Abstract:This paper aims at solving the important problem of how to optimally allocate the limited R&D subsidy funds between the innovation “main force” (enterprises) and the innovation “national team” (scientific research institutions). By constructing a three-stage dynamic game model involving “government-enterprise-scientific research institution”, the optimal allocation strategy of government subsidies is explored theoretically, and the influence of a series of factors such as industry scale, industry-research cooperation and technology spillover on the formulation and implementation of government R&D subsidies is explored. The research proves that only when the spillover effect reaches or exceeds a certain level, the government will implement R&D subsidy to enterprises, and with the increase of the spillover effect, the government will increase the proportion of subsidy budget funds to enterprises. Industry-research cooperation not only strengthens the willingness of market enterprises and research institutions to invest in R&D, but also improves corporate profits, consumer welfare and social welfare level. A higher degree of industry-research cooperation will enhance the government's willingness to subsidize R&D to research institutions and reduce the government’s support to enterprise R&D.