Recent research finds that grade span affects academic achievement, but only speculates about the mechanisms. In this prepublication version of the study, we examine one commonly cited mechanism, the top dog/bottom dog phenomenon, which states that students at the top of a grade span ("top dogs") have better experiences than those at the bottom ("bottom dogs"). Using an instrumental variables strategy introduced in Rockoff & Lockwood (2010) and a longitudinal data set containing student survey data for New York City public middle school students, we estimate the impact of top dog and bottom dog status on bullying, safety, belonging, and academic achievement. This paper provides the first credibly causal evidence that top dog status improves the learning environment and academic achievement. We further find that the top dog effect is strongest in 6th grade and in schools with longer grade spans and that the top dog effect is not explained by new students to a school or student height. [This paper was published in "American Educational Research Journal" v53 n5 p1450-1484 2016.]