We examine the impact of informal risk sharing on risk taking incentives when transfers are organized through a social network. A bilateral partial sharing rule satisfies that neighbors share equally a part of their revenue. In such a society, correlated technologies generate interdependent risk levels. We obtain three findings. First, there is a unique and interior Nash-equilibrium risk profile, and it is in general differentiated and related to the Bonacich measure of the risk sharing network. Second, more revenue sharing enhances risk taking on average, although some agents may lower their risk level. Last, we find that under investment might often be observed.

We study the formation of a directed communication network in which agents distribute a fixed amount of resource over links. Indirect benefits transit through the path maximizing the product of link strength. In this environment, the wheel architecture is shown to be both the unique efficient and the unique Nash architecture.

We study rival firms' incentives in quality-improving Research and Development (R&D) networks. The analysis stresses the role of free riding associated to collaboration and three major consequences emerge: R&D efforts decrease with the number of partners, networks of alliances are over-connected as compared to the social optimum and the profitmaximizing number of alliances is possibly non monotonic (decreasing then increasing) with respect to inverse measure of product differentiation.

In a vertically differentiated oligopoly, firms raise cost-reducing alliances before competing with each other. It is shown that heterogeneity in quality and in cost functions reduces individual incentives to form links. Furthermore, both differentiated Cournot and Bertrand competition qualitatively similar incitations to form alliances.