Gender bias in journals of gender studies
Authors
- First Online:
- Received:
DOI: 10.1007/s11192-012-0661-5
- Cite this article as:
- Kretschmer, H., Kundra, R., Beaver, D.d. et al. Scientometrics (2012) 93: 135. doi:10.1007/s11192-012-0661-5
- 13 Citations
- 485 Views
Abstract
The causes of gender bias favoring men in scientific and scholarly systems are complex and related to overall gender relationships in most of the countries of the world. An as yet unanswered question is whether in research publication gender bias is equally distributed over scientific disciplines and fields or if that bias reflects a closer relation to the subject matter. We expected less gender bias with respect to subject matter, and so analysed 14 journals of gender studies using several methods and indicators. The results confirm our expectation: the very high position of women in co-operation is striking; female scientists are relatively overrepresented as first authors in articles. Collaboration behaviour in gender studies differs from that of authors in PNAS. The pattern of gender studies reflects associations between authors of different productivity, or “masters” and “apprentices” but the PNAS pattern reflects associations between authors of roughly the same productivity, or “peers”. It would be interesting to extend the analysis of these three-dimensional collaboration patterns further, to see whether a similar characterization holds, what it might imply about the patterns of authorship in different areas, what those patterns might imply about the role of collaboration, and whether there are differences between females and males in collaboration patterns.
Keywords
Gender biasCo-operationSocial networksCo-authorshipCollaboration patternsMathematical Subject Classification (2000)
62689194JEL Classification
C0C02C3C31C46Introduction
The scientific and scholarly systems reflect a strong gender bias favoring men which makes it more difficult for women researchers to fully develop their potential and careers. The causes of that gender bias are complex and related to overall gender relationships in most of the countries of the world. An as yet unanswered question is whether in research publication gender bias is equally distributed over scientific disciplines and fields or if that bias is reflects a closer relation to the subject matter.
Bibliometric indicators of gender co-operation
Author order in the by-line and concentration measures
Three-dimensional collaboration patterns of the journals “PNAS”, “Psychology of Women Quarterly” and of the mixed bibliography of 14 journals of gender studies
The special methods and indicators are explained in each section.
Data
Titles of the 14 journals
Journal | Number of papers | Number of authors |
---|---|---|
Affilia | 620 | 1,058 |
Feminism and psychology | 666 | 999 |
Gender and society | 704 | 1,034 |
Gender technology and development | 230 | 319 |
Men and masculinities | 298 | 433 |
Psychology of women quarterly | 1,111 | 2,473 |
Sexualities | 427 | 536 |
Signs | 1,396 | 1,676 |
Social politics | 283 | 364 |
Women’s studies international forum | 1,631 | 2,056 |
European journal of women studies | 366 | 454 |
Feminist theory | 228 | 258 |
Indian journal of gender studies | 291 | 373 |
Feminist economics | 398 | 658 |
SUM | 8,649 | 12,691 |
Bibliometric indicators of gender co-operation
- Participation
counts the number of publications with at least one author of a given gender
- Contribution
measures the involvement of each gender in the production of a publication assuming that each author contributed the same amount
- Number of authors
total count of the authors of a given gender in each publication
Calculation of female participation, contribution and total count
Gender | Female participation | Female contribution | Female total count | |||
---|---|---|---|---|---|---|
F | M | M | M | 1 | 1/4 | 1 |
F | F | M | M | 1 | 2/4 | 2 |
F | F | F | M | 1 | 3/4 | 3 |
F | F | F | F | 1 | 4/4 | 4 |
Bibliometric indicators of gender co-operation
Participation of women in % | Participation of men in % | Contribution of women, in% | Number of female authors in % | |
---|---|---|---|---|
Naldi et al. | 45.8 | 94.7 | 19.5 | 22.2 |
COLLNET bibliometric | 65.3 | 76 | 45 | 47.9 |
14 Journals of gender studies | 91.6 | 17.3 | 87.5 | 85.6 |
Naldi et al. (2004)—Science and Technological Performance by Gender
COLLNET (COLLNET-Collaboration Network in Science and in Technology: www.collnet.de)
Journals of Gender Studies (Kretschmer, Kundra, Beaver and Kretschmer)
The COLLNET results in comparison with the results by Naldi et al. have already been published by Kretschmer and Aguillo (2004).
The bibliometric study of Naldi and Parenti (2002) is based on a data sample of 10,000 items published during the year 1995 in scientific journals of international relevance and written by 35,000 authors from six European countries. Women’s Participation amounted to only 45.8% of all items as opposed to the much greater male Participation of 94.7%. Women’s Contribution amounted to about 1/5 (19.5%), approximately the same as the Number of female authors, 22.2% of all authors. Although there are differences in these results related to disciplines and countries, in general the low position of women in co-operation is striking.
The bibliometric study of 64 COLLNET members from 20 countries examined lifetime productivity until June 2003. This study is based on a data sample of 223 multi-authored publications between at least two COLLNET members. Women’s Participation amounted to 65.3% of all items and men’s Participation 76%. Although the difference between the participation of women and the participation of men is statistically significant (χ^{2} test, p < 0.01) it is clearly less than in the Naldi and Parenti study. Women’s Contribution, 45%, almost equalled the Number of female authors, 47.9%. In sum, female and male COLLNET members are rather equally distributed in co-operation.
The results of the analysis of the co-operation among COLLNET members differ strongly from those of gender studies in the natural sciences (Naldi et al.) which show a very low participation rate of women in collaboration activities. Female COLLNET members’ collaboration patterns are nearly equally as distributed as male members’. But further, the new results found in Journals of Gender studies even surpass the COLLNET results, insofar as female collaboration is concerned.
