Best studies and empirical evidence strongly supports diversionary theory---especially for U.S. declines

In addition to advancing the theoretical literature on diversionary war, this paper tests these hypotheses with far better data than previously available. The new data for this paper consists of over 30,000 US diplomatic events between 1945 and 2006 drawn from a new corpus of all New York Times articles on foreign affairs over this period. Events are classified into 20 specific types (such as praise, threaten, or coerce) and four aggregate categories (material cooperation, verbal cooperation, verbal conflict, and material conflict). Verbal actions are those that occur in speech only (such as appeals or rejections) while material actions are those that involve physical action (such as the provision of aid or the exhibition of force posture). This represents the most detailed and comprehensive event dataset for this period. Its granularity allows for a better examination of domestic coalitional effects, and for an examination of diversionary cheap talk for the first time. Though other datasets like Militarized Interstate Disputes 3.0 under the Correlates of War project11 have greater historical range, they are censored in the sense that they only contain major events. While other day-level international relations event datasets exist, most notably King and Lowe (2003), they lack historical breadth, with most extending back only into the 1990s. Therefore, this new dataset is an advance in terms of detail and range. Modeling choices include fixed effects regression and Bayesian vector autoregression. I find strong empirical support for the theories outlined above. For Democratic administrations, a one percentage point increase in unemployment is associated with 12.85% more verbal conflicts per month than the baseline average (significant at the 1% level), but no more material conflicts. Unified government is associated with over 25% more verbal and material conflicts per month than divided government for both Democratic and Republican administrations (significant at the 1% level). Throughout, the estimates presented in this paper are conservative: I operationalize conflict as events the United States participated in, but if the sample is restricted to international events the United States initiated, the point estimates are larger and the statistical significance greater. This paper examines only the US case due to the wealth of data available on events, economic conditions, and public opinion ratings. However, future studies might broaden the analysis to OECD countries more generally. The broadest inference—that elected leaders verbally divert abroad to satisfy their core domestic constituency through rally around the flag effects, while avoiding paying the costs of war—should have theoretical traction in all democratic party systems. In the remainder of this paper, Section 2 reviews the literature, Section 3 advances a theory of diversionary behavior, Section 4 introduces the data, Section 5 presents the empirical analysis, and Section 6 concludes 2 Existing Literature Testing the diversionary aggression hypothesis has become somewhat of a cottage industry in political science. A recent review concludes that though the internal logic of diversionary war is “compelling and theoretically well supported,” the empirical evidence is “decidedly mixed.”12 Several studies have found empirical support for diversionary aggression in US foreign policy.13 A commonly cited example is President Reagan’s 1983 invasion of Grenada following the incident in Lebanon in which 200 Marines were killed. Other incidents that have raised scholarly interest are President Clinton’s strikes in Serbia, Sudan, and Afghanistan in 1998 and 1999 concurrent with his impeachment scandal.14 Other studies have found support for diversionary aggression in non-US and cross-national contexts.