SPECIFYING THEORIES IN COMPARATIVE POLITICS: TOWARD A MORE THOROUGHLY DEDUCTIVE APPROACH

Abstract

Models in political science are often poorly specified prior to testing. In practice, most analysts rely on regression analysis to determine the weights for each independent variable (causal factor) identified in the model. We demonstrate a method for determining the relative weights of causal factors prior to testing of the model. The approach makes use of expert opinions in the qualitative literature in order to construct a more completely specified model in a deductive manner prior to testing. We also demonstrate several methods for assessing the degree of confidence we might have in the model based on the relative degree of divergence among the experts concerning the causal factors.