Comparing standard, collaboration and fractional CNCI at the institutional level: Consequences for performance evaluation

The average Category Normalised Citation Impact (CNCI) of an institution’s publication output is a widely used indicator for research performance benchmarking. However, it combines all entity contributions, obscuring individual inputs and preventing clear insight and sound policy recommendations if it is not correctly understood. Here, variations (Fractional and Collaboration [Collab] CNCI)—which aim to address the obscurity problem—are compared to the Standard CNCI indicator for over 250 institutions, spread globally, covering a ten-year period using Web of Science data. Results demonstrate that both Fractional and Collab CNCI methods produce lower index values than Standard CNCI. Fractional and Collab results are often near-identical despite fundamentally different calculation approaches. Collab-CNCI, however, avoids assigning fractional credit (which is potentially incorrect) and is relatively easy to implement. As single metrics obscure individual inputs, institutional output is also deconstructed into five collaboration groups. These groups track the increasing international collaboration trend, particularly highly multi-lateral studies and the decrease in publications authored by single institutions. The deconstruction also shows that both Standard and Fractional CNCI increase with the level of collaboration. However, Collab-CNCI does not necessarily follow this pattern thus enabling the identification of institutions where, for example, their domestic single articles are their best performing group. Comparing CNCI variants and deconstructing portfolios by collaboration type is, when understood and used correctly, an essential tool for interpreting institutional performance and informing policy making.