Abstract
Occupation is a key factor in human thinking, feeling, and behavior. Theoretically derived occupational groupings or classes are typically used to transform occupations into a variable suitable for statistical manipulations. We argue that such groupings are unlikely to produce groups that are homogeneous across a broad set of attributes. Instead, we offer a data-driven approach to identify groups of occupations based on respondents’ mobility data using network analysis. The vertices of the network are codes of occupations, and the edges reflect the number of transitions between them. Using modularity maximization, we identify four communities and evaluate the stability of the resulting partition. As an example demonstrating the efficiency of the resulting grouping, we present a comparison of the predictive power of this grouping and one of the generally accepted groupings of occupations, that is ESeG (European Socio-Economic Grouping), in relation to the human attitudes and values found in previous publications. The results indicate the preference of our grouping.