War is a rare event in world politics, but it is always with us. How can we say this? On the one hand, if we note that the number of territorial states in the global system has ranged from fewer than 30 after the Napoleonic Wars to nearly 200 at the end of the twentieth century, that gives us about 400 nondirectional pairs of states in 1816 and about 18,000 pairs today. And even if we recognize that most wars are between neighbors, and thus reduce the possible pairs at war in a given year to the 40 bordering neighbors in 1816 and the 317 in 1993, the potential is never even approached. There were no wars under way in 81 of the 180 years since the modern interstate system came into being, and seldom more than one in any given year. On the other hand, 75 interstate wars led to a total of more than 30 million battle-connected fatalities among combatants, not to mention tens of millions of additional deaths among civilians in the context of these wars.

Another way to make this point is by looking at the frequency of war involvement by the members of the interstate system.

Introduction

In 1961, Singer noted that the analysis of the international system provides the most comprehensive level for the study of interstate relations. He observed that formation and dissolution of alliances, shifting capability distributions among major powers, and the effectiveness of international organizations and norms all could be subsumed within system-level study. However, 30 years later, Wayman and Singer (1991) state that the production of substantive data-based findings on associations between system-level characteristics and war has been less successful than dyadic-level studies. While this observation is valid, it is nevertheless possible to extract a series of consistent and cumulative empirical findings on war from the quantitative literature on international systems (references listed in appendix 2, table A.4). Many of the difficulties associated with early systemic-level studies stemmed from conceptual ambiguity in the construction of measures or the misspecification of variables. However, these problems do not fundamentally detract from the set of important observations on the factors associated with war-prone international systems.

Polarity/alliance

The literature of international politics is rife with discussions on the effects of the “polarity” of the international system on war. Unfortunately, due to the ambiguity of the concept of polarity and the multiple ways in which it can be operationalized, confusion rather than clarity has characterized the debate. The essence of this confusion involves the question of whether the polarity of the international system is determined by the number of major states or by the number of distinct clusters of states emerging from the configuration of alliances.

Introduction

The onset and seriousness of World War I (1914) is explained on the basis of empirical uniformities established by systematic quantitative analyses. Scientific explanations of individual events may be provided through inductive subsumption under probabilistic laws, as well as through deductive subsumption under universal laws. This inductive explanation demonstrates that the case of World War I is a specific instance of a set of patterns which have appeared in a much larger number of cases.

World War I conforms to a set of probabilistic laws based on empirical regularities identified at the state, dyadic, and systemic levels of analysis. The states involved in the onset of the war were contiguous or proximate major powers (Austria-Hungary, Russia, Germany, France, and Great Britain); three of these major powers (Germany, Great Britain, and Russia) were moving through critical points in their power cycles; two of these major powers had an unstable capability balance (Germany and Great Britain had recently experienced a power transition); three of these major powers lacked democratic political systems (Austria-Hungary, Germany, and Russia); and the international hierarchy was unstable with a declining leader (Great Britain). Add to this the classification of the Germany/ France dyad as an “enduring rivalry” and the highly polarized alliance systems (Germany/Austria-Hungary and Great Britain/ France /Russia), and the occurrence of a war of enormous magnitude, duration, and severity was a high-probability event consistent with a broad array of empirical war patterns.

In 1970, Singer outlined his scientific approach toward uncovering the causes of international war and speculated as to the outcome. Specifically, he postulated that state attributes, relational characteristics within dyads, and system-level attributes might combine in creating a potent source of war:

The analysis presented here has attempted to identify those factors and to demonstrate how they conjoin in influencing the onset of war among nations.

The preceding eight chapters of this book were designed to provide a scientific explanation of war in international politics grounded on data-based empirical research. The study began with an examination of two basic decisionmaking models – the rational and the nonrational – in relation to the explanatory framework of the volume.

Introduction

Wars follow from decisions. Therefore, any explanation of war must incorporate, explicitly or implicitly, a model of decisionmaking. Decisionmaking models are distinguished by their assumptions about rationality. The two alternative assumptions undergirding these models are, simply, the rational and nonrational. Nonrational models, whether focusing on psychological variables or organizational interests and routines, maintain that decisions are frequently distorted by systematic perceptual, cognitive, or bureaucratic biases. Although the definition of “rationality” is a subject of both philosophical and conceptual debate, minimal criteria for an instrumentally rational decision would include the logical requirements of consistency and coherence in goal-directed behavior. Indeed, Singer (1981:14–15, 1989a:14) notes that:

Nonrational models focus on psychological or bureaucratic factors associated with individuals or organizations and often require detailed information on specific people or bureaucratic structures and processes, whereas rational models offer the simplifying assumption that psychological or bureaucratic biases have little impact and that decisionmakers (or decisionmaking units) all calculate in more or less the same way.