How Not to Measure Intergenerational Occupational Persistence

Abstract

This article is an invited critique of Rytina's proposal that canonically weighted scales for detailed occupations be used to assess intergeneratinal occupational persistence. These canonical scales yield higher intergenerational correlations than the Duncan SEI primarily because they capitalize on chance fluctuations in the data used to construct them. Thus, there is no reliable basis for revising previous findings about the relative importance of ascription and achievement in the stratification process. Canonical scaling (and related log-linear and log-multiplicative models) are no more than a useful adjunct to models that locate the sources of intergenerational occupational persistence in the measurable characteristics of occupations and persons.