Science Journal Ranking by Average Impact Factors

    Last Updated in December 2002

 by Ioan-Iovitz Popescu
 
 


Links dedicated to my dearly beloved Denisa (Daisy): 
Denisa's discovery of multiphoton spectra (1.16 MB) 1973
Damned impact factors April 2004 
Dr. Denisa Popescu - In Memoriam by Carl B. Collins 27 November 2003
Nominated for the Nobel Prize in Physics (1981, 1995, 1996, 1997) 
For her discovery of atomic multiphoton spectra (1973)
Pictures of Dr. Denisa Popescu at her apogee 
   
Our wedding (August 10, 1963)
Her look while discovering multiphoton spectroscopy (1973)
With Professor Carl B. Collins, Center for Quantum Electronics, UT Dallas (1973)
Working out spectral records (1973)
Denisa and her Dallas team (1973)
Heralding Laser Spectroscopy - Vita et Opera of Dr. Denisa Popescu (2003)
Beloved, sleep: �Weep not, she is not dead but sleepeth� Luke 8:52 November 2004
Denisa & Iovitzu - Photo Album April 2006

Added in June 2003:
On a Zipf�s Law Extension to Impact Factors (218 KB)

Added in March 2003:
Science Journal Titles (125 KB) and their Vocabulary (85 KB)

Added in December 2002:
SUMMARY_Version_2003 (39 KB)
Science_Journal_Ranking_Version_2003_for_8011_journals(308 KB)
SUMMARY_Versions_2001_2002_2003 (36 KB)
Science_Journal_Ranking_in_2001_2002_2003 (214 KB)

SUMMARY_Version_2001 (42 KB)
SUMMARY_Version_2002 (42 KB)
SUMMARY_Version_2001_versus_2002 (35 KB)
Science_Journal_Ranking_Version_2001 for 7557 journals (285 KB)
Science_Journal_Ranking_Version_2002 for 7832 journals (305 KB)
Science_Journal_Ranking_Version_2001_versus_2002 (178 KB)
Science_Journal_Titles_Word_Frequency_Version_2002 (65 KB)
ISI Journal Title Abbreviations  (104 KB)

 
 
Summary figures of the present article
Illustrating a typical non power law by a downwards curved �log (rank) � log (frequency)� distribution of science journal ranking by average journal impact factors, JIF (left figure), and the corresponding distribution of journal ranks, JRK (right figure), as separately assigned within 107 disciplines. Incidentally, the journal rank distribution coincides with that of a uniform sequence of random numbers. Similar relationships hold also true for various subsets such as scientific fields and disciplines (see figures in ADDENDUM at the end of the article). 

 

Annual impact factor dynamics in science (1974 - 2004)
see the Lavalette fitting in the Addendum below
as well as at Popescu (2003) and Mansilla et al. (2007)
 

 
 
 Science Journal Ranking by Average Impact Factors

Version December 2001

Created by Acad. Prof. Dr. Ioan-Iovitz Popescu

Based on ISI annual data sets of SCI-JCR (1974-1999)

 
Motto: �There is something still worse, however, than being either criticized or dismantled by careless readers: it is being ignored. Since the status of a claim depends on later users' insertions, what if there are no later users whatsoever? This is the point that people who never come close to the fabrication of science have the greatest difficulty in grasping. They imagine that all scientific articles are equal and arrayed in lines like soldiers, to be carefully inspected one by one. However, most papers are never read at all. No matter what a paper did to the former literature, if no one else does anything with it,  then it is as if it never existed at all. You may have written a paper that settles a fierce controversy once and for all, but if readers ignore it, it cannot be turned into a fact; it simply cannot. �You may protest against this injustice; you may treasure the certitude of being right in your inner heart; but it will never go further than your inner heart; you will never go further in certitude without the help of others.� (Bruno Latour, Science in Action, p. 40 (1987))
 
