Opening Up the Black Box of War

What factors influence victory or defeat in war? Why are some wars longer and more costly than others? How does public support for a war change during a conflict? Although these types of questions dominate approaches to military history, until recently they have played a relatively small role in influencing how political scientists study war. Realist approaches to international relations, with their emphasis on system structure, suggest that wartime decisions and behavior are largely unimportant in determining the nature of politics during and after a conflict. Rather, realists view war as a scale that balances bundles of prewar capabilities (Waltz 1979; Morgenthau 1954). Given the dominance of the realist paradigm during the past 40 years, studies of behavior within wars are rare compared with analyses of the initiation of conflict. Siverson and Starr (1991, 2) write that traditionally, the study of war has almost exclusively been concerned with the subject of initial war outbreak. Licklider (1993, 7) states that for both civil and interstate wars, their causes have been extensively studied while the equally interesting question of how and why they end has been, for the most part, curiously neglected. During the cold war, the study of conflict largely meant the study of how to avoid conflict initiation (i.e., deterrence), and scholars generally ignored intrawar behavior. Even the Correlates of War data set of interstate wars treats each war as an exclusive, single observation (Small and Singer 1982). Furthermore, most challenges to realism, such as the democracies-don't-fight literature, fail to examine wartime behavior. As a result, most studies of conflict have black-boxed war, treating it as a single aggregated observation with no disaggregated or internal dimensions. Those studies of wartime behavior that were produced tended to be either case specific (Kecskemeti 1958) or descriptive (Ikle 1971), and only a few general, rigorous analyses of wartime behavior were conducted from 1960 to 1990 (Schelling 1966; Klingberg 1966; Mueller 1973; Bueno de Mesquita 1978; Wittman 1979; Morrow 1985).