Structural equation modeling of political discussion networks

This study conducts structural equation modeling (SEM) of political discussion networks. It examines multiple relationships between political discussion networks—network size and non-kin composition, political efficacy, and neighborhood conversation. Based on a two-step approach, it first analyzes and revises the measurement model and then analyzes and revises the structural model given the revised measurement model. The proposed SEM model includes ordered categorical variables as factor indicators in the confirmatory analysis and outcome variables in the structural regressions. Traditional estimation and regression methods need to be adjusted accordingly. This study uses WLS estimation and adopts a latent variable approach to study the categorical outcome variables in the SEM. The results show that the hypothesized SEM model is fully supported. Neighborhood conversation positively and directly contributes to political discussion network size as well as the non-kin composition of the networks. It also indirectly affects network size through political efficacy. Political efficacy also has a direct effect on network size.