Humbly dedicated to cited scientists
FOREWORD
    Bibliometric indicators currently used to examine and evaluate the published knowledge production are primarily based on impact factors of journals covered by Science Citation Index database and published annually since 1975 in the Journal Citation Reports [Garfield (Editor)]. This concept has been introduced by Garfield [Garfield, 1972, 1979] as a measure of the average citation frequency for a specific citable item (article, review, letter, discovery account, note, and abstract) in a specific journal during a specific year or period. Commonly, the impact factor of a journal is defined as the ratio between citations and recent (previous two years) citable items published or, in other words, as the average number of citations in a given year of articles published in that journal in the preceding two years. Thus, for instance, the impact factor for 1990 of Physical Review Letters (PRL) has been calculated as the cumulated number of 22,007 citations received in 1990 for articles published in the considered journal in 1988 (11,497 citations) and 1989 (10,510 citations), divided by the cumulated number of 2901 citable articles published in that journal during the same two-year period, i.e. in 1988 (1430 articles) and 1989 (1471 articles). The impact factor of PRL in the year 1990 results accordingly from the ratio 22,007 citations / 2901 papers = 7.586 citations per paper and has the meaning of number of citations received by the "average PRL article" during the considered two-year period. Obviously, the definition can be extended over longer time spans. For more detailed recent information about impact factors and their applications and extensions, see also [Garfield, ISI Essays, 1994; PRESTìGIXTM, 2000, 2001]. Thus, for instance, while the impact factor is based on 4 independent variables (citations in the considered year to articles published in the preceding two years, and their respective number of articles), the prestige factor uses a more comprehensive algorithm (PRESTìGIXTM) containing 6 independent variables (citations in the considered year to articles published in that year and in the preceding two years, and their respective number of articles). In other words, the prestige factor uses more recent data, namely from 1998-2000 for the year 2000, and from 1999-2001 for the year 2001.
   Developed originally from the need to compare the journal influence or performance, the impact factor provides nowadays the main quantitative tool for ranking, evaluating, categorizing, and comparing journals. Thus, it provides librarians a tool for the management of journal collections and publishers a quantitative evidence in evaluating the position of their journals. But data can as well be ranked to reveal interesting facts about individual or collective performance and trends, such as highly cited papers and authors (hot papers, hot authors), most active laboratories, institutions or research fronts, up to countries and world science mapping and policy [Aguillo, Braun, Garfield, Katz]. Accordingly, many institutions worldwide are devoted to information science and technology based on citation analysis and variously known as scientometrics, bibliometrics, informetrics, cybermetrics, and webometrics � visit more sites at USEFUL LINKS in the References. In Romania a specialized department has recently been created under the Romanian Ministry of Education and Research, namely CENAPOSS, an acronym for the National Center for Science Policy and Scientometry, set up at the end of 1999.

    "Perhaps the most important and recent use of impact is in the process of academic evaluation. The impact factor can be used to provide gross approximation of the prestige of journals in which individuals have been published. This is best done in conjunction with other considerations such as peer review... Again, the impact factor should be used with informed peer review" [Garfield, 1994]. Methods and techniques are currently designed for evaluation and comparison of research groups and individual scientists, such as the so called ISI�s Expected Citation Rates (ECR) System [Garfield, 1994] and ISSRU�s precursory Mean Expected Citation Rates (MECR), Mean Observed Citation Rates (MOCR) and Relative Citation Rates (RCR) [Braun, 1985-1988].

     A simple scientometric evaluation of individual and group contributions in fundamental science has recently been proposed and a particularly relevant scientometric indicator has been introduced, namely the cumulative impact factor [Popescu, 1994], defined by the sum :

