Scope:
This document is a guide for institutions on how to interpret their research counts as used by QS. It explains the terminology we use, the sources of information, the time period and the exclusions. In particular, it may serve as a useful companion to both our Fact Files, and our Underlying Data Sheets.
All of our research indicators use data from Elsevier Scopus. We receive an extract from Elsevier at the beginning of each year. This extract is used for the QS World University Rankings (WUR) and every other ranking in the cycle following the WUR.
The following terms and definitions should be noted:
- Paper. A document published in an academic source.
- Paper type. Elsevier classify academic output according to certain types, such as article / book chapter / review.
For our analyses, QS uses the following paper types:
- Article
- Review
- Conference Paper
- Book
- Book Chapter
- Article in Press
- Business Article
- Citation. An academic citation of a paper in another paper, as published in an academic source.
- Self-citation. A citation of an author’s own work by said author.
- Affiliation. One or more institutions / units / departments affiliated with a paper. A university may have several affiliations associated with them. Any affiliation may have its own ID in Scopus (Scopus Affiliation ID) which is attributed to a university by Elsevier and reviewed / amended by QS (e.g. overseas campuses are not currently included in QS counts).
- Scopus ASJC code (All Science Journal Classification code). A code used to identify the academic field of the source (predominantly serial title) indexed by Scopus. More than one code may be used per source. A list of codes can be found here.
- Affiliation cap. A variable cap per subject area applied to papers with high numbers of affiliations in the corresponding subject area (list of subject areas defined by QS). This is an internal step. See here for a technical walkthrough.
- Normalization of citations / papers. A process by which QS seek to normalize the imbalance in research output across disciplines to make it more equitable. This is an internal step. Details can be found here.
Prior to the final analysis, we will send institutions with an underlying data sheet check (Fig. 1). Part of this will be to check institution data, such as student and faculty numbers. The other part will state the numbers of papers and citations that will be used in the analysis.
Figure 1. Papers and Citations as shown in a QS Underlying Data Sheet.
This is how those figures are arrived at:
- Papers (gross)– the total number of papers, over a 5 year window, prior to any filter applied. You can approximate our counts using either SciVal (Fig. 2 in Appendix) or Scopus (Fig. 3 in Appendix).
- Papers (net)– the total number of papers, over a 5 year window, that have been filtered by paper-type exclusion and had a variable affiliation cap applied. Note also that overseas campuses are not currently included in our counts. Only this paper number is used in our normalized total paper count, while papers gross is there just to highlight the proportion of papers lost by our filters.
- Citations(gross, including self-citation) – the total number of citations including author-level self-citations and belonging to papers gross (see 1), prior to any filter applied.
- Citations(net, including self-citations) – the total number of citations including author-level self-citations and belonging to papers net (see 2).
- Citations (net, excluding self-citations) – the total number of citations excluding author-level self-citations and belonging to papers net (see 2). Only this citation number is used in our normalized total citations count, while 3 and 4 are there just to highlight the proportion of citations lost by our filters.
Understanding Differences in Paper and Citation Counts
When asked to compare our numbers (Fig. 1) with yours, you might find that the total papers (gross) and citations (gross) appear different from what you see on Scopus (Fig. 2). A variation of up to 5% in gross papers is to be expected, but deviations beyond 5% may also occur. Please, note that QS only considers the following paper types: articles, reviews, conference papers, books, book chapters, articles in press, business articles, data papers, reports, and retracted articles.
Citations (Gross), Papers (Net) and Citations (Net) are calculated based on the Papers (Gross) after removing ineligible paper types. Additionally, we remove affiliations located in a different country than the parent institution (according to the Scopus database).
The Papers (Net) and Citations (Net) are calculated after removing papers exceeding the affiliation cap. Although the affiliation cap only removes 0.1% of all papers published in each subject, the impact on citations can be sizeable, as these papers tend to be highly cited and involve collaborations with many institutions.
Differences in the SciVal vs QS numbers:
The first point to note is that SciVal is a live database and so should be expected to have a larger number of papers and citations overall. Our Scopus extract is static and provided to us early in the year (end of January for 2024). The second point to note is that SciVal does not have a feature for applying a year window to citations (but only to papers) and excluding author-level self-citations, while Scopus does.
Example:
QS |
SciVal |
Scopus |
|
Paper (Gross) |
547 |
550 |
551 |
Papers (Gross, after applying a filter by paper type) |
n/a* |
528 |
529 |
Papers (Net) |
517 |
n/a** |
n/a** |
Citations (Gross, Including Self-Citations) |
3210 |
n/a*** |
3182 |
Citations (Gross, Excluding Self-Citations) |
n/a* |
n/a *** |
2313 |
* normally, we do not provide this intermediary number in our documents, but you can estimate the impact of our paper-type filter vis Scopus / SciVal
** SciVal / Scopus does not have a filter by affiliation caps, but you can estimate the numbers through some data analysis using Scopus paper export by ASJC code or SciVal paper export by QS Subjects and using our cap values ( see article attachement)
*** SciVal does not have a feature of limiting a citation window by year and excluding author-level self-citations
Appendix
Figure 2. Documents as shown in SciVal, filtered by year.
Figure 3. Documents and citations as shown in Scopus, filtered by year only