*Digital matching platforms promise to reduce frictions on the labor market. In this study, we propose to evaluate the impact of a communication campaign by email designed to encourage the use of an online matching platform maintained by the French Public Employment Service, Pôle emploi. The intervention consists in sending emails to job seekers, providing information, help and motivation to register or update their profiles on the platform. We propose an innovative experiment design, where we optimize the take-up rate and discover the best e-mail allocation within the experiment using contextual bandits. Our design also allows us to derive results on the impact of the platform's use on labor market outcomes. This presentation focuses on the design and the analysis of the first pilot results.

**The endogeneity of network formation has been a major obstacle to the empirical study of peer influence for many important types of networks, including friendship networks, buyer-supplier networks, banking networks, etc. This paper puts forward the first causal identification strategy in the literature to study the effect of non-randomly formed peer relationships. I prove that causal identification holds under general conditions and needs neither a network formation model nor an outcome model to be specified. This is because the propensity scores of the unobserved confounders can be non-parametrically identified and estimated from the distribution of network links. Using the proposed method, I empirically estimate the causal effect of high school friendships on female students’ bachelor’s degree attainment. While previous literature finds that being exposed to more high-achieving boys in high school makes girls less likely to obtain a bachelor’s degree, I show that this is not true when these high-achieving boys are considered friends by the girls. In fact, one additional high-achieving male friend increases the probability that a female student graduates from college by 3 p.p. Further analysis suggests that this positive influence is not a result of increased academic ability but rather comes from a significant confidence boost. These results imply that rather than shielding girls from high-achieving boys, it would be more effective to foster friendship and close interactions among them.