Statistical Models of Social Fluidity

Abstract

Hypotheses about social fluidity in log-linear models of social mobility are hypotheses about patterns of odds ratios; yet odds ratios are only indirectly related to the interaction parameters of log-linear models. I propose rewriting such models so that the interaction parameters are equal to log odds ratios. This allows straightforward tests of difference and similarity between the odds ratios of different mobility tables. The approach is illustrated using three cross-nationally comparative data sets.