“Whether a text is interesting and you get something out of it is more important than whether it is published somewhere important”.
PhD candidate, humanities
Hopefully, your PhD research will make an impact by advancing knowledge in your field or by contributing to real-world applications. While these kinds of impact are difficult to measure validly, more or less useful approximations of the degree of impact originate in data on how often and how broadly research is cited. There are, however, important aspects of research beyond those captured by citation-based metrics, and recent initiatives have spurred a growing interest in a broader and fairer basis for research assessment. On this page you will learn about
Evaluating research and researchers is not easy. While citation-based impact metrics, such as the journal impact factor, are convenient and have been popular, they have serious limitations and drawbacks as research assessment tools.
The Declaration on Research Assessment (DORA) is a global, cross-disciplinary initiative that embodies an awareness of the need to develop better methods for evaluating research and researchers. The number of signatories is growing, and in 2018 it has been signed by the Research Council of Norway, the Norwegian University of Science and Technology and UiT The Arctic University of Norway. Several major research funders are also among its signatories (e.g. the Wellcome Trust).
Signatories of DORA commit to not using journal-based metrics (such as the impact factor) in decisions regarding funding, hiring and promotion, to be explicit about the criteria used for making such decisions, to consider the value and impact of all research outputs (e.g. datasets and methods innovations), and to expand the range of impact measures to include such things as influence on policy and practice.
DORA represents an important development. Arguably, it implies that you may benefit from considering ways in which you can describe and provide documentation of any influence your work may have that are not captured by citation-based impact metrics. That said, citation based metrics continue to be important and to evolve. In the following sections, we will provide an introduction to some of them.
The most well-known measure of journal rank is the journal impact factor (often abbreviated to IF or JIF). It was developed in order to select the journals to be included in the Science Citation Index (Garfield, 2006). The impact factor is a measure of how often articles in a particular journal have been cited on average in a given year. The central idea is that the impact factor and similar measures of journal rank indicate the journal’s relative influence among journals within the same subject category.
A journal’s impact factor is based on citation data from the Web of Science database, owned by Clarivate analytics. If your institution has purchased the appropriate licence, you may be able to look up a journal’s impact factor and related statistics there. While using the impact factor in research evaluation is controversial (see Critical remarks, below), as a PhD candidate, you should know what it is, and how it is calculated.
The impact factor (IF) is the ratio of the number of citations (A) in the current year to items published in the previous two years and the number of citable articles (B) published in the same two years: IF=A/B.
Figure 1: General formula for calculating a journal’s impact factor.
Consider the journal Proceedings of the National Academy of Sciences (PNAS). This journal published a total of 6 467 citable articles in 2015-2016. In 2017, the total number of citations to articles from these two previous years was 61 460 (see table below).
|IF for 2017||2015||2016||Sum|
|Citations in 2017
to articles published in
|35 338||26 122||61 460|
|3 282||3 185||6 467|
Table 1: Citations and publications involved in the calculation of the impact factor for PNAS for 2017
Over the years, criticism has been raised against the impact factor. You can read more about some of the critical remarks here.
The impact factor is associated with a journal’s prestige, and is sometimes considered a proxy for the scientific quality of the work it publishes. Unfortunately, there is no verifiable association between journal impact factor and reasonable indicators of quality (for an overview of the relevant research, see Brembs, 2018).
As can be seen from the formula and example calculation above, the journal impact factor is, roughly, a mean. Means are sensible indicators of central tendency if the distribution of values is symmetrical. However, citations to scholarly articles are not symmetrically distributed. Most published articles receive few, or even no, citations, while a small number of articles become very highly cited. This skewness means that a journal’s impact factor is a poor predictor of the citation count of any given single article published in that journal (Seglen, 1997; Zhang, Rousseau & Sivertsen, 2017).
Because the impact factor is field dependent only journals within the same scientific field are comparable. Nevertheless, the impact factor has been used to compare different fields.
Figure 2: Amount of references by age to articles published in 2011 (the figure is based on Adler, Ewing, & Taylor, 2009)
This figure demonstrates the field dependency of the impact factor. The citations contributing to the calculation of the impact factor are the ones inside the two-year citation window, marked grey. Citations outside the window do not count, even though most of them lie outside and refer to older articles. For rapidly developing fields (blue line), the impact factor is considerably higher than in slowly developing fields (red line). A lower impact factor does not mean a lower quality of one field compared to another. The citation window may be too short to be representative for subjects which develop slowly.
