Models and trends of log-linear association

Abstract

Log-linear models afford a wide range of tests regarding specific aspects of the association between two ordinal-level variables with identical attributes. This range is considerably broadened when orthogonal contrasts are used to trace changes in these aspects over time. This paper provides descriptions and graphic illustrations of six common log-linear models' applications to mobility data on U.S. white males. The models are then expanded to examine temporal variations in aspects of the mobility regime captured by the respective models. The analysis shows a significant increase over the period of study in the odds that white males from blue collar backgrounds attain higher occupational statuses. Moreover, the findings indicate that blue collar workers are more likely to be drawn from lower status origins during years of economic prosperity, and from higher status origins at times of economic dearth.