On Schelling and the Fallacy of Positive Doctrines

History has rightly praised the Harvard economist and Nobel laureate Thomas Crombie Schelling for developing the foundational concepts of bargaining and strategic coercion. Each of these concepts has inspired numerous others. The bargaining model gave rise to attempts to understand how parties weigh the costs and benefits of armed conflict, even as they prosecute such conflicts, and why they choose to settle when they do.[i] Similarly, studies of coercion have examined Schelling’s theories of compellence and deterrence in efforts to increase our understanding of them and to make both more effective.[ii]

Unfortunately, as seminal as they were, Schelling’s concepts of bargaining and coercion represent the type of “positive doctrines” Carl von Clausewitz eschewed. “We must remind ourselves,” said Clausewitz, “that it is simply not possible to construct a model for the art of war that can serve as a scaffolding on which a commander can always rely on for support.”[iii] This statement has been taken to mean the Prussian opposed theories that prescribed the steps to be taken for success. But it also reflects his abiding disdain for oversimplified models, those that focus on one causal factor over others, such as chance, uncertainty, fear, warlike passions, talent, and genius. Such theories clearly neglected reality. The best role for theory, he concluded therefore, was to analyze the relationships between ends and means, effects and their causes, and then to convert the insights gained from those analyses into subjective (personal) knowledge, or skill.

Schelling’s error was not so much that he developed theories to predict rather than to explain, though he is guilty of that to a degree, but that he oversimplified war by attempting to reduce it to a rational sequence of decisions, a decision-logic. In effect, he deliberately excluded war’s irrational factors, much like the many theories Clausewitz rejected nearly two centuries ago, to gain a better understanding of what thought processes contributed to an individual’s decision to concede. In so doing, however, he fell victim to the analyzer’s paradox: isolating elements to gain a better understanding of each individually distorts the role they play collectively. Still, Schelling’s theories hold lessons for the contemporary strategist; this essay discusses some of their strengths and weaknesses.

Schelling’s curriculum vitae

Schelling was born into a US Navy family in 1921. But unlike his well-known contemporaries, Bernard Brodie and Robert Osgood, both of whom served in stateside (and largely safe) assignments during the Second World War, Schelling had absolutely no military experience. Both the US Army and US Navy declared him physically unfit for military duty; hence, Schelling worked instead for the US Bureau of the Budget as an analyst. He obtained a bachelor’s degree in economics from Berkley in 1944. After the war, he became involved in administering the Marshall Plan, and served as a foreign aid advisor to the Truman administration. He completed a doctorate in economics with Harvard in 1951, and by 1956 he had joined the Yale faculty.

Shortly thereafter, he published two important articles on bargaining: “An Essay on Bargaining” and “Bargaining, Communication, and Limited War.”[iv] The latter essay, especially, is said to mark the beginning of his transition from an economist to a strategist, a transition some critics contend he never successfully made.[v] Schelling sent a prepublication copy of the latter essay to Brodie who, unimpressed, replied he had already addressed many of the same issues several years earlier.[vi] Despite an inauspicious start, the two maintained what Brodie later described as a “rugged friendship” in which each commented candidly on the other’s draft manuscripts.[vii]

By 1958, Schelling had connected his theory of conflict behavior to game theory, then heavily based on the works of John von Neumann, by arguing for a complete “reorientation” of the field toward “bargaining.”[viii] Shortly thereafter, he joined the faculty at Harvard and began a consulting tenure with the RAND Corporation. Through his relationship with RAND, he became an advisor to Assistant Defense Secretary John McNaughton regarding the strategic direction of the Vietnam conflict, especially the 1965 bombing campaign known as “Rolling Thunder.”[ix] The full extent of Schelling’s involvement remains unclear, but presumably his theories of coercive bargaining helped shape McNaughton’s expectations for the campaign.[x] In 1970, Schelling terminated his capacity as policy advisor in protest over the Nixon administration’s expansion of the war into Cambodia.[xi] Thereafter, he turned his attention to applications of game theory to endeavors outside war.

