<論文>コーホート分析における代数的要因分解法の多時点間比較への拡張

Abstract

As many sociologists argue, the birth of a new cohort and the death of an old cohort may have some effects on social change. This is called the cohort replacement effect, which can be estimated with several statistical methods, decomposing social change into cohort-replacement and individual-change effects. One of these methods is the algebraic decomposition method which Firebaugh proposed, applying Kitagawa's classical decomposition method to cohort analysis. As Firebaugh admits, however, the algebraic decomposition method has a limitation in that the estimation bias is large when it is applied to a long term social change. In this paper, we resolve this limitation through generalizing it with data sets covering three or more periods. We show that long term social change can be divided into several successive short term changes whose cohort replacement effects are estimable with little bias. The sum of these cohort replacement effects on the successive short term social changes is proved equal to the long term cohort replacement effects. We illustrate the advantage of our generalized method with a simple example.