Comparing QCA Methods for Exploring Conjunctural Causation －From the Perspective of Type I and Type II Errors－

摘要

The existing literature in the field of social sciences does not analyze conjunctural (interaction) effects adequately because of the limitations in comparative methods. For example, statistical analyses, such as regression analysis, tend to commit type II error in examining conjunctural effects and overlook true conjunctural effects. Qualitative Comparative Analysis (QCA) has come to be employed in many studies as a method for exploring conjunctural causations, especially against nonlarge N cases. QCA, especially the classical form (crisp-set QCA: csQCA)，however, has been criticized for reasons such as its inability to include control variables and the dichotomization of variables. In other words, it has been criticized for committing type I error in testing conjunctural effects. In order to overcome this criticism, a developed version of QCA (fuzzy-set QCA: fsQCA) was introduced and has been applied in preceding studies. fsQCA allows us to employ nonbinary variables and is considered to be superior to csQCA. The purpose of this paper is to discuss whether fsQCA is better than csQCA for exploring conjunctural causations. From the perspective of type I and type II errors, the present paper claims that fsQCA is not always preferable to csQCA. As an illustrative analysis, the present paper analyzes an empirical research question concerning countries' frequencies in filing dispute settlements under GATT (General Agreement on Tariffs and Trade) /WTO (World Trade Organization) using the two methods. The results also provide an implication on comparative methods beyond QCAs; although comparative methods, especially statistical analyses, emphasize the importance of avoiding type II error and using detailed measurement, this may not always be the case.