Fig. 1: Relationship between an individual’s SES and friends’ SES.

a, The mean SES rank of individuals’ friends versus their own SES percentile ranks. The series in green circles is calculated using the entire friendship network for each individual. The series in orange squares is constructed using each individual’s ten closest friends based on the frequency of public interactions such as likes, tags, wall posts and comments. SES is constructed by combining information on 22 variables to predict median household incomes in individuals’ residential block groups and then ranking individuals relative to others in the same birth cohort (Methods: ‘Variable definitions’). b, Comparison of estimates of homophily in the Facebook data and the Add Health survey. The series in purple squares plots the mean parental income rank of children’s friends against their own parents’ income percentile rank in the Add Health data. The series in green circles presents the analogous relationship in the Facebook data using our SES proxies, restricting the sample to individuals born in 1989–1994 and using their five closest friends from high school to match the Add Health sample as closely as possible (Supplementary Information A.5.2). For each series, we report slopes estimated from a linear regression on the plotted points, with heteroskedasticity-robust standard errors in parentheses.