Assessing resource allocation in R&D organizations is an important issue that requires a comprehensive measure to characterize it. To provide a greater picture, we first construct a dynamic three-stage network DEA model, which evaluates the R&D efficiency, technology-diffusion efficiency, and value-creation efficiency of Taiwanese R&D organizations over the period 2005–2009. Before integrating window analysis and network data envelopment analysis (DEA) to estimate dynamic efficiencies, we apply Analytic Network Process (ANP) to determine the relative importance of each stage. Subsequently, we employ panel data regression to examine whether the capital stock of patents, quality of human resources, and capability of service support affect the dynamic efficiencies of the R&D organizations. Our findings show that the mean R&D efficiency score is greater than that of the technology-diffusion efficiency, with the value-creation efficiency score being the lowest, suggesting that R&D organizations have to firstly work on improving the technology-diffusion inefficiency, and finally improving the value-creation inefficiency. Our panel data regression analysis indicates that the capital stock of patents do affect the efficiencies of the R&D organizations, even including the quality of human resources and capability of service support. That is, managers should focus on technological development and innovation to improve their corporate performance.