S [(journal impact factor, q) / (article author number, a)]
or, shortly, S (q/a), extended over the whole list of scientific publications of the assessed individual or group. Obviously, the cumulative impact factor has the meaning of author�s total number of citations per author in the first two years after publication, with its unit cites/author at this paper age. This unit is equivalent to a single-authored (a = 1) article, published in a journal with impact factor unity (q = 1).The cumulative scientometric indicator defined above has successfully been tested for promotion thresholds in Romanian physics research institutes and faculties [Romanian Ministry of National Education, Order No. 5103, Appendage 1-II, dated on 05.07.1999] and for accreditation of research excellence centers in Mathematics, Physics, and Chemistry [Romanian CNCSIS - National Council of Scientific Research in Higher Education]. The necessary thresholds resulted from the "Romanian experiment" for any candidate to a high academic position or grade are, for instance, a minimal score of 6 to 8 points for associate university professor, of 9 to 12 points for full university professor, and of 14 points for Ph.D. supervisor. In other words, these minimal promotion scores require the equivalent of 6-8, 9-12, and 14 single-authored papers respectively, published in unity-valued impact factor journals. The conversion of these figures from the field of physics to any other scientific field can easily be carried out by multiplying with the ratio of the corresponding average impact factors (AVEJIF) given in the linked SUMMARY Version 2001(or in SUMMARY_Version_2002, inasmuch as the AVEGIF-values remain essentially the same). Thus, for instance, for ENGINEERING SCIENCES, the minimal promotion scores given above for PHYSICAL SCIENCES should be reduced by a factor of
 
