defense policy issues have only a very marginal electoral impact. See also Fenno (1978). Mayhew (1974, p. 122) claims that congressional principals (at least) are essentially (even solely) electorally motivated and so do not engage on those policies, like most military issues, that bring no electoral benefit.

41

The rationalist explanation is the logical place to begin theorizing, even if there is reason to believe that other explanations may also be relevant, for instance an ideational one based on the construction of the military and civilian identities. Without a rationalist baseline explanation there is nothing against which to compare alternative explanations, Fearon (1995).

42

Coase (1988). This grounding was suggested to me by Hamilton and Schroeder's analysis of an EPA administrator's decision to use informal rules (i.e. decide a matter internally) or to use formal rules (i.e. post the rules and involve the public and other actors as required by statute). Hamilton and Schroeder (1984), p. 128.

43

Feaver (1992) and Bawn (1995).

44

Spence (no date).

45

This is the essence of Posen's (1984) theory of civilian control of doctrine. Elsewhere (Feaver 1992, pp. 73-76), I discussed two contradictory effects of external threat on civilian control. A general increase in the perception of a threat, raises the importance of military readiness and so might encourage civilians to delegate control so as to minimize the chance that an enemy could preempt with a decapitating strike. On the other hand, a crisis is characterized by the sudden focusing of attention on security matters. This might prompt a civilian government to assert more control where they had previously delegated. I found that at least for nuclear policy, the former effect captures well standard military arguments for greater delegation; senior military commanders regularly asked for greater delegation over nuclear weapons, pointing to an enhanced Soviet threat for justification. Civilians, however, were less likely to initiate this argument although they sometimes were persuaded by it. Accordingly, I emphasize the opposite effect in this general formulation. It is an empirical question if it holds in the more general case.

46

For instance, Sarkesian et al. have found that junior military officers are more likely to place a higher priority on aggressive military policy than are senior military officers. He attributes the change to exposure to professional military education which civilianizes the military viewpoint. This may explain Betts' apparently contrary finding that senior military advisors did not differ with their civilian counterparts in the priority they placed on an aggressive military policy. The military institution does have the bias, but as officers become more senior and gain broader exposure to the civilian world, the bias is reduced. Sarkesian's evidence is also consistent with Betts' finding of a significant difference in the civilian and military viewpoints on how best to use force, once a decision to use has been made. Sarkesian et al. (1992) and Betts (1991).

47

Obviously, if there is no divergence in viewpoint at all, the costs of delegating to the military reduces to zero -- in short, there is no principal-agent problem. Brehm and Gates (1992a), pp. 26-27.

48

This, in terms of my model, is what is meant by Huntingtonian and Janowitzean professionalism. The expectation (although not the certainty) that military are likely to obey the next time, other things equal. It is the visible demonstration of competence and respect for civilian authority that makes the costs of delegating authority to the military less than it otherwise might be.

49

This is similar to what Bawn calls procedural uncertainty, the possibility that after delegation and after monitoring the agent's behavior, the agent will still deliver policies that deviate from the principal's ideal point. Bawn argues that procedural government is higher in divided government, because the agency will be subjected to conflicting pressures. Bawn (1995), pp. 65-66.

50

Moe (1984), p. 763.

51

Smith (1951), p. 3. Likewise, before the rise of professional armies, the military was customarily "given an interest in the economic residual" through the privileges of booty; before delivering the conquered territory to the king, the army would pillage what it could.

52

Feaver (1993) and Feaver (1996a). Clausewitz (1976), pp. 605-6. Brodie (1973), p. 494. Huntington (1957), pp. 83-85. Smith (1951), pp. 50-51. Finer (1962), pp. 39-56, especially 47-56, and 141-144. Betts (1991), p. 10. Hendrickson (1988), p. 11.

53

Welch (1976), pp. 33 and 318.

54

Bianco and Bates (1990).

55

They argue, in other words, that adverse selection trumps moral hazard. Solve the adverse selection problem and you do not need to worry about moral hazard. Fail to solve adverse selection, and no reasonable amount of monitoring can prevent moral hazard. Brehm and Gates (1992a) and Brehm and Gates (forthcoming).

56

Janowitz (1971), p. 368.

57

According to the Folk Theorem, certain games, including the iterated principal-agent problems of interest in this study, have many equilibria, some involving cooperation (or compliance) and some involving defection (or insubordination). Formal analysis can not say which particular outcome will obtain or even how players can influence the game to move to their desired outcome. Kreps (1990) and Miller (1992) treat organizational culture as Schellingesque focal points, strategies that naturally draw the attention of players and which distinguish salient equilibria as likely outcomes.

