Discrete-time logit model as one type of event history analysis has become applied in the studies on social stratification and mobility. However, in these studies the hazard probability and the coefficient estimated about one occupational category of occurred event are not comparable with those about another occupational category of event, because the definitions of survival which means nonoccurrence of the event are different each other. This paper proposes a discretetime multinomial logit model which estimates comparable hazardprobabilities and coefficients based on the same definition of survival.
The effect of coefficient estimated in logit models has been interpreted and explained as increased or decreased ratio of probability by odds ratio. However, another interpretation and evaluation would appear when the effect of coefficient is interpreted by the criterion of increased or decreased difference of probability. The latter half of this paper examines the difference between when interpreting coefficients by odds ratio and when by increased or decreased difference of probability, using the result of data analysis. The examination shows that the two criterions of interpretation bring different evaluations about the effects of coefficients.

特集 : 人間科学#寄稿論文