Abstract: Patent and paper are important scientific outputs for research institutes (RIs). Utilizing NB and ZINB count model to control the possible biases from the zero-inflated and over dispersion, we for the first time evaluates the impact of various of S&T inputs and characteristic variables on the output of patent and paper based on a panel dataset consisting 1813 of China’s research institutes from 1998-2005. We obtain three results: (1) S&T personnel without graduate degree is not significantly promote the output of patent. Among the other three S&T inputs with the positive estimates, S&T personnel with graduate degree have a larger impact on the patent output. The impact of S&T intermediate inputs on paper is not significant. Among the other three S&T inputs with the positive estimates, high-level S&T personnel have a larger impact on the paper output. (2)Compared with the RIs supervised by local government, RIs by central government publish more papers, and apply less patent. Compared with RIs located in the eastern region, RIs in central and west region apply less patent. Paper publication is not significantly different among the three regions. RIs located in manufacture industries have more patents and less papers, and RIs in public service industry have less patents and more papers. (3)Compared with the non-converted RIs, S&T enterprises apply more patents to satisfy the demand of innovation market, however, NPR and NPNR, which focus more on basic R&D and public services, significantly increase their publication.