1998 年 13 巻 p. 217-225,275
We usually choose one of two kinds of data on analyzing voting behavior in a certain area. One is survey data which is collected by researchers. The other is aggregate data which is prepared by using statistics. We know quite well the advantages of each. And selecting each data naturally leads to fixing some procedures and variables in many cases.
I intended to use both, and feed all 9 variables into a computer.
In this paper, I have aimed to predict an election in terms of three categories, that is, characters, local political activities, and political environment of the French deputies. I judged the most effective factors of these, which had controlled each other.
I calculated the following result of a logistic regression analysis: z=-1.6875+0.6308 (passed)+2.7082 (left)-0.1767 (long)+1.3034 (mayor)-0.9164 (prefect)+0.5869 (community)+0.4133 (decent.)-0.5868 (activity)-0.1108 (money) And r=0.2143 (passed), 0.1668 (mayors), 0.1370 (left), etc.
From these figures, it appears that the most important factors are characters of the French deputies. Compared with other elections, the 1997 French legislative election is not special. I think that we can apply this implication to other legislative elections in France.