The Evolution of Cooperation in Patent Races: Theory and Experimental Evidence

In a dynamic patent race model we analyze the formation and breakup of joint ventures in relation to: (a) the relative as well as absolute position of the firms in the race; (b) the degree of competition in the ex post market. Fudenberg et al. (1983) studied the main features of a patent race when firms compete in R&D, showing that firms in the same position compete fiercely, dissipating the rent from innovation. By contrast, we show that if firms can cooperate or compete in R&D, and if they start in the same position, they cooperate at the outset but break their agreement in the last stage if they will be serious competitors in the downstream market, while, if they can collude in the ex post market, they cooperate from the outset and they innovate jointly. When the firms are lagged by one step, cooperation does not take place, except in the case the value of the race is negative and the cost saving due to cooperation is large. However, cooperation never occurs if the leader is more than one step ahead. Finally, when the firms cooperate in R&D they proceed to the discovery at low speed. We test these conclusions via experiments on the incentive to cooperate during the course of a race. The results of a sample of 86 races support our theoretical conclusions, although the experimental findings are less clear-cut than the theoretical ones.