2010 年 77 巻 p. 225-242
Some of recent media studies pointed out the increasing of media effect in politics and public opinion. In what meaning, why and how has the media effect increased? The aim of this paper is to demonstrate the time-varying effect of media coverage on the cabinet approval rate by analyzing aggregate time-series data. I examine two hypotheses about factors that increased the media effect on the cabinet support rate in Japan. Hypotheses I; the growth of floating voter, what is called on "mutouha", who are sensitive to political information made the media effect increase. Hypotheses II; Prime Minister Junichiro Koizumi who used media for getting support of electorate make the media effect increase, what we call "the Koizumi effect". In examining the relation between the media coverage and the change of the cabinet support rate, I introduced "the sentence-final modality" model as the new method of the contents analysis. The method is used for specifying positive/negative information about prime minister or cabinet in editorials of newspapers and converting its information into positive/negative score. In addition, I adopt the recursive regression method for analyzing time-varying effect of media. I can acquire three findings. First, the positive/negative evaluations in media coverage make a clear effect on the cabinet support rate. Second, the media effect has been significant after 1993 when floating voter grew rapidly. Third, "the Koizumi effect" is not able to confirm in this analysis.