Southern Methodist University

We extend the model of friendship networks developed by Brueckner (2006) in two ways. First, we extend the level of indirect benefits by incorporating benefits from up to three links and explore its implication for the socially optimal and individual effort levels. Next, we generalize the magnetic agent problem by allowing for more than 3 players by restricting ourselves to regular networks that include payoffs from the magnetic agent. The authors would like to thank Pascal Billand, Christophe Bravard, and Daniel Millimet for their helpful suggestions. The paper has also benefitted enormously from the comments of the anonymous referee and the editor. Sudipta Sarangi gratefully acknowledges the support o