Schelling later claimed his theories never directly influenced US Cold War policies. Admittedly, influence can be difficult to prove and policymakers rarely give credit to others in any case. Yet Schelling clearly left a lasting impression on game theory and on the concepts of bargaining and coercion, all of which informed American Cold War policy in important, if sometimes subtle ways. In addition, his most important works, Strategy of Conflict (1960) and Arms and Influence (1966), remain foundational to the disciplines of international relations and strategic studies.[xii] Nobel Laureate Roger Myerson recently went so far as to declare Strategy of Conflict a “masterpiece that should be recognized as one of the most important and influential books in social theory.”[xiii]

The Bargaining Model

Like most of his contemporaries, Schelling drew a sharp distinction between limited and all-out war. He equated the latter to a zero-sum game—one party gains all, while the other loses all. The former, in contrast, constituted a non-zero-sum game—each party achieves an acceptable outcome. He further claimed the “maneuvers and actions of limited war” amounted to a “bargaining process.” A bargain is struck, he said, whenever “someone makes a final, sufficient concession,” that is, a rational decision to stop unfavorable things from happening, or to enable favorable things to begin happening.

Bargains normally take place by means of “tacit agreements” and “explicit negotiations,” and generally pertain to the following categories: (1) the conduct of the war itself, which includes agreements over what types of weapons could be used, who could participate in the fighting, and how, etcetera; (2) the details of the armistice or surrender document that specifies the conditions for halting the conflict; (3) the status of the enemy regime, or regimes, and their fates after the fighting; (4) the disposition of the territories involved in the fighting; (5) long-term agreements, such as disarmament or inspection protocols, to preserve the peace; and (6) the status of countries, nations, alliances, and coalitions necessary to satisfy the conditions of peace.[xiv] Bargaining, thus, underpinned not only how limited wars ended, but also how they unfolded, as well as their direct and indirect consequences.

If bargaining is prevalent in limited wars, argued Schelling, then the way strategy is thought about requires a reorientation, if not a revolution. Strategy is less about destroying an opponent’s material and psychological capacity to resist than it is about modifying the opponent’s behavior—whether through “threats” or “threats and promises”—until the desired concession is obtained. Strategy is, thus, an interactive process because the “ability of one participant to gain [its] ends,” depends “to an important degree on the decisions that the other participant will make.”[xv] In addition, the strategy of conflict, Schelling maintained, applies not only to war, but to virtually any form of conflict or competition that is not a zero-sum game.

Schelling later admitted that ambiguity and uncertainty can obscure the bargaining process. Bargaining could proceed in fits and starts, he said, or be carried out by people who had little experience with it, or who were under strict time constraints or other pressures, or who might look upon the idea of bargaining as a form of “appeasement” or of “collaboration with the enemy.”[xvi] Nonetheless, he believed such frictions did little more than modify an individual’s decision-logic: “Why does [a competitor] concede?” he asked; the answer was “Because he thinks the other will not. I must concede because he won’t. He won’t because he thinks I will. He thinks I will because he thinks I think he thinks and so [on].”[xvii]

Because it was deductive in nature, this focus naturally appealed to the rapidly growing field of game or decision theory.[xviii] But, at the same time, it excluded explanations that approached decision-making in terms of organizational, emotional, or sociological behaviors. While subsequent generations of scholars would fault Schelling’s Strategy of Conflict for this shortcoming, it was already duly noted by contemporary reviewers.[xix]

Schelling’s bargaining model has other faults as well. He admitted that friction might obscure the bargaining process, but he still assumed all parties would act rationally, they would allow themselves to endure only so much pain and would prefer to negotiate rather than pay too much for what they wanted, or risk not getting it at all. He excluded irrational forces, such as warlike passions, and the possibility one party might be willing to pay any price for what it wanted, regardless of the likelihood of success. As he later said of the Viet Cong, “We wanted to convince them that we could tolerate more pain than they could, but they weren’t rational.”[xx] In short, if one or both warring parties did not behave as if they were in a bargaining situation, did bargaining exist?

