Testing random assignment to peer groups

Identification of peer effects is complicated by the fact that individuals under study may select their peers. Random assignment to groups has proven useful sidestep such a concern. In absence formal randomization mechanism, it needs be argued “as good as” random. This paper introduces simple yet powerful test do so. We provide theoretical results for this test. As by-product, we equally obtain an approach popularized Guryan et al. (2009). These help explain why suffers from low power, as been observed elsewhere.