2019 年 95 巻 p. 15-25
Statistical causal inference refers to the effort to clarify causal directions,
which go beyond simple correlation. Causal inference using survey data is an
extremely useful analytical tool in social science, which is a comparatively difficult
subject for carrying out experimental research. Its techniques are becoming
more sophisticated and its applications are rapidly expanding. However, in
media studies, analysis of statistical survey data utilizing causal inference is
rarely carried out, especially in Japan.
In this article, in an effort to deepen the understanding of causal inference
for the readers, the author investigates what significance it holds with regard
to media studies. In addition, the author explains the basic concepts and the
analytical framework of causal inference using longitudinal panel survey data.
Panel survey data has many advantages in carrying out causal inference, but
there are also several weak points. In contrast, though cross-sectional survey
data is generally regarded as less suitable for causal inference, it has a practical
advantage to compensate for the weak points of panel surveys.