This paper compares research and development (R&D) efficiency of 23 countries by taking into account input congestion under the framework of data envelopment analysis (DEA). R&D stock capital and R&D manpower are inputs, while patents, scientific journal articles, and royalties and licensing fees (RLF) are outputs. Empirical findings are as follows. First, there exists congestion effect in the R&D activities of sampling countries. Second, European countries have the most serious R&D input congestion, followed by Asian countries and American countries. Moreover, non-OECD countries have more serious input congestion phenomenon than OECD countries. Third, Mann-Whitney U test shows that the R&D efficiency scores under traditional DEA model and Tone and Sahoo (2004) congestion model have significant difference. Fourth, the government action, intellectual property rights protection, knowledge transfer, public spending on education and economic freedom have positively impacts on R&D efficiency when there exists congestion.