Face-to-Face Meetings, Worker Mobility, and Referrals

We study how face-to-face interactions shape worker mobility through social networks. Using granular cellphone geolocation and sociodemographic data on 3.3 million workers in a large urban labor market, we exploit settings in which multiple friends of the same worker relocate to jobs within the same destination area. Within the same worker and destination, face-to-face interaction with a friend increases the probability of moving to that friend’s workplace by 36-43 percent relative to remote communication, with effects substantially larger than those of phone calls or digital messaging. Consistent with referral models, the effect emerges only after the friend joins the destination firm and disappears in pre-move placebo periods. A data-driven heterogeneity analysis using regularized machine learning reveals a pronounced targeting gradient: effects in the highest predicted-return settings are more than twice as large as the population average. We interpret these findings through a simple referral model in which interaction technology shapes the effectiveness of information transmission, showing that mobility depends not only on the presence of social ties but on how information flows within them.