The role of baseline granularity for benchmarking citation impact. The case of CSS profiles

• Published in: Scientometrics Vol. 116; no. 1; pp. 521 - 536
• Main Authors: Glänzel, Wolfgang, Thijs, Bart
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.07.2018; Springer Nature B.V
• Subjects: Agglomeration; Citation analysis; Computer Science; Information Storage and Retrieval; Library Science; Quantiles; Characteristic scores and scales; Journal citation measures; Granularity; Citation impact
• ISSN: 0138-9130; 1588-2861
• DOI: 10.1007/s11192-018-2747-1

In this paper we study the effect of granularity on Characteristic Scores and Scales (CSS). Unlike the traditional indicators that are mostly based on means and quantiles, CSS require the reduction of the citation distributions collaboration of the underlying reference population to four states (classes) and thus higher a different level of granularity. While the question of the choice of granularity is at higher levels of aggregation usually not critical since countries and university have rather multidisciplinary profiles, at lower aggregation levels specialisation becomes more typical. Inappropriate granularity might not warrant the depiction of the publication profiles at these levels in a correct and adequate manner and thus not add accurate citation profiles either. In order to be able to process one complete annual volume of the Web of Science, we decided to calculate CSS thresholds and classes for two levels of granularity, namely sub-fields and WoS Subject Categories. With about 5% deviation, we did not find a real significance. However, we identified journals with similar impact measures but different citation profiles, independently of the granularity. Finally, we have pointed to the limitations in the choice of granularity—in terms of both too broad and too narrow subjects.