Gartzke and Li (2003b) formulate a mathematical relationship between 鈥渢rade share,鈥� 鈥渢rade dependence,鈥� and 鈥渢rade openness,鈥� and use this to argue that the disparity between the findings in studies by Barbieri and those of Oneal and Russett and others can partly be explained by features of variable construction. This article argues that although Gartzke and Li's formulation is very useful, the authors draw some conclusions from their analysis that only hold under very restrictive assumptions. 鈥淭rade share鈥� does not in general measure a state's disconnectedness from the world trading system, and 鈥渢rade share鈥� is not associated with a high risk of interstate dispute if 鈥渢rade dependence鈥� is associated with a low risk. What Gartzke and Li attribute to 鈥渢rade share鈥� is rather an effect of trade substition from one partner to another. Finally, the trade openness variable may be a proxy for country size. The discussion also illustrates some problems involved in controlling for variables that are related by definition to a key explanatory variable.
Hegre, H氓vard (2005) Identifying How Trade Matters in Empirical Studies of Interstate Conflict, Conflict Management and Peace Science 22 (3): 217鈥�224.
