Using citation-based indicators to compare bilateral research collaborations

A standard approach to compare research collaborations between pairs of countries is to look at the citations accrued by all publications with authors from both countries. This approach is often misleading, as aspects only marginally related to the collaboration between the country pairs may bias the result considerably. Among them, the main aspect is the number of co-authors. Publications with many co-authors have on average higher citation impact. If the mix of co-publications between two countries has a high share of such publications, the citation impact will likely be high. Moreover, publications with many co-authors tend to include many countries and are thus only to a limited extent characterising the actual collaboration between the selected pair of countries. The purpose of this study is to develop methods for comparisons of country pairs useful for policy makers, who use SciVal or similar tools. Five methods to compare international collaboration are developed and tested. It is noted that the standard approach for comparisons deviates the most. Fractional methods to calculate the citation impact are recommended, as they allow for the use of citations to all co-publications with a higher weight on the citations to publications in which the country pair dominates. As fractionalisation is laborious to carry out based on SciVal data, a more convenient option is also suggested, which is to use co-publications with maximum 10 co-authors. Elsevier should introduce better methods for comparisons of international collaborations and, until this has been made, help its users understand the limitations of the standard approach featured in SciVal. A by-product of the study is that international co-publications deliver a higher citation impact also when publications with the same number of co-authors are compared.