Endogenous formation of coalitions in a model of a race

Endogenous Network Formation in Patent Contests And Its Role as A Barrier to Entry

“…Corollary confirms the results found in Goyal & Joshi [] and Joshi []: if the complete network (grand coalition) is formed, then no firm has an incentive to deviate. However, this does not exclude the existence of other pairwise stable networks.…”

“…First we show that our set up allows us to endogenize both R&D effort and network formation. Second, like Goyal & Joshi [] and Joshi [], we find that the complete network is always pairwise stable when the cost of forming a link is insignificant. However, when the benefit of forming a link (i.e., a partnership) is large enough, there exist asymmetric pairwise stable networks which have the dominant group architecture: a non‐empty set of firms does not participate in the contest while all participating firms form links with one another.…”

“…Their main finding is that the complete network is the only one to be pairwise stable when the linking costs are low. On the other hand, Joshi [], in a similar setting, studies a coalition formation game with endogenous R&D effort. Joshi thus answers the two questions simultaneously in a setting of strong cooperation: firms both share the prize and coordinate the R&D effort at the coalition level .…”

“…In the second stage, firms are involved in a patent contest which is influenced by the network structure realized in the first stage. Differently from Goyal & Joshi [] and Joshi [] we formalize such a stage as an all‐pay auction rather than a classical patent race . The reason behind this is threefold.…”

“…In order to perform this backward induction approach, the expected payoffs of the patent game need to be tractable. As has been pointed out by Joshi [], in a model of coalition formation in a classical patent race, when research intensities are chosen non‐cooperatively the equilibrium is no longer tractable. The all‐pay auction with private information allows us to circumvent this problem by using the results of Parreiras and Rubinchik [] and [].…”