Abstract:    Scientific research needs stable and sustainable funding. The funding policy determines the allocation of scientific research investment resources, and then affects the efficiency of funding. Therefore, the funding policy plays an important role in scientific research investment. Based on a quasi-natural experiment and a difference-in-difference method, the thesis explores the impact of NSFC′s policy of increasing funding on scientific research output. 
    We take general project of NSFC as an example in 2011, and make an empirical analysis on physics. We also use "number" and "number + member + keyword" two different retrieval methods to screen the project output data to test the "net effect" of the change of funding policy after excluding the irrelevant output. It is found that the increase of NSFC funds has a significant policy effect on the impact of the research output, and notability improves the quantity and quality of output. After excluding irrelevant output and considering the "net effect" of the policy, the increase of funds has a more significant impact on the quality of project output. 
   We make the heterogeneity analysis on the experience of principal investigator. With "number" retrieval method, the increase of funds can significantly improve the scientific research output (quantity and quality), but it has a stronger impact on the "young" researchers who have won only one project. Excluding irrelevant output with "number + member + keyword" retrieval method, the increase of funds will increase the number of directly related output for the "young" researchers who have won only one project. On the contrary, the increase of funds improves the quality of output for the "senior" researchers who have won projects for many times. 
    We make the heterogeneity analysis on the institutional type. We find the increase of funds improves the outputs for the top institutions (985 universities and Chinese Academy of Sciences) and for these two types of institutions, the "net effect" of research quality is more significant than quantity after excluding irrelevant output with "number + member + keyword" retrieval method. After excluding irrelevant output and considering the "net effect" of the policy, the increase of funds has no significant impact on the directly related output of the project for other institutions (211 universities and general universities). Therefore, removing the phenomenon of "nominal", there is a function of "the strong is always strong" in the "net effect" of the policy of increasing funds. 
    We make the heterogeneity analysis on the team size. With the increase of funds is more obvious for the improvement of the output quantity of large teams, and more significant for the improvement of the output quality of small and medium team with "number" retrieval method. After removing the phenomenon of "nominal", only the medium scale team will benefit from the policy of increasing funds. 
    In order to improve the policy funding effect, we can appropriately increase the scale of funds and guide scientific researchers to "focus on quality rather than impulse" academic publication. We should implement fine and differentiated funding, and improve the accuracy of funding policies. 
   We can carry out gradient funding for different groups to general projects. For example, for "senior" researchers who have won many times, we can increase the funding size and form a "cumulative effect". For "young" researchers with little experience, we should carry out diversified funding. For researchers of top institutions, we should appropriately increase funding size to highlight the "Matthew effect" and increase exchange and cooperation projects between top institutions and other institutions in order to accelerate the mobility and sharing of scientific research resources. 
    We also need to pay attention to the work of team size and team structure in fund evaluation. At least, NSFC should show solicitude for the correlation between research results and fund projects at the time of project application and conclusion. Moreover, we should also further improve the achievement management of NSFC, get rid of "nominal" phenomenon and pay more attention to the actual output of projects.

Key words: science foundation, funding scale, scientific research output, difference-in-difference, policy "net effect"