Peace and Conflict Studies - Spring 2014

Peace and Conflict Studies Volume 21, Number 1 91 construct a scale of 1-21. I adopt the “weak link” principle in determining the effects of regime type on conflict as well and create a democracy low variable. As in the economic development variable, all data are lagged by one year. Control Variables. Capability ratio : To determine the capabilities of each country, I use the Correlates of War (COW) data (Singer & Small, 1995), which gauges the National Capabilities of states from their population, industry, and military forces. Capability ratio is calculated by taking the ratio of the stronger state’s military capability index to that of the weaker member in each dyad. A higher score indicates higher power discrepancy, or less power parity, in a dyad. The final variable Logcapratio , is the natural log of the capability ratio in a dyad. Alliance : The variable alliance equals one if countries A and B are formally allied through either a defense pact, entente, or non-aggression pact; it is zero otherwise. I use COW’s data on alliances. Major power : To control for the higher conflict-proneness of major powers, I use a major power variable, which equals 1 if a dyad includes at least one major power and 0 otherwise. The US, the USSR, Britain, France, and China are considered as major powers for the entire period I analyze; Germany and Japan are regarded as major powers after 1989. State age (longevity) : Several developing countries gained their independence during the time period covered by this research. Younger states are expected to focus on state building and internal problems and to avoid external conflicts. To control for lower conflict propensity in early statehood, I create a state age variable, which equals the number of years since independence for each state. As in other variables, I use the state age of the younger state in each dyad. The COW’s “state system” data specifies the dates for each state’s entry into the international system. Developed state : Lastly, to control for the possible distinct relationship between all- developing-state dyads and the mixed (developing-developed) ones, I include a developed state dummy, which equals 1 if a dyad includes a developed state and 0 otherwise. --- Because my data have a cross-sectional time-series nature, I introduce measures to correct or relieve temporal autocorrelation and cross-sectional heterogeneity. Following Beck, Katz, and Tucker’s (1998) suggestion to correct temporal dependence and using Tucker’s (1999) btscs program, I created a peaceyears variable, which counts the years since the last MID, and three cubic splines. Finally, I report robust standard errors clustered on