Correlates of War

David Singer and Melvin Small of the University of Michigan initiated the Correlates of War project in 1963. The basic strategy was to (1) define "war" operationally, (2) collect all sorts of statistical data about every war from the fall of Napoleon to the present, and (3) search the data for variables correlated with the outbreak of war.

Singer and Small operationally defined war as a “military conflict waged between national entities, at least one of which is a state, which results in at least 1000 battle-deaths of military personnel.” The original analyses used data from wars that took place from 1816–1965. Later analyses extend as far back as 1495 and as far forward as 1992. Data such as start and end dates, location, battle deaths, total deaths, pre-war population, pre-war armed forces, initiator, nature (interstate, colonial, or imperial war), and political status of combatants was collected. Civil wars were coded separately with data including dates, location, intervention by outside powers, their political status and on what side they intervened, who won, battle deaths, total population, and total number of pre-war armed forces. Virtually every hypothesis anyone could devise was tested, even astrological ones. The unsurprising answer in those cases was “no connection.”

The Correlates of War Project was a search for consistent empirical patterns rather than an argument for a particular one.

The major findings of the Correlates of War Project and the research studies stemming from it are of four types, beginning with those pertaining to individual states. The evidence reveals the lack of a connection between wars and business cycles, form of government, geographic size, population density, and total population. There is weak evidence of a link between domestic and foreign conflict, although there is no consistent cause from one case to another. Domestic violence may break out to protest a war but the government may try to divert domestic discontent by creating a foreign crisis. Dictatorships such as North Korea and Cuba have long used the tactic; President Clinton was satirized in the movie Wag the Dog for doing the same to distract from his personal behavior.

There appears to be a strong relationship between war and both capabilities for war and number of alliances. Major powers are more likely to engage in war than are minor ones, and they are more likely to fight severe wars. Highly militarized societies (customarily measured by the ratios of military personnel to population and defense expenditures to GNP) lean toward an aggressive foreign policy.

The second major set of conclusions pertains to pairs of countries. At the economic level, there exists the long-standing idea that there is a positive correlation between peace and economic development, free trade, and market economies. Correlates of War data combined with trade data from the UN and IMF corroborate this hypothesis. At the political level, researchers using the Correlates of War reach three contradictory findings. First, relative parity (that is, balance of power) deters war because victory becomes problematic. Second, relative parity encourages war because both sides see a chance for victory. Third, war is more likely under conditions of changing capabilities—that is, the erosion of a dominant nation’s capability or the rise of a new power. This leaves us in approximately the same position as Richardson’s work on arms research: anything is possible and there is little basis for predicting which is most likely.

The third major set of conclusions pertains to world regions, defined as the five most populated continents. Over the past two centuries, major rather than minor powers have been the most likely to be involved in military action, and those actions are more likely to be severe. In the decades since 1945, regional shifts in the frequencies of interstate conflict have become evident, with the newer regional subsystems of sub-Saharan Africa and Asia accounting for the majority of militarized interstate disputes. The thesis that wars occur in cycles of one sort or another disappears under scrutiny.

The fourth set of conclusions takes a systems view. Since 1816, the number of wars has increased but so has the number of countries, the ratio remaining constant. From the late fifteenth century to the last quarter of the twentieth, great power wars have decreased in number and increased in destructiveness. Some efforts, like the Peace of Westphalia and the Congress of Vienna, were successful in resolving issues that led to war and mitigating future conflict, but others such as the Versailles system have been disastrous, so that the prospects for keeping the peace through intergovernmental organizations are not hopeful.

As with any groundbreaking study, the Correlates of War Project has been both damned and praised for its operational definitions, the variables chosen for or omitted from the study, its chronological framework, its analytical methods and the hypotheses tested, which is to say just about every aspect of the study. These criticisms may be divided into technical and theoretical objections, the latter being the more important for our purposes.

The study was a search to identify variables that are the most significant for a study of war. That is, it echoes Galileo's effort (Chapter 1) to identify a few key variables with great explanatory power. It is limited to wars involving at least one recognized state. The technology of war changed radically over the period covered, and is changing even more so now. It considers wars fought for many different purposes by many different kinds of forces. Humans are likely to use anything science learns to their own advantage, making the study itself an independent variable. This is a significant problem for the study of all human behavior. Despite these difficulties, the Correlates of War is falsifiable, parsimonious, and usable. It remains to be seen whether it is generalizable. It has changed the way we study war and provided a starting point for other thinkers.