Is Grand Strategy Predictable Appendix and Data

This study used three different models with different regression techniques to analyze the effect of relative economic and military power on grand strategy choice. The first model used a time series logit analysis to look at the effect of a series of lagged independent variables on whether a power adopts a pro or anti status quo posture. The second model used a time series multivariate OLS regression model to analyze the effect of the same independent variables on the risk of major war which a great power assumes the following year. The third model combined the first two models by using a robust multinomial logit analysis to examine the effect of the lagged independent variables on the odds of a nation adopting each of the four overarching grand strategies the following year.The dependent variable for the first model, whether a country is Anti-status Quo, is a dummy variable coded 1 if a nation is attempting to dramatically change its place in the current world order (ex. going from minor to major power, or going from major power to hegemon) at the expense of another power, or if a nation makes significant efforts to isolate itself from the current world order. The dependent variable for the second model is the Probability of a Major War, generated from a time series multivariate logit analysis using a dummy dependent variable coded 1 for years a country was involved in a major war, and independent lagged dummy variables coded 1 for years a country used the following strategies: Voluntary War, Binding, Blackmail / Subversion, Bloodletting, Internal Balancing, External Balancing, Buck passing / Retrenchment, Appeasement and Bandwagoning. Finally, the third dependent variable, Grand Strategy, was coded 0 for Legalist, 1 for Isolationist, 2 for Interventionist and 3 for Revisionist. The Grand Strategy variable represents a simplified version of Trubowitz’s typography of grand strategies found in his book Politics and Strategy, with his listed strategies condensed down to four overarching strategies. With respect to the independent variables, all three models use lagged versions of the following variables: relative military expenditure, relative military personnel, relative GDP, relative GDP per capita, nation age, homeland vulnerability, government type, economic system, and polity 2 score. These variables were either taken directly or derived from publicly available data provided by the Correlates of War Project, Polity IV Project, and OECD. Each of the sub strategy variables was coded based on the historical record. Other potential measures were considered, such as years great power, exports, and others, but many of these were dropped after correlational analysis indicated significant overlap with other variables. Finally, when appropriate, robust regression techniques were used in order to address potential problems with heteroscedasticity, auto correlation and others. <br>