58

Brehm and Gates (1992b and 1993).

59

Huntington (1957), pp. 70-78; Welch (1976), p. 33; Smith (1951), p.5; Hendrickson (1988), p. 26.

60

McCubbins and Schwartz (1984).

61

For an in depth analysis of rules of engagement see Sagan (1991).

62

Kiewet and McCubbins (1991), pp. 33-4.

63

Huntington (1957, pp. 418-423) notes this disapprovingly. In a subsequent work, however, he calls it "a key aspect in the maintenance of civilian control." Huntington (1961) p. 378. These are not necessarily contradictory observations, if the latter is a resignation to a condition that the former disparaged. This view seems implied in Huntington's observation, "The services undoubtedly found it easier and more virtuous to tangle with each other than to challenge civilian groups and arouse the hallowed shibboleths of civilian control." Huntington (1961), p. 381. Huntington does not investigate the implications of this for his model of civil-military relations.

64

A notable example, cited later in the text, is Lupia and McKubbins (1994).

65

Thus, some people argue that any pollution is bad and so we must monitor potential polluters with an eye to eliminating any amount of pollution. This is called a lexicographic preference, where the principal will always prefer more of a certain item and will not be sensitive to costs.

66

Kiewet and McCubbins (1991), pp. 34-35.

67

Kiewet and McCubbins (1991).

68

Lupia and McCubbins (1994), p. 102.

69

The divergence might also grow if the military agent's preferences creep over time .

70

This helps explain the particularly intense problems surrounding President Clinton's effort to lift the ban on homosexuals serving openly in the military. Since President Clinton was the first prominent civilian principal to attempt to change this policy, he was confronting a military that had been screened according to old preferences on this issue. It also suggests that this problem will ameliorate over time as applicants are screened according to the new set of civilian preferences (assuming President Clinton's preferences prevail over time).

71

Hamilton and Schroeder (1994); Spence (no date).

72

Niskanen (1971).

73

Gordon (1992).

74

This holds at a general level. At a more detailed level, this produces rather counterintuitive secondary expectations. For instance, within oversight, the preference for greater autonomy should translate into a military preference for fire alarms as against police patrol mechanisms. It is not obvious that the military markedly prefers the watchdog oversight of the news media to the direct monitoring of OSD civilians. Certainly, the military's relationship with the media has been stormy and has been characterized by the military's determined effort to keep the media at arms length (especially during operations). On the other hand, this may be a function of the military's overall desire for autonomy and given a direct choice between a news media (which might be more easily duped) and a vigorous civilian secretariat (which would be in the military's knickers at all times) the military might prefer the former. Likewise, it is not clear that the military resents interservice rivalry, as is implied by this ranking of institutional checks low in the preference order. Here the apparent anomaly is more easily explained away. Once established, interservice rivalry has an element of self-perpetuation about it. Military members get conditioned to think as service members and so are less likely to adopt the ab initio perspective implied by the preference ranking. Since Congress finally established institutional incentives for jointness in Goldwater-Nichols, the grip of the services has weakened considerably. In any case, the more general expectation that the military would prefer greater reliance on screening mechanisms to a larger number of civilian appointed officials in the OSD is eminently plausible.

75

This is why Janowitz and Huntington stresses civilian respect for military professionalism. From Janowitz: "...civilian supremacy is effective because the professional soldier believes that his political superiors are dedicated men who are prepared to weigh his professional advice with great care." Janowitz (1971), p. 367.

76

Table II in the appendix summarizes all the hypotheses.

77

Witness the decline in ROTC, which alienates the military from society, and the decline in service experience among politicians, which alienates the political elite from the military.

78

See, for instance, Dunlap (1994), pp. 349-52.

79

Cushman (1994); Dunlap (1993); Dunlap (1994); Kohn (1994); Luttwak (1994); Powell, et al., (1994); and Weigley (1993).

80

American civilians and military apparently agreed, for the most part, that a major Desert Storm-level commitment of U.S. forces to bring victory to one side was out of the question for political reasons. The force deployed in December 1995, while large, came in with a very narrow mission and has assiduously avoided anything resembling combat.

81

This is particularly manifest in the ability of civilian, even UN officials, to call off air strikes while the planes are enroute.

Back to Top