Two other problems are worth mentioning. First, Schelling’s model assumes strategy is interactive in nature, but that is not always the case. While strategy’s variables are indeed interdependent, they do not all depend on the actions of the adversary. Some of strategy’s activities are noninteractive in nature, that is, they occur independent of a foe’s actions or decisions because these are not known; or they occur in relation to achieving greater security but not in relation to a specific foe. Strategy, therefore, has the potential to be interactive but also the potential to be insular. Second, zero-sum games and non-zero-sum games can coexist even in limited wars. An example is the repatriation of prisoners of war in the Korean conflict. Both sides saw that issue as nonnegotiable, hence, as a zero-sum game, in a conflict that was a non-zero-sum game in most other respects.[xxi] A theory of armed conflict that cannot accommodate a mixture of zero-sum and non-zero-sum games cannot accommodate real war.

Despite its faults, Schelling’s bargaining model advanced modern strategic thinking by reinforcing the critical, but easily overlooked role of negotiations in bringing wars to a close. It also served as a reminder that the parameters of war themselves are continuously shaped by implicit and explicit agreements.

Coercive Strategies: Compellence & Deterrence

Published one year after the 1965 Rolling Thunder campaign, Schelling’s Arms and Influence explored the ways in which one might use military force coercively, that is, through threats or threats and promises, to make an adversary concede. He assumed that making rivals give up something differed qualitatively from forcibly taking it from them. The former relies on intimidation, or potential harm; while the latter involves actual harm. The threat of more harm to come offered more leverage, in his view, than harm already inflicted. “[B]rute force succeeds when it is used,” Schelling wrote, “whereas the power to hurt is most successful when it is held in reserve.” Moreover, “the power to hurt is bargaining power;” whereas exploiting that power is diplomacy, “vicious diplomacy,” to be sure, but diplomacy nonetheless.[xxii]

To illustrate the difference between intimidation and brute force, Schelling offered a number of historical examples, in particular the US military’s campaigns against Native Americans in the nineteenth century. “To hunt down Comanches and exterminate them was brute force,” he said; whereas “to raid their villages to make them behave was coercive diplomacy, based on the power to hurt.” To be sure, “the pain and suffering to the Indians might have looked much the same one way as the other.” But “the difference was one of purpose and effect.” If Indians were killed because “authorities despaired of making them behave” that was “pure unilateral force.” But if “some Indians were killed to make other Indians behave, that was coercive violence—or intended to be, whether or not it was effective.”[xxiii]

Schelling identified several prerequisites for the use of force coercively: (1) the situation had to afford both sides some negotiating space and, thus, could not be a zero-sum game; (2) the potential harm had to be sufficient to make the costs of noncompliance discernably higher than its benefits; (3) the threat of harm had to be credible, that is, within the capability and willingness of the coercer to apply it; (4) the promise of relief if the foe complied also had to be credible; and (5) one had to provide the adversary a specific but adequate timeframe within which to comply.[xxiv] Obviously, these conditions presupposed the ability to control not only one’s own force, but also much of the general situation.

Because Schelling saw deterrence as coercive, he coined the term “compellence” to refer to coercive strategies that were active or offensive in nature. Compellence meant “inducing” an opponent to take an action, such as withdrawing its forces from an area. Deterrence, then, referred to coercive strategies that were passive or defensive. Compellence usually required administering punishment until an adversary took appropriate action; deterrence required inflicting punishment only if an opponent did not comply.[xxv]

This construct appealed to game or decision theorists because it separated two logical trains of thought for the development of branches and sequels. It also appealed to limited war theorists because it created the impression that coercive pressure could be applied in graduated ways, either to compel or to deter. Presumably, one could influence a rival’s behavior without committing more military force than necessary.

