Modeling the Profiles of Japanese Independent Voters in the Early 1996

Abstract

This article illustrates that careful attention must be paid to “I do not know” (DK) answers when modeling the profiles and political inclinations of the Japanese independent voters. The DK answers, concentrated on answers to the questions regarding the support for the current cabinet and political ideology, are treated as missing because there is no valid reason to treat them as one separate category. We perform logistic regression analysis of binary response — whether the interviewee has a political party to support or not — on several political and economic explanatory variables, within which the EM algorithm is implemented. This enables us to incorporate the cases with missing-values into the analysis. We found that missing-value problem could not be ignored in the sense that marginally significant or insignificant explanatory variables can easily turn otherwise when the same model is fitted only for the data with no missing-values.