Structural Equation Models for Ordinal Variables

Abstract

This article presents structural equation models with discrete dependent variables that refer to occupational destinations. After assuming threshold measurement models and structural models with normally distributed error terms, it is shown that the linear relationships among observed and latent continuous variables can be treated in a way that is similar to conventional structural equation models in the LISREL framework. The models that are presented are simple cases of a large class of models that can be estimated with the computer program LISCOMP. Reanalyzing data from the 1962 Occupational Changes in a Generation Survey, the models incorporate discrete occupational destinations into the classic Blau-Duncan model. Results indicate that although the likelihood of becoming a manager, official, or proprietor is directly affected by parental social status, the chances of becoming a scientist or engineer are affected only indirectly through education.