Response to Hauser and Logan and Grusky and Van Rompaey

Abstract

Hauser and Logan's (hereafter H&L) response to has many facets, but its core is a statistical critique. Reduced to essentials, their claims are that lacks cross-sample validity and that the apparent increase in the intergenerational correlation is due to over-response to sampling error. But their attempt to demonstrate this is flawed. Symmetric scaling of intergenerational continuity is a procedure (or algorithm) for determining a hierarchical pattern from mobility data. It results in, as I state above, SSIC scores only for the data on which they are based, and other more or less similar arrays would arise from different samples. This means that comparison across samples requires new estimates. But H&L omitted this central step and thus failed to correctly grasp cross-sample performance. Direct estimation reveals four central contrary results. First, is empirically indistinguishable from application of a maximum-likelihood method due to Goodman (1979a, 1979b, 1981a, 1981b), and thus falls within a central tradition of mobility analysis. Second, application of to the OCGII data reveals almost identical summary estimates, which is hardly consistent with an absence of cross-sample validity.2