In the article by Bennett and Stam, they present an event history model on war duration. The model incorporates both domestic political and real politics variables. The main focus of the article is to establish that factors such as government type, capabilities, terrain, and strategies play a central role of determining war duration. The authors use hazard analysis to test their hypotheses and to be able to support the main arguments. The study found out that war duration is strongly determined by real politics variables and not behavior or type of government. The authors note that there is little empirical work that focuses on war duration. The ones that are present use methodologies that are questionable. Most of them rely on data driven models which are considered to give little insight into war in addition to being seriously miss-specified.
The article reviews several literature materials with regards to the general model and the real politics approach. In addition, the authors also look at other approaches that do not generally suit into either ‘regime behavior and type’ or ‘real politics’. The authors draw their argument on two bodies of work. These bodies point out on the theoretical models of war duration. The first body of work focuses on quantitative models that predict duration of war based on real politics. The quantitative model includes variables such as level of military technology, military skill, military strategy, relative capabilities, and military exchange rates.
The second body of work focuses on the nature of political decisions to retreat during war. Researchers in this area say that the decision to join, continue fighting a war, or surrender from a battles is primarily political in nature. The impact of these decisions is felt all through duration of war. Due to the nature of the political decisions, the area has become an important point of focus for research on war duration.
Each of the approaches discussed in the literature review section have been subjected to the major variables under consideration. Under real politics, strategy has been described to influence the outcome of wars and also affect the duration of a battle. Strategy is categorized into three, punishment, attrition, and maneuver; each is influences war duration differently. Terrain also determine outcome of war in the sense that wars fought on a dense terrain tend to last longer than those fought on a flat terrain. Other variables identified to influence war duration include interaction between terrain and strategy, balance of capabilities, population, mobilization, issue salience, and military quality. With regards to types and behaviors of regime, repression and democracy offer varying influence to the duration of war.
The test the hypothesis on war duration, the authors have applied hazard analysis. They focused on the wars that started between 1816 and 1985. Hazard method is appropriate for the analysis because data on duration of event is normally positive. This rules out the appropriateness of other methods of analysis such as OLS regression. The hazard model effectively estimates the instantaneous rate for the occurrence of transition. The hazard model is also specified by parametric function forms that include gamma distribution, log-normal, normal, weibull, and exponential distributions. The choice for a duration distribution is based on the preliminary data, convenience, and economic theory. However, Weibull duration distribution is most preferred for some various reasons.
The result indicates that the model used in the study can and actually does make predictions of both long and short wars, although it is more accurate predicting short wars than long wars. The individual effect of both models has been proven to have statistical significance. Most of the empirical effects of the variables under consideration were able to match the expectations of the researchers based on the literature review.
Bennett, D Scott and Allan C Stam. "The duration of interstate wars, 1816-1985." The American Political Science Review (1996): 239-257. ProQuest Central.