A new approach for modeling mobility tables: Based on community detection methods

Based on community detection methods in social networks, a new approach is presented for understanding and fitting social mobility processes. As a community detection algorithm, eigenspectrum decomposition is applied to identify optimal community structures in which most shares of the relations are within community ties. Through this approach, one may construct a log-linear model by adding community effects to an independence model until the model fits the given data, where the community effects are derived from the community detection analysis of the residuals. To illustrate, this approach is employed in mobility tables from the General Social Survey. Consequently, a set of parsimonious models that fit the observed mobility tables is presented, which offers novel and interesting interpretations of mobility patterns.