Political participation has long been viewed as a social act. But the influence of social relationships on participation is often impossible to disentangle from the factors that select people into these relationships. To overcome this challenge, we study randomly assigned college roommates, thus reducing these selection biases and other confounds. We examine short-run social influence of roommates on voter participation in 2016 and longer-term effects in the 2018 and 2020 elections. We collected consent from over 2,000 first-year students, allowing us to obtain a matched voter file indicating which students voted and the public voting histories of students’ parents, an indicator of students’ pre-college political environment socialization. Our evidence suggests that roommates’ influence on turnout decisions rivals the association between students’ turnout and that of their parents. Yet this parity masks gender differences. For women, the effect of roommates is larger. For men, the student-parent association exceeds the roommate effect.
