Introduction

Any inquiry into escalation and negotiation in an international context must pay special attention to arms races and their inherent tendency to escalate. Since these phenomena deal with quantities – numbers of ballistic missiles, aircraft, tanks, troops, etc. – it is natural that their scientific treatment includes quantitative analysis. On the other hand, arms races and escalation depend fundamentally on international negotiations, and international relations in general, so that the qualitative methodological tools of political science are also appropriate. The published literature on international negotiation and escalation contains considerable quantitative work on arms races as such, but little quantitative work on negotiations about such races. Thus, it is not surprising that very little quantitative work addresses both subjects at the same time. It is the purpose of this chapter to develop some new ideas on this and to identify directions for future research.

Richardson (1939, 1960) was the first to describe arms races quantitatively and phenomenologically using a system of coupled linear differential equations. His work was brought to a wider audience by the well-known review article of Rapoport and Lewis (1957), which unleashed a veritable avalanche of arms-race research, including both formal theory and empirical analysis. The wealth of theoretical and empirical studies of arms races has been well documented in bibliographies (Cioffi-Revilla 1979) and literature reviews (Busch 1970; Intriligator 1976, 1982; Moll and Luebbert 1980; Avenhaus and Fichtner 1984).

Introduction

The many negotiation tactics called “escalation” have at least one feature in common: they are all risky. One famous metaphor for escalation, recounted by O'Neill (1991; see also PBS 1983) makes this point graphically:

The struggle in Vietnam can be interpreted as an implicit negotiation in which the status quo was simply the continuation of the conflict at its current level. From this perspective, Johnson's escalation strategy was to render the status quo less preferable for the opponent – ideally, to make it unendurable – and thereby pave the way for a resolution that Johnson preferred. This tactic, which can be called settlement pressure escalation, is implemented by reducing the opponent's comfort level – withholding services, impounding assets, supporting its domestic opposition, etc. Settlement pressure escalation is sometimes seen as a good if costly bet, which the initiator “wins” if the target gives in first. However, as more recent events in the former Yugoslavia clearly demonstrate, this bet is risky – success is far from certain, and the initiator's costs can easily exceed those of the target.

The fundamental tenets of classical deterrence theory are incompatible with models based on rational choice. Deterministic threats, which rest on the presumption that states can commit themselves to irrational actions, violate the canons of both classical deterrence theory and instrumental rationality. While the threat-that-leaves-something-to-chance maintains consistency with the rationality postulate, consistency is achieved only by assuming that irrational actions are carried out by a disinterested third party. And the resolution suggested by a metagame analysis serves merely to bring the paradox of mutual deterrence into sharper focus: when the costs of war are so high that all-out conflict is the worst outcome for both players, mutual deterrence is unlikely. It might appear, then, that there is no escape from this pernicious puzzle.

We believe otherwise. In this chapter we propose still another resolution. Our proposal is based on the concept of perfectly credible retaliatory threats. Like other attempts to eliminate the paradox, this resolution involves a modification of the underlying game form; but unlike other proposals, ours retains the core realist assumption of an anarchical international system while remaining faithful to the definition of instrumental rationality. In a narrow technical sense, then, the paradox of mutual deterrence stands. Technicalities aside, however, the resolution we propose permits an intuitively satisfying and empirically plausible respecification of classical deterrence theory that helps to explain the dynamics of a wide range of deterrence relationships, including those that occur under the conditions that delimit the paradox.

By design, the deterrence models explored in part II are extremely austere. To be sure, the “no-fat” modeling approach we adopt allows us to focus directly on the role of uncertainty and credibility in both mutual and unilateral deterrence games. But axiomatic austerity cuts both ways. The ability to penetrate core theoretical structures and analyze the role of a few fundamental variables is not altogether costless. Parsimony is inversely related to the complexity and range of questions that a model can fruitfully address. For example, in the simple models developed in part II, conflict is an all-or-nothing proposition. As a consequence, these models are unable to shed any light on the conditions associated with either limited conflicts or escalation spirals. Nor do our rudimentary models capture well the subtleties of some more complex deterrence situations. Thus, to address these and related limitations, we now begin to complicate, ever so slightly, our bare-bones deterrence models and to explore a number of questions associated specifically with extended deterrence relationships.