The pool of selected journals has a strong Anglo-American bias. Influential journals written in other languages are rarely captured by Clarivate Analytics Journal Citation Reports.
“Typically, when the author’s bibliography is examined, a journal’s impact factor is substituted for the actual citation count. Thus, use of the impact factor to weight the influence of a paper amounts to a prediction, albeit coloured by probabilities.” (Garfield, 1999)
The impact factor is not only used for ranking journals according to their relative influence, as initially intended, but also for measuring the performance of individual researchers. Given the skewness of citation distributions described above, this is a misapplication. The use of the impact factor when applied to individual researchers has been criticized by a broad scholarly community, not least the co-creator of the Science Citation Index, Eugene Garfield, himself.
The impact factor can be manipulated. It is influenced by the point in time when a journal issue is published. Issues published at the beginning of a year have a greater chance to accumulate citations than those published at the end of the year. Furthermore, editors may influence the value of their journal’s impact factor by writing editorials containing references to articles in their journal (journal self-citations). In addition, references given in the editorial count to the numerator, while editorials do not count to the denominator. By definition the denominator only consists of citable articles and editorials are not regarded as such.
References given in the articles may be incomplete and incorrect. Incorrect references are not corrected automatically and therefore are not added to the citations. This fact influences the value of the impact factor and other citation indicators such as the h-index.
In order to compensate for some of the weaknesses of the impact factor (field dependency, inclusion of self-citations, length of citation window, quality of citations), efforts have been undertaken to develop better journal indicators. More advanced metrics are usually based on network analysis, such as the SCImago Journal Rank (SJR) and the Source Normalized Impact per Paper (SNIP), both based on data from Scopus. While such measures arguably do a better job of ranking journals, they are still only applicable to journals and should not be used to evaluate research output at the level of individual researchers. For that purpose, the h-index, introduced below, is better suited.
The h-index is a measure of the total, citation-based impact of a researcher measured by how often she/he is cited.
When exploring the literature of your research field, the h-index may give you a picture of the impact of individual researchers and research groups. You may retrieve the h-index from e.g. Web of Science, Scopus and Google Scholar.
When applying for a scholarship, project funding or a job, you may be required to state your h-index.
The h-index for a given author (Karen) calculated step-by-step:
Step 1: Search for author Karen in a given database
Figure 3: Search result for author Karen’s publications in a given database
The figure illustrates the search result for author Karen in an arbitrary database. It also indicates all citing publications (counting lines) within the same database. In our example, author Karen has 10 publications (a, b, c, d, e, f, g, h, i and j). NP = 10.
Step 2: Sort publications by decreasing number of citations
Table 2: Karen’s publications sorted by decreasing number of citations. View table as graph
Step 3: Fulfil the condition: h of the publications have at least h citations each
h = max(Rank) provided Cit ≥ Rank
In our example, four publications have received more than four citations. These are publications c, i, a and g. The remaining (NP – h) publications do not have more than h citations each. In our example, the remaining six publications (f, h, d, j, b and e) do not have more than four citations each.
Result: Karen’s h-index is equal to 4.
In this example, we use a renowned Norwegian researcher in ecology and evolutionary biology: Nils C. Stenseth. We demonstrate that his h-index is different in each of the databases due to their different coverage of content.
Step 1: Search for the author, making sure you cover all possible versions of the author’s name
Step 2: Go to the statistics available in the three databases
The results presented here are based on data as of April 2018. Citation counts typically increase with time and so does the h-index. To determine the present value, perform a new search.
h-index = 76
To make sure that all publications by the author are retrieved, the example shows search results for stenseth n*, stenseth nc*, and stenseth nils. It is possible to add rows to make sure that different spellings of the name are included. The time span is 1945-2018, and the number of publications is 630. The h-index, including self-citations, is 76.
h-index = 75 (80)
The time interval covers publications from 1974 to 2018. In Scopus, the h-index excluding self citations is 75. With self-citations, the h-index would be 80. The number of publications is 595.
h-index = 100
Google Scholar covers a wider range of publication types; therefore the h-index is higher here. The number of publications is 756. The h-index for all years is 100, while the h-index since 2013 is 57.