(AVEJIF)ENG / (AVEJIF)PHYS = 0.42/1.41 »0.3.
    Clearly, a high number of citations mean a major impact in the specific field or a high utility. However, as pointed out above, it is critical to take into account, among other aspects, that publication and citation rates, as well as the peak impacts, vary widely from field to field, and among different disciplines, and we need to know what the average citation rate is within a field and a discipline to assess an individual. To illustrate this diversity, the current average values of impact factors for 12 scientific fields and 107 scientific disciplines are given in the SUMMARY_Version_2001 or, respectively, in the updated version SUMMARY_Version_2002 . A convenient alternative way to consider this requirement consists in the use of the relative rank, r, of the journal within its discipline instead of the impact factor, q. The grounds consist in the fact that, according to the Lavalette ranking law [Lavalette, 1996; Popescu, 1997; see also  ADDENDUM ], there exists a simple functional dependence between r and q, namely
q = c [Nn/(N-n+1)]-b = c [(N+1)/r - N]-b
where the positive exponent b (roughly 1/2) and the scaling factor c are two fitting parameters, N is the total number of journals in the considered discipline, and
r = 1 - (n - 1)/N
defines the journal (relative) rank, corresponding to its (descending absolute) ranking number n. Thus, for instance, a value r = 0.75 means that 75 % of journals of the considered discipline have a rank (and the corresponding impact factor) lower or equal to that of the considered journal. Incidentally, the journal rank distribution coincides with that of a uniform sequence of random numbers TOP (right). For many reasons, mainly in interdisciplinary comparisons and evaluations, it appears therefore more appropriate to use a "normalized" indicator such as the cumulative rank, defined by:
S [(journal rank, r) / (article author number, a)]
or simply, S (r/a), with its natural unit rank. This unit is equivalent to a single-authored (a = 1) article, published in the discipline top journal (r = 1). The major advantages of the "ranks scale" S (r/a) in comparison with the "cites scale" S (q/a) consist in both (i) a bibliometric equivalence of journals belonging to various disciplines but having the same rank, and (ii) a much higher stability as compared to the corresponding impact factor. Generally, the journal rank r = 1 - (n - 1)/N ranges from unity (for top journals) to 1/N (for bottom journals). It follows, thus, that the journal average rank of scientific disciplines (or fields) is given by (1+1/N)/2 » 1/2, as far as N >> 1. This narrow distribution of average rank values, in contrast to that of average impact factors, indicates the "normalization" of the rank scale notwithstanding the variety of disciplines and fields displayed in the attached SUMMARY_Version_2001 or SUMMARY_Version_2002 tables, and illustrated in the TOP (right)  figure by the normalized linear distribution of journal ranks. Consequently, the rank scores and the promotion, appointment, and accreditation thresholds, established in the cites scale, S (q/a), can be roughly converted into the rank scale, S (r/a), by simply multiplying by the ratio 0.5/AVEJIF.
    Obviously, for the use of the rank scale it is of major importance the proper placement of journals into various possible disciplines. In the present database version, as a general rule, each journal has been "optimally assigned" to one (and only one) discipline, namely to the 1999-ISI discipline in which that journal has the optimal position, i.e. the highest rank number. Thus, for instance, the journal Int. J. Mod. Phys. C has the rank r = 0.80 in the discipline Computer Sciences (Interdisciplinary) and the rank r = 0.40 in the discipline Mathematical Physics. In a similar way, the journal J. Magn. Magn. Mater. has the rank r = 0.82 in the discipline Material Sciences and the rank r = 0.57 in the discipline Condensed Matter Physics. In the previous database version (July 2000), both these journals were located in the field PHYSICS, namely in the disciplines Mathematical Physics and Condensed Matter Physics, respectively. In the present version, however, by virtue of the aforementioned principle of maximal positioning (ranking) within possible disciplines, the two journals of the example given above have been assigned to the discipline Computer Applications of the field COMPUTER SCIENCES, respectively to the discipline Materials of the field MATERIALS SCIENCES. Thus, it happens that some interdisciplinary journals rather get out of their "traditional" or "intuitive" discipline and move into more related ones. The biggest number of disciplines into which ISI arranged journals in 1999 is 6 (six) as, for instance, the journal Chem-Biol. Interact., which appears in Chemistry (r = 0.83), Toxicology (r = 0.82), Chemical Medicine (r = 0.77), Biology (r = 0.74), Pharmacology and Pharmacy (r = 0.72), and Biochemistry and Molecular Biology (r = 0.54). According to the rule adopted here, the journal in this later case has been allotted to the Chemistry discipline. Perhaps the most useful table needed for further improvement of discipline definition and journal allocation consists in the word frequency listing of scientific journal titles. The one corresponding to the latest records, containing the vocabulary and the occurrence frequency of (mostly abbreviated) 4196 distinct words used in the titles of 7832 science journals, can be found in the TXT file: Science_Journal_Titles_Word_Frequency_Version_2002 (65KB).
    In order to meet the increasing needs of journal impact factors for a variety of purposes, this work stands for a completed and updated continuation of the previous versions (Popescu, version December 1999, with 2,935 journals; version July 2000, with 5,762 journals). It represents at this time (of version 2001) a collection of 7,557 journals and their 93,618 annual impact factors from SCI Journal Citation Reports of general and special interest to scientists engaged in fundamental, life and engineering sciences, and covering all available journal impact factors in the window 1974-1999 (excepting the not edited 1976 year). Thus, the "ever-changing river of journals", has an average lifetime of [93,618 (years) / 7,557 (journals)] » 12.4 years/journal, i.e. of about one half of the considered effective 25 years of ISI quotation. The linked SUMMARY_Version_2001 (and, similarly, SUMMARY_Version_2002) presents the data of the corresponding 12 scientific fields and 107 scientific disciplines, with their brief definition, the journal number N, the discipline (field) weighted average journal impact factor (AVEJIF), and the discipline or field top journal title and its impact factor in 1999 (or in 2000, respectively). As to the news in the version 2001,  compared with previous versions, these are mainly the domains of MEDICAL SCIENCES (1906 journals) and AGROSCIENCES (487 journals), so that the present database displays the main stream of all scientific journals.  For convenience, the actual database will be divided into separate Excel files and links to each of the 12 scientific fields, as given in the following list. Also the currently updated database version 2002 is added below, confirming, as expected, the stabilization of the average impact factors:
 
NEW! SUMMARY_Version_2003 (39 KB)
NEW! SUMMARY_Versions_2001_2002_2003 (36 KB)
NEW! Science_Journal_Ranking_Version_2003_for_8011_journals (308 KB)
SUMMARY_Version_2002 (42 KB)
SUMMARY_Version_2001_versus_2002 (35 KB)
Science_Journal_Ranking_Version_2002 for 7832 journals(305 KB)
SUMMARY_Version_2001 (42 KB)
Science_Journal_Ranking_Version_2001 for 7557 journals (285 KB)
 
