Document Type Assignment Accuracy in Citation Index Data Sources

Introduction The observed citation counts of publications can be divided by the average of a reference set of similar publications in order to get a relative impact measure. It is customary to define the reference set by publication date, scientific discipline and document type. Different document types (DT) have very different citation distributions, leading to very different results in calculations of indicators when separating reference sets by DT and disregarding this kind of normalization (Sirtes, 2012). Thus, when computing relative impact, the correctness of the assignment of document types to publications is crucial. The correctness of DT assignment in citation indexes has been called into question by studies of van Leeuwen et al. (2007), drawing attention to the treatment of letters and ‘research letters’ from medical journals as the same type in Web of Science and by Harzing (2003), illustrating how WoS is using some highly questionable assignment criteria. In this contribution DT assignments in WoS (Thomson Reuters, 2013) and Scopus (Elsevier, 2014) by their respective staff are compared to those of the publishers.