The h-index alone does not give a complete picture of the performance of an individual researcher or research group.
The h-index underrepresents the impact of the most cited publications and does not consider the long tail of rarely cited publications. In particular, the h-index cannot exceed the total number of publications of a researcher. The impact of researchers with a short scientific career may be underestimated and their potential undiscovered. Read more about this below: “Problem: The Matthew effect in science”.
“Using a three-year citation window we find that 36% of all citations represent author self-citations. However, this percentage decreases when citations are traced for longer periods. We find the highest share of self-citations among the least cited papers.” (Aksnes, 2003)
Citing is an activity maintaining intellectual traditions in scientific communication. Usually, citations and references provide peer recognition; when you use others’ work by citing that work, you give credit to its creator. Citations are used for reasons of dialogue and express participation in an academic debate. They are aids to persuasion; assumed authoritative documents are selected to underpin further research. However, citations may be motivated by other reasons as well.
Citations may also express
Applicable across fields?
Note that scholarly communication varies from field to field. Comparisons across different fields are therefore problematic.
However, there are attempts to make citation indicators field independent. For example, The Times Higher Education World University Rankings involve citation indicators which are field independent, i.e. normalized (Times Higher Education, 2013).
Citations are basic units measuring research output. Citations are regarded as an objective (or at least less subjective) measure to determine impact, i.e. influence and importance. They are used in addition to, or as a substitute for peer judgments.
There is a strong correlation between peer judgments and citation frequencies. For this reason, citations are relied on as indicators of quality and are used for e.g.
Citations must be handled carefully when evaluating research.
Citation data vary from database to database, depending on the coverage of content of the database.
Furthermore, two problematic factors are different motivations for citing, and the the considerable skewness of the distribution of citations.
To those who have, shall be given…
When sorting a set of publications by the numbers of citations received, the distribution shows a typical exponential or skewed pattern. Works which have been cited are more visible and are more easily cited again (vertical tail in figure), while other works remain hidden and are hardly ever cited (horizontal tail in figure). This phenomenon is referred to as the Matthew effect in science.
What is the problem with skewed distributions? Skewed patterns make it difficult to determine an average citation count. Different approaches may be applied, see the figure.
Good research + high visibility
Your best chance to make an impact
Being aware of how academic performance is evaluated allows you to make informed decisions and devise strategies to build and document your impact, and thereby improve your career prospects. Our general advice centres on making your work visible, accessible and understandable.
Make your work visible to other researchers:
Make your work accessible to other researchers by adopting open science practices:
Make your work understandable to other researchers:
Adler, R., Ewing, J., & Taylor, P. (2009). Citation statistics. Statistical Science, 24(1), 1-14. Retrieved from https://www.jstor.org/stable/20697661
Aksnes, D. W. (2003). A macro study of self-citation. Scientometrics, 56(2), 235-246. https://doi.org/10.1023/A:102191922
Brembs, B. (2018). Prestigious science journals struggle to reach even average reliability. Frontiers in Human Neuroscience, 12(37), 1-7. https://doi.org/10.3389/fnhum.2018.00037
Garfield, E. (1999). Journal impact factor: A brief review: Canadian Medical Association Journal, 161, 979-980. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1230709/
Garfield, E. (2006). The history and meaning of the journal impact factor. JAMA, 295, 90-93. https://doi.org/10.1001/jama.295.1.90
Hirsch, J. E. (2005). An index to quantify an individual’s scientific research output. Proceedings of the National Academy of Sciences, 102(46), 16569-16572. https://doi.org/10.1073/pnas.0507655102
McKiernan, E. C., Bourne, P. E., Brown, C. T., Buck, S., Kenall, A., Lin, J., . . . Yarkoni, T. (2016). How open science helps researchers succeed. eLife, 5, 1-19. https://doi.org/10.7554/eLife.16800
Seglen, P. O. (1997). Why the impact factor of journals should not be used for evaluating research. BMJ, 314(7079), 498-502. https://doi.org/10.1136/bmj.314.7079.497
Zhang, L., Rousseau, R., & Sivertsen, G. (2017). Science deserves to be judged by its contents, not by its wrapping: Revisiting Seglen’s work on journal impact and research evaluation. PLoS ONE, 12(3), e0174205. https://doi.org/10.1371/journal.pone.0174205