This paper suggests a procedure to control for independent variables in the measurement of segregation. It links the well known Index of Dissimilarity to the Multinomial Logit Model. This combination of the regression approach and the segregation approach turns out to be a useful tool in analyzing structures of social inequality. Using data from the German population census of 1970 and the German microcensus of 1995 the trend toward family-type assimilation on the part of immigrants in Germany exemplifies the technique. In this case the regression approach would lead to an ambiguous picture of the general development, while the standard segregation approach provides a dubious answer as it does not take into account demographic change. However, combining both approaches allows us to control for structural changes in the assessment of dissimilarity. In the end, this lends support to the hypothesis that the family types of Germans and of immigrants have converged over the last decades.

[No Keywords available]

[No Keywords available]