By the 1980s, however, critics had begun to scrutinize the dual theories of compellence and deterrence more closely. As a result, some of the accepted differences between were called into question. It was not necessarily valid, for instance, to assume that because compellence was more active than deterrence that it was also more difficult to achieve.[xxvi] Nor was it clear that proving a positive, like compellence, was any easier than proving a negative, like deterrence.[xxvii] As Henry Kissinger, Schelling’s contemporary and fellow Harvard alumnus, once explained, deterrence “can only be tested negatively, by events that do not take place;” however, it is “never possible to demonstrate why something has not occurred.”[xxviii] Years later, it is still not clear why Rolling Thunder failed to compel Hanoi to negotiate; or what finally compelled Slobodan Milosevic to comply with NATO’s demands in 1999.[xxix]

Conclusions & Implications

Understanding the strengths and weaknesses of Schelling’s theories brings us back to Clausewitz’s issues with positive doctrines. The same theme—the presumption of definitive causes or explanations—runs through the bargaining model as well as his concepts of deterrence and compellence. To ask why certain strategies have not worked, a common habit of late, is to presume they should have worked. That presumption encourages us to continue to refine our explanatory models and, ultimately, if unconsciously, to construct positive doctrines: if we determine in one case that X caused Y, then we will naturally expect X to cause Y in the future.

A positive doctrine also assumes we can get at primary sources (recordings, notes, memos, interviews, etc.) that will pinpoint what a decisionmaker’s exact reason was for taking a specific action. However, such sources are often not available until years, if not decades, after the fact. Even when they become available, they offer little genuine assurance that what they reveal, in fact, reflects what was on the decisionmaker’s mind when the critical moment arrived. This is true even when data is abundant, as is the case with the Cuban missile crisis (1962) and the Mayaguez incident (1975).[xxx] In each case, key elements such as the US presidents’ estimate of the foes’ intentions, his understanding of the potential costs and benefits of military action, as well as of the impressions such action might have on American and world opinion, among other factors, all combine to make it impossible to isolate an individual’s decision-logic without distorting it in the process. This, again, is the classic analytical paradox: taking something apart prevents us from appreciating how the whole may be greater than the sum of its parts.

Certainly, drawing precise cause-effect relationships in war is problematic in any case since multiple actions are underway at any given time. We cannot necessarily determine exactly which action, or group of actions, caused an opponent to act, or to refrain from acting. Nor did Clausewitz necessarily intend for us to do so. Rather, the point is to avoid taking analysis too far, to refrain from pushing it until we discover the so-called definitive cause. Inductive analysis suggests war and strategy have neither independent nor dependent variables, only interdependent ones. Any examination of strategy, accordingly, must be capable of accommodating interdependence.

To return to Schelling’s example of the US government’s campaigns against Native Americans—in practice, the US government’s strategy fell into a cycle consisting of threats, negotiations, attacks, renegotiations, renewed threats, and so on. Several strategies were used at different points in the sequence. Among these were attrition, exhaustion, terror, and dividing-and-conquering (turning some tribes against others), all of which formed an overall synthetic strategy that compelled some actions and deterred others. The loss of loved ones and of charismatic leaders, fear for the future of the tribe, weariness and a sense of despair, the lack of food and shelter in a harsh climate, balanced against promises of relief—all likely had a bearing on why the Indians complied. Attempting to distinguish which strategy was most important, therefore, risks losing the fact that they might only have been effective in combination with the others.

The answer, therefore, lies more in attempting to preserve the synthetic nature of any strategy we study, rather than reducing it to what we presume to be its decisive element. Equally important, our strategies must work even in environments characterized by ambiguity, uncertainty, and imprecision. Hardly a strategist today would reject Clausewitz’s concept of friction or dispute how its many types impede thought and action in war. Yet, so often irrational and nonrational factors are politely set aside as if they only affected practice, not theory. Instead of repeating that error, we must construct theories that allow for and anticipate, rather than seek to exclude, such forces. In a word, our theories must embrace interdependence.

Even though Schelling acknowledged the importance of uncertainty and other forms of friction, his theories could not actually accommodate them. While they remain popular among students of decision and game theory, theirs is the only world in which positive doctrines work. Let’s not mistake it for the real world.