An important reason for studying networks is to understand how networks affect behavior. People often form relationships because they wish to be connected to other individuals for economically relevant reasons like the benefits of collaboration, exchange, and sharing of information. This endogeneity poses a huge challenge in analyzing how network structure affects behavior. Although selection and endogeneity issues are well-known to economists working with observational data, the issues in network settings can be acute. The aim of this research proposal is to tackle this key problem by proposing three different field experiments where the network formation can be considered as “random” and thus exogenous. The first project studies network effects on worker productivity and on-the-job-training. For that, we will use a unique dataset about an in-house call centre of a multi-national mobile network operator in the Netherlands. The exogeneity of the network stems from the fact that co-workers work at different hours and days during the week and thus the time they work together and interact can be considered as random. The second project will investigate how being randomly assigned to a group of students which centrality (which measures their popularity in friendship networks) is high can affect the group and individual grades of students in Bangladesh. The last project will focus on political opinions in an elite school in France. In this school, freshmen are randomly allocated to study groups for one year. We investigate whether being randomly exposed to people who have different political opinions can change own political opinion.