Abstract:     Since the implementation of the innovation-driven development strategy, China has increased support for technological innovation year by year, and successively introduced various policies such as direct subsidies, pre-tax additional deduction policy and preferential tax rates for high-tech enterprises. In this context, the consistency and sustainability of policies issued under different time and backgrounds should be considered. Previous studies have examined the effect of R&D policies on R&D activities based on samples at the company, regional, or industry level. Although some literatures have introduced spatial correlation into the research of R&D policies or R&D activities, few studies consider the spatial correlation of R&D policies and R&D activities at the same time. The first law of geography indicates that geographically adjacent units tend to have stronger correlation. Researchers have pointed out that China′s government R&D subsidies have a significant positive dependence. Correspondingly, the unbalanced distribution of regional R&D activities in China has continued to exist in recent years. It can be seen that the phenomenon of the first law of geography also exists in the process of R&D policies supporting R&D activities. If each region is regarded as an independent unit and the spatial correlation between regions is ignored, it is easy to produce bias in the estimation of the true effects of R&D policies. Therefore, it is urgent to accurately capture the incentive effect of R&D policies in the innovation process from the perspective of spatial correlation and explore the spatial interaction of R&D policies and R&D activities. To this end, what is the spatial impact of a single policy after considering the spatial correlation? Do the policies of a region have spatial spillover effect on the neighboring region′s R&D activities? What is the impact of policy mix when a company benefits from multiple policies at the same time?
     On this basis, using the data of large and medium-sized industrial enterprises in 30 provincial-level regions in China from 2009 to 2015, this paper employed the spatial econometric methods to reveal the spatial correlation of R&D tax policies and R&D activities. This article analyzed the impact of R&D tax policy mix on R&D activities by introducing spatial econometric models to the knowledge production function. The study delivers the following results: (1) There is a strong spatial-autocorrelation in R&D tax policies or R&D activities; (2) Both R&D tax policies have significant positive effects on the R&D input or output after considering the spatial relevance. However, the impacts are significantly weakened compared with the research without considering spatial correlation; (3) From the perspective of policy mix, when the impact of two tax policies on R&D activities is considered simultaneously, its impact and significance are weakened compared with the research only considering a single tax policy; (4) From the perspective of spatial spillover effects, R&D activities have significant positive spatial spillover effects. Meanwhile, direct tax credits have significant positive spatial spillover effects on R&D input. 
    Relevant research results not only shed new insights on how geography matters in the process of R&D tax policy supporting R&D activities, but also provide in-depth understanding on the influence of different R&D tax policies at different stages of R&D process. The results of the paper can provide a useful reference for policy making to design effective policy mix, thereby enhancing policy efficiency and China regional innovation capabilities.

Key words: R&D tax policy mix, Spatial Durbin Model, R&D input and output, regional innovation