 
AGROSCIENCES � 487 journals (94 KB)
AMBIENTAL SCIENCES � 183 journals (45 KB)
BIOSCIENCES � 1456 journals (248 KB)
CHEMICAL SCIENCES � 422 journals (84 KB)
COMPUTER SCIENCES � 365 JOURNALS (75 KB)
ENGINEERING SCIENCES � 1067 journals(188 KB)
GEOSCIENCES � 253 journals (57 KB)
MATERIAL SCIENCES � 294 journals (63 KB)
MATHEMATICAL SCIENCES � 378 journals (77 KB)
MEDICAL SCIENCES � 1906 journals (322 KB)
PHYSICAL SCIENCES � 575 journals (115 KB)
SCIENCE AND EDUCATION � 171 journals (44 KB)
      Each sheet is self-explanatory, has a heading, a legend, and the same Excel format, with the following columns, namely: identification number (ID), DISCIPLINE, journal rank (JRK), abbreviated JOURNAL TITLE, average journal impact factor (JIF) for the entire time span of ISI quotation, the corresponding number of years of ISI quotation (YRS), and the impact factor standard deviation (DEV). The rank JRK of  journals has been established FOR EACH DISCIPLINE SEPARATELY in terms of the ranking by average journal impact factor JIF. These particularly important columns JRK and JIF (yellow filled) stand by on the left and right side, respectively, of the JOURNAL TITLE column. Journal ranking by average scientometric indicators, such as by the average impact factors in the present application, could serve for simple uses of extant data in any assessment of research performance for hiring, promotion, tenure, appointment, accreditation, and other academic rewards. The TOP FIGURE  summarizes both scientometric evaluation alternatives, i.e. by journal absolute average impact factor and/or by journal rank within the corresponding discipline. For link to the corresponding scientometric tables in TXT files (easy to copy and paste) click on the following addresses:

Science_Journal_Ranking_Version_2001 for 7557 journals (285 KB)
Science_Journal_Ranking_Version_2002 for 7832 journals (305 KB)

As expected, the annual data �shift� of the considered consecutive years, as measured by the average standard deviation, ranges within percents, namely of 5.1% for JRK and of 2.6% for JIF. Still lower becomes this deviation at larger journal set scales (disciplines < fields < science), such as that of weighted average journal impact factors (AVEJIF) of disciplines (1.15%) > of fields (1.10%) > of the complete science journal listing (0.86%), click on SUMMARY_Version_2001_versus_2002 (35 KB).


Generally, the (counted or estimated) citations in the scientific literature appear to be the most objective measure of any hierarchy by scientific output. Though far from perfect (see, for instance, a recent "anti-scientometric" essay [Kutzelnigg, 1998], criticism [Amin and Mabe, 2000; Adam, 2002], or the sophisticated scientometric algorithm PRESTìGIXTM), the simple �poor man�s citation indicators� proposed above enable to examine several key facets of individual and collective knowledge production. Almost the same results one gets appealing to rather expensive and time consuming direct citation counts. Actually, a �high resolution scientometry� becomes nonsensical inasmuch as needs merely consist in gross definitions of thresholds, classes, and trends.  Wondering to which scientometric classes might really belong angry anti-scientometric people, it would be, however, gratifying if some day scientists will bear with pride objective cumulative ranks above traditional scientific ranks and distinctions.
 
 

Acknowledgements. The author gratefully appreciates the interest in this work of  Professor Tibor Braun, Professor Daniel Lavalette, and Professor Mircea Oncescu. Special thanks are due to Dr. Victor Sofonea for documentation, Mr. Claudiu Cezar Vasilescu, Dr. Savu-Sorin Ciobanu and Dr. Magdalena Nistor  for computer assistance, and Dr. Andrei Barborica, Dr. Eugen Aldea and Mr. Sorin Vizireanu for database programs and homepages. The Alexander von Humboldt-Foundation is gratefully acknowledged for generous donations of computer facilities.
 
