Introduction
In the field of bibliometrics, researchers often need to allocate scientific credit to specific countries or institutions to rank productivity and impact. For papers with multiple authors from different locations, a common approach is "straight counting," where credit is assigned to either the first affiliation listed or the corresponding (reprint) affiliation. This study investigates whether these two methods are interchangeable by analyzing a massive dataset of over 18 million papers published between 2000 and 2015.
Methodology
The study utilized the Web of Science (WoS) Core Collection, focusing on the 16 most productive countries. To measure consistency, the authors calculated the percentage of match ($P_{ctry}$) between the first and corresponding affiliations. They employed Levenshtein (edit) distance with a 90% similarity threshold to handle variations in institutional naming (e.g., "Tsinghua University" vs. "Tsinghua Univ"), ensuring that minor typographical differences did not skew the results.
National vs. Institutional Consistency
The analysis reveals that the robustness of bibliometric rankings depends heavily on the level of aggregation. At the national level, the choice between first and corresponding author is largely academic. The average match rate was 98.57%, suggesting that national productivity rankings are stable regardless of the counting method used. China showed the highest national consistency (99.58%), while Canada had the lowest (97.33%).
In contrast, institutional rankings are far more sensitive to the choice of metric. The average match rate dropped to 91.43% at the institutional level. This ~9% mismatch implies that nearly one in ten papers could be attributed to a different university depending on whether the first or corresponding author is prioritized. In some countries, such as Brazil, this mismatch was as high as 14%.
The 2013 Data Anomaly
A significant finding of the longitudinal study was a sharp decline in affiliation consistency starting in 2013. This drop—approximately 3 percentage points at the country level and 7 percentage points at the institutional level—was identified as a "data glitch" rather than a shift in scientific collaboration. Investigation revealed that Web of Science changed its recording practices in 2013 to more granularly link authors to affiliations. Paradoxically, this increased accuracy made the "match" percentage appear lower because the data became more precise in distinguishing between different affiliations that were previously grouped together.
Conclusion and Future Work
While national-level bibliometric assessments are robust, institutional evaluations require careful consideration of attribution methods. The study also highlights the vulnerability of bibliometric trends to changes in database indexing rules. Future work should involve cross-database comparisons with Scopus or Dimensions to determine if the 2013 shift is unique to Web of Science and to develop more sophisticated entity disambiguation algorithms.
Summary updated