In this paper we estimate peer group effects on academic performance based on the grit of the study network, taking into account information on socioemotional skills and networks of first semester students from one of the best universities at the academic level in Colombia. To identify the peer group effect, we propose using an instrumental variables model in order to solve the endogeneity problem reflected in the first-degree peer network, and using the second-degree peer network as an instrument. The results suggest that the grit of the peer network increases the student's academic performance by 0.11 standard deviations. Similarly, we find significant heterogeneous effects of the estimate between males and females, the Big-5, other non-cognitive factors such as intelligence malleability and self-control, by subject area, and school characteristics. In addition, we studied peer group effects considering students' academic behaviors. Finally, we conducted a robustness test using the third-degree peer network as an alternative instrument, given the closeness of the network in second-degree with the network of friends. These results indicate that the grit of the study network behaves as a good determinant of academic performance.