 

Prof. Dr. Ioan-Iovitz Popescu

Full Member of the Romanian Academy
Current Personal Scientometric Scores:
18 Ranks / 45 Cites / 1506 Citations
Bucharest, October 2001
 
 

REFERENCES

FURTHER BIBLIOGRAPHY ON RANKING LAWS


1
ADDENDUM

 THE LAVALETTE RANKING LAW

    The ranking law, established in 1996 by the French biophysicist Daniel Lavalette (http://www.curie.u-psud.fr/U350/), states that the impact factor q of a set of N scientific journals, ordered by the descending ranking number n (a positive integer in the [1, N] range), obeys the general relationship

q (n)= c [Nn/(N-n+1)]-b

with only two fitting parameters, namely the exponent b and the scaling constant c = q(1). In other words, the expression proposed by Lavalette represents a linear function in the double logarithmic log(q), log [n/(N-n+1)] scale. Offering the promise for various applications and theoretical investigations, this is barely more complex than the well known rank-frequency Zipf�s law q = c n-b. It is important to point out that the independent variable in the Zipf�s law is the descending ranking number, n, whereas in the Lavalette�s law this is the ratio n/(N-n+1) between the descending and the ascending ranking numbers, thus explicitly enclosing the set size number N.
 
 
 
 
 
 
 
 

Fig.1.The normalized Lavalette ranking function q/c =  [Nn/(N-n+1)]-b  in terms of the descending ranking number n for typical values of the parameters N and b

.

Fig.2. Illustrating the Lavalette ranking law for three random subsets (1000, 2000, and 4000 journals) excerpted from the present collection (7557 journals) and ranked by average journal impact factors (JIF). The journals with JIF = 0 have been excluded.

 
 
Fig.3. Illustrating the Lavalette fitting for 4 random subsets of journals with title initial letter A, B, C, or D, as excerpted from the present collection (7557 journals) and ranked by average journal impact factors (JIF). 

 
Fig.4. Ranking of 26 random subsets of journals with title initial letter belonging to various letters of  the alphabet, as excerpted from the present collection (7557 journals) and ranked by average journal impact factors (JIF). The non power law shaping is obvious and the fitting pleasure is left to the reader (power law means straight line on log-log plot). 

 
Fig.5. Illustrating the Lavalette ranking law for the present collection (7557 journals) and two disjoint subsets of the fields of Medicine (1906 journals) and Physics (575 journals) respectively, ranked by average journal impact factors (JIF). The journals with JIF = 0 have been excluded.  

 
 
Fig.6. Ranking of 12 natural subsets of journals belonging to various scientific fields, as excerpted from the present collection (7557 journals) and ranked by average journal impact factors (JIF). The non power law shaping is obvious and the fitting pleasure is left to the reader (power law means straight line on log-log plot).

 
 
Fig.7. Testing of Lavalette�s variable Nn/(N-n+1) for the average impact factors (JIF) of 7557 journals, as assigned within 12 scientific fields, shows a systematic departure from a perfect Lavalette fitting, i.e. from a straight line on a log(JIF), log [Nn/(N-n+1)] plot. 

 
 
Fig.8. Illustrating the King�s effectin the particular case of the Science & Education field, as revealed by the first three positions held by Nature (JIF = 15.68), Science (JIF = 14.31), and P. Natl. Acad. Sci. USA (JIF = 9.53), in contrast to the field average journal impact factor amounting only to AVEJIF =  0.88. The matching to the rest of the sequence can be substantially improved by simply eliminating the anomalous members.

 
 
This is a page of the site:

Prof. dr. Ioan-Iovitz POPESCU
Member of the Romanian Academy
Vita et Opera: Plasma, Lasers, Scientometrics