This paper aims to fill the gap in the extensive literature on the results of interaction betweenpublic and private R&D expenditures. Two types of public R&D subsidies, the appropriationsand loan interest, are investigated in the effect of public subsidies by applying for a uniquemicro data set of Chinese manufacturing firms combined with the Innovation Fund forSmall and Medium Technology-based Firms (Innofund). The results are consistent with theprediction that loan interest subsidies based on more competitive selective system are moreeffective for the public R&D subsidies than the appropriations. Firms’ heterogeneity,measured by state-owned enterprises (SOEs) and private-owned enterprises (POEs), is testedfor the different impact of public subsidies on corporate R&D investment. The results showthat the crowding-in effect is mainly driven by POEs, not by SOEs, which confirms thetheoretical assumption by Aghion and Tirole that the agency problem play a dominant rolein corporate R&D activities. An improved econometric method, combined propensity scorematching with difference-in-differences estimation strategy, is applied for correcting theselection bias. Our results are robust to different matching techniques.