In this chapter we begin by describing a generic two-level extended deterrence/escalation model and discuss its characteristics under complete information.

The consensus view of classical deterrence theory is deficient, both empirically and logically. Empirically, the theory is hard put to explain, inter alia, the stability of the Cold War period before the Soviet Union achieved essential equivalence with the United States, the absence of an all-out conflict between the Soviet Union and China, especially during the most contentious stretches of this strategic relationship, and the historical tendency of major-power wars to occur under parity conditions. Logically, the theory is marred by a fundamental incompatibility between its tenets and the canons of rationality. This is the paradox of mutual deterrence. More specifically, logic implies that the status quo should unravel as higher and higher costs render mutual conflict worse and worse for both sides. Classical deterrence theory, however, asserts the opposite.

States clearly do not always behave the way classical deterrence theory suggests they do or should. Waltz (1993: 53–54) notwithstanding, the glaring discrepancy between logic and fact, between prescription and description, is troubling. Accordingly, in this chapter and the next, we inquire whether classical deterrence theory can be resuscitated, that is, whether it can be rendered logically coherent and, ultimately, empirically accurate. To this end, we now explore the merits of three potential resolutions of the paradox of mutual deterrence that attempt to explain general deterrence stability without violating the rationality postulate: deterministic threats, “threats-that-leave-something-to-chance,” and a solution suggested by the theory of metagames.

Even a cursory examination of typical behavior patterns under an all-or-nothing defense posture reveals the weakness of a policy that forgoes a tactical-level threat. The only remotely plausible deterrence equilibrium associated with all-or-nothing deployments requires that Challenger be Soft and likely unwilling to wage a high-level conflict. This, however, is the easier case. Relying exclusively on a threat to escalate is insufficient to deter Challenger when Challenger is likely Hard and, therefore, likely to respond, tit-for-tat, to any escalation.

It is no wonder, then, that the Eisenhower administration's deployment policy came under intense criticism among defense analysts almost as soon as it was announced. Critics asserted, prematurely it now appears (Gaddis, 1997), that the threat of Massive Retaliation lacked credibility (Kaufmann, 1956). And they claimed that the New Look, by stressing “more bang for the buck,” placed undue reliance on strategic weapons to deter Soviet aggression in Europe and elsewhere, leaving little room for maneuver during periods of acute crisis. To avoid the stark choice of either all-out nuclear war or capitulation, they proposed that United States conventional forces be strengthened, and augmented with an arsenal of tactical nuclear weapons (Kissinger, 1957b).

When the Democrats came to power in 1961, these changes were pursued under a policy labeled Flexible Response. In 1967, after extensive debate and compromise, NATO formally adopted Flexible Response (Stromseth, 1988).

For some, the mere idea of deterrence conjures up stark images of the bipolar world that existed before the Soviet Union splintered apart, an unwelcome vestige of the superpower rivalry that dominated most of the latter half of the twentieth century. Now that the Cold War is over, some analysts have concluded that deterrence, and all its attendant concepts, are no longer relevant. Regional and ethnic conflicts seem sure to dominate the new millennium, so the argument goes, and therefore deterrence theory can be safely relegated to the dustbin of history, or sent to a home for outdated or decrepit theories. Notwithstanding the recent accession of India and Pakistan to the nuclear club, the interstate war between NATO and Serbia, and the all-but-inevitable proliferation of weapons of mass destruction, other theories and other concepts are required to explain conflict and cooperation, and to guide policy, in the postmodern era.

Admittedly, the world is different from what it was during the heyday of the Cold War, and the international system will undoubtedly continue to change in the twenty-first century. But this does not mean that deterrence is dead, either as an objective, as a policy, or as a theory.