The counting of papers and citations is fundamental to the assessment of research productivity and impact. In an age of increasing scientific collaboration across national borders, the counting of papers produced by collaboration between multiple countries, and citations of such papers, raises concerns in country-level research evaluation. In this study, we compared the number counts and country ranks resulting from five different counting methods. We also observed inflation depending on the method used. Using the 1989 to 2008 physics papers indexed in ISI's Web of Science as our sample, we analyzed the counting results in terms of paper count (research productivity) as well as citation count and citation–paper ratio (CP ratio) based evaluation (research impact). The results show that at the country-level assessment, the selection of counting method had only minor influence on the number counts and country rankings in each assessment. However, the influences of counting methods varied between paper count, citation count, and CP ratio based evaluation. The findings also suggest that the popular counting method (whole counting) that gives each collaborating country one full credit may not be the best counting method. Straight counting that accredits only the first or the corresponding author or fractional counting that accredits each collaborator with partial and weighted credit might be the better choices.