Explaining contact patterns in acquaintanceship networks: A new covariate-based model

Aggregate relational data (ARD) on individuals’ ties to groups in a population offer valuable insights into the size of personal networks and the extent of segregation in contact with those groups. However, existing ARD models face two key limitations. First, they mask heterogeneity in network sizes across individuals who may differ markedly in relevant characteristics despite exhibiting similar response patterns on the ARD instrument. Second, although existing models can measure the overall level of segregation in contact with groups, they cannot reveal the determinants of segregation. To address these limitations, we introduce the Covariate Model, a regression-based framework that incorporates respondent covariates into analyses of ARD. We illustrate this model using ARD on contact with occupational categories. In addition to obtaining more substantively plausible network size estimates than existing approaches, the Covariate Model uncovers novel segregation patterns. For example, covariates — driven primarily by educational differences — account for a considerable portion of the segregation in contact with Higher Service occupations (e.g., lawyers and professors) but contribute little to explaining barriers to interaction with other occupational classes. By modeling the determinants of contact patterns in acquaintanceship networks, the Covariate Model extends the analytical reach of ARD and opens new avenues for research on social capital and segregation. • Covariate Model uncovers heterogeneity in network size masked by existing models. • Covariate Model reveals the determinants of segregation in contact with subgroups. • Covariate Model provides improved fit over existing models. • Covariate Model is widely applicable to research on social capital and segregation.