A Comparison of Methods to Quantify Homophily in Social Networks
Individuals in social networks have a tendency to form connections with others who are similar to themselves in some given manner. This phenomenon, called homophily, can be found among many attributes of individuals such as age, education level, gender, race, and socioeconomic status (Lazarsfeld & Merton, 1954; McPherson et al., 2001). Numerous measures of homophily exist, so this project aims to serve as a guide for five commonly used measures. Using both simulated and real-world network data, these homophily measures were compared under conditions varying in network size, the level of homophily, the proportion of individuals possessing the attribute of interest, and the ratio of the average connections of the individuals that do and do not possess the attribute of interest. Each metric was at least slightly sensitive to changes in the previously described conditions. Additionally, some of the measures contradicted one another regarding the presence of homophily in a given network and assigned different rank orderings of the homophilies of multiple networks.