Bibliometric funding allocation at KTH

This page describes the calculations and implementation of the funding allocation model based on the journal citation indicator used at KTH.

Introduction

On behalf of the President of KTH, Peter Gudmundson, the Unit for Publication Infrastructure (PI) at the ECE School has put forward a journal indicator for allocation of research funds. The indicator rewards publication in journals, which are highly cited relative to the subject fields the journals belong to. It is combined with a volume measure: the number of faculty full time equivalents, a combination that constitutes the KTH bibliometric indicator for funding allocation (BIFAKTH). Values of BIFAKTH are calculated per department.

The KTH model for funding allocation is intended to give incentive to the researchers of KTH to publish in highly cited journals.

Units involved in the generation of BIFAKTH values and the fund quantity that is allocated

Values of the journal indicator are generated by PI, whereas values of the volume measure are generated by the Human Resources Department. The Finance Office then calculates the values of the combined indicator, i.e. of BIFAKTH.

Year 2015, the fund quantity that was allocated by the model was 21.6 million SEK. Allocation of funds occurs the year after the year of analysis.

Delimitation of unit publications and the treatment of unit changes

There are different ways to define which publications that belong to a unit. One way is to first define which researchers that at present are employed at the unit, and then base the calculations on the publications these researchers have published. According to this definition, also publications that the researchers have published at other institutions than KTH are counted. For such publications to be counted, however, they must be recorded in DiVA, the publication database of KTH.

Another delimitation approach is to include publications in which affiliation to the unit occurs. The bibliometric funding allocation at KTH is based on this second alternative: the point of departure is the publications that are affiliated to the respective department, and where DiVA is the data source for affiliation.

When units have changed name, been moved, been divided or merged, units or parts of units are attributed to existent units. Example 1. The departments A and B have been merged to department C. Each publication with at least one affiliation to A or B are then attributed to C.  If individual researchers have changed department, earlier publications are not attributed to the new department, given that the former department has not been integrated into the latter. Thereby, publications do not follow individual researchers. Example 2. Researcher X belongs to department A, but has earlier belonged to department B, which has not been integrated into A. X has 1 publication, PA, with affiliation to A, 1 publication, PB, with affiliation to B. A is then assigned PA and B PB (we assume that X is the sole author of the two publications).

Detailed overview of BIFAKTH

The volume measure of BIFAKTH is based on the preceding year, in relation to the year of analysis. For instance, if the year of analysis is 2015, the number of faculty full time equivalents concerns year 2014. Professors, Associate Professors and Assistant Professors are counted as faculty.

The other component of BIFAKTH, the mean field normalized journal impact (mjcf) for a unit A (a department, for example), is an indicator that shows the citation impact of the journals in which A has published. The indicator normalizes for the variation of citation patterns between subject fields. It is calculated as the mean of the field normalized citation impact of the journals (Jcf of the journals) in which A has published. Regarding publication period, publications belonging to A and published from x – 3 to x – 1, where x is the year of analysis, are taken into account. The publications used for the calculation of indicator values are Web of Science publications recorded in DiVA. Only publications of the types “Article” and “Review” are taken into account.

The Jcf of a journal is the average of the field normalized citation rates of the articles and reviews in the journal. In the calculation of a Jcf value, the citation rate of each publication P in the journal is compared to a field reference value, a world average citation rate across the publications of the field (or fields) associated with P, where these publications belong to the same document type and are published the same year as P. If A has published multiple publications in the same journal, the Jcf value of the journal is counted one time for each publication. For a publication belonging to A and published in year y (x – 3 ≤ yx – 1, where x is the year of analysis), the Jcf value of the journal of the publication is based on the years y – 5 to y – 1.

In the calculation of mjcf values, fractional counts are used: the author’s fraction of each publication is counted as 1/n, where n is the number of authors of the publication. A unit’s fraction is calculated as m/n, where m is the number of authors affiliated to the unit. However, if an author is affiliated to several units (schools/departments) within KTH, the author’s fraction is equally distributed between the units. A unit’s fraction is then calculated as the sum of its author fractions. The unit’s fraction of a publication is applied also to the Jcf value of the publication: the Jcf value is multiplied by the fraction (cf. Example 3 below).

Since mjcf is based on a mean, a lower limit for the number of (whole) publications is reasonable. The limit is set to 5. For each department with less than 5 publications, the journal indicator value is set to 1, which is the Jcf world average.

There are journals in Web of Science for which Jcf values cannot be obtained. We consider two cases. (a) The journal of a publication published year x’ was established later than year x’ – 1. (b) Each field reference value that is used in the calculation of the Jcf value for the journal is less than 0.5. In both cases (a) and (b), the Jcf value of the journal is set to 1, the Jcf world average.

Let A1, …, Am be the departments of KTH. For a given department Ai (1 ≤ i ≤ m), the BIFAKTH for Ai, BIFAKTH(Ai), is calculated as follows. The mean field normalized journal impact for Ai, mjcf(Ai), is calculated. Then the number of faculty-employed full time equivalents for Ai is obtained. Let FTE(Ai) be this number. Then BIFAKTH(Ai) is given by 

Example 3 provides a numerical example.

Example 3. Let A be a department. Assume that A has three publications, P1, P2, and P3, published in three different journals. Further assumptions are given in the Table 1.

Table 1. Author fractions, Jcf values and author fractions combined with Jcf values

  A author fraction (of publication) Jcf (of journal for A publication) [A Author fraction] x Jcf
P1 1 (1/1) 0.8 0.8
P2 0.25 (1/4) 1 0.25
P3 0.5 (5/10) 2 1
1.75   2.05

Given the assumptions, mjcf(A) is equal to

Let the number of faculty full time equivalents for A be 3, i.e. FTE(A) = 3.  Then BIFAKTH(A) is equal to

For formal definitions of mean field normalized journal impact for A, and of author fraction with respect to a unit and a publication, can be found here (pdf).

Now, for a given department Ai and for the KTH bibliometric funding allocation, BIFAKTH(Ai) is transformed to a relative value: BIFAKTH(Ai) divided by the sum of BIFAKTH values over the KTH departments. Let BIFA-relKTH(Ai) be this relative value. Formally:

where m, as indicated above, is the number of KTH departments.

Funds are not allocated to the KTH departments directly, but to their schools. Let S be a school, and let A1m, …, Akm be its departments. The BIFA-relKTH value for S, BIFA-relKTH(S), is the sum of the corresponding values of its departments. More formally,

Finally, S is assigned the proportion, expressed by BIFA-relKTH(S), of the funds associated with bibliometric funding allocation at KTH.      

Top page top