Alternative Mechanisms of Peer Achievement Spillovers: Implications for Identification and Policy∗

This paper provides a link between the theoretical and empirical models of peer spillovers in the education context, informing the interpretation and identification of the parameters of the empirical model and their applicability to policy. The point of departure is to consider why peer achievement spillovers in educational production. I compare two theories, that peer achievement proxies for unobserved peer ability (a predetermined characteristic) or effort (a simultaneous choice). While the literature does not tend to distinguish between these mechanisms, I show that the implications for identification and policy are quite different and that the tendency to focus only on spillovers from predetermined characteristics of students may be misguided. The theoretical model also helps to clarify the interpretation of spillovers from peer charac-teristics. I show that the marginal effect of peer characteristics may actually differ in sign depending on the underlying mechanism of peer influence, potentially helping to reconcile some of the mixed evidence regarding contextual effects in the literature. Fi-nally, I discuss the relevance of these results for policy, particularly focusing on optimal grouping