Decisions about college are highly consequential, yet they are often made with poor information. This paper studies how information about college programs spreads through peer networks and affects application behavior using data from Ontario, Canada. We build and estimate a structural model of application and enrollment decisions. Ontario’s unusual admissions rules and detailed microdata make all inputs into admissions decisions observable. This allows us to recast the NP-hard application portfolio problem as a tractable problem, by making a student’s admissions probabilities to different programs independent conditional on observables. We also leverage student interactions across cohorts and with neighbors who attend other high schools to resolve standard challenges in estimating peer effects. We find that the probability of applying to a college-major pair increases significantly if a peer in an older cohort has applied to or enrolled in that same program, with stronger results if the peer enrolls. These findings shed light on the consequences of racial and socioeconomic segregation in schools and the long-term evaluation of information interventions.