Bibliometric analysis of 14 journals of gender studies shows that Women’s Participation amounted to 91.6% of all papers as opposed to the much lesser male Participation of only 17.3%. As before, women’s Contribution, 87.5%, and the Number of female authors, 85.6%, were approximately equal. Although there are some minor differences in these results related to each journal individually, the very high position of women in co-operation is striking and most probably related to the subject matter of the journals.
Author order in the by-line and concentration measures
The order of the authors in the by-line is taken into consideration with help of concentration measures. The concentration of females (COF) in position x (x = 1, 2, 3, 4…) in the by-line is defined here as the ratio between the percentage of females in position x and the percentage of females in total (in the present study of gender studies’ journals: Number of female authors as a percentage of all authors = 85.6%).
If there is equal concentration by gender, the expected percentage of females in position x of the by-line should be equal to the percentage of females in total: COF = 1. If, for example, COF is higher than 1 in position 1 (first author), then females are relatively overrepresented as first authors, and vice versa if lower.
By analogy, the concentration of males (COM) in position x in the by-line is defined here as the ratio between the percentage of males in position x and the percentage of males in total (in the present study of gender studies’ journals: Number of male authors as a percentage of all authors = 14.4%).
Three-dimensional collaboration patterns of the journals PNAS, “Psychology of Women Quarterly” and of the mixed bibliographies of 14 journals
For about a decade social network analysis (SNA) has been used successfully in the information sciences, as well as in studies of collaboration in science. A variety of applications of SNA is available (Wasserman and Faust 1994; Otte and Rousseau 2002) both for studies in large and in small networks.
“Many investigations of scientific collaboration are based on statistical analyses of large networks constructed from bibliographic repositories. These investigations often rely on a wealth of bibliographic data, but very little or no other information about the individuals in the network, and thus, fail to illustrate the broader social and academic landscape in which collaboration takes place” (Pepe et al. 2009, p. 1).
In other words, in investigations of large networks information about “Who is collaborating with whom” is mostly missing.
However, the model of well-ordered three-dimensional distributions of co-author pairs’ frequencies in networks (Kretschmer and Kretschmer, 2007, 2009) says that, depending on the personal characteristics of collaborators (for example, author productivity or others), a special fundamental principle of social group formation is a determining factor in shaping preferences in co-authorship between individual scientists. This principle is based on similarities/dissimilarities and the corresponding consideration of this and other complementarities are a crucial determinant of the mathematical model.
Consequently, fundamental principles of social group formation produce well-ordered structures (called “Social Gestalts”) with different shapes depending on changing personalities and situations. This model has already been applied to 52 large co-authorship networks (Kretschmer and Kretschmer 2009). For 96% of them the squared multiple R is larger than 0.98 and for 77% of the 52 networks even larger than 0.99.
Method of counting co-author Pairs, based on SNA
For the purposes of analysis, a social network can be considered as consisting of two sets, a set of n nodes (individuals) and a set of m edges (undirected relations) between pairs of the nodes. The degree of a node F_{x} with x (x = 1, 2…n) is equal to the number of nodes (or edges) that are attached to the node F_{x}. In co-authorship networks between two authors (nodes) F_{x} and F_{y}, there exists an edge if both have published at least one publication together.
An author’s productivity is measured by his number of publications. The number of publications i per author F_{x} or j per possible co-author F_{y}, respectively, are determined by using the ′normal count procedure′. Each time the name of an author appears, it is counted.
Artificial table of co-author pairs N_{ij}
i/j | 1 | 2 | 3 | N_{i} |
---|---|---|---|---|
1 | 30 | 20 | 10 | 60 |
2 | 20 | 25 | 5 | 50 |
3 | 10 | 5 | 2 | 17 |
N_{j} | 60 | 50 | 17 | N = 127 |
In other words: N_{ij} is equal to the sum of co-author pairs of authors who have the number of publications i in co-authorship with authors who have the number of publications j. N is equal to the total sum of degrees of all n nodes (all authors F_{x}) in a network, equal to the total sum of pairs.
Logarithmic binning procedure
Distributions of this kind of co-author pairs’ frequencies (N_{ij}) have already been published (Kretschmer and Kretschmer 2007; Kundra et al. 2008; Hanning et al. 2008). However, these distributions were restricted to i_{max} = 31.
Usually the stochastic noise increases with higher productivity because of the decreasing number of authors. We intend to overcome this problem in this paper with help of the logarithmic binning procedure. Newman has already proposed in 2005 using the logarithmic binning procedure for the log–log scale plot of power functions. To get a good fit of a straight line (log–log scale plot of power functions, for example Lotka’s distribution), we need to bin the data i into exponentially wider bins. Each bin is a fixed multiple wider than the one before it. For example, choosing the multiplier of 2 we receive the intervals 1–2, 2–4, 4–8, 8–16, etc… For each bin we have ordered the corresponding first value of i (or j) to this bin. Thus, the sequence of bins i’ or j’ is:
i’ (i’ = 1, 2, 4, 8, 16, 32, 64, 128, 256…). The same holds for the bins j’. The sizes or widths of the bins (∆i’) are: 1, 2, 4, 8, 16 etc… The same holds for (∆j’).
Matrix of N_{ij}^{S} (PNAS) with N = 634,014
i’(bin)/j’(bin) | 1 | 2 | 4 | 8 | 16 | 32 | 64 | Sum |
---|---|---|---|---|---|---|---|---|
1 | 13,2068 | 77,429 | 41,720 | 18,484 | 6,847 | 1,954 | 196 | 278,698 |
2 | 77,429 | 54,516 | 30,168 | 14,087 | 5,564 | 1,708 | 192 | 183,664 |
4 | 41,720 | 30,168 | 17,390 | 8,203 | 3,184 | 1,021 | 105 | 101,791 |
8 | 18,484 | 14,087 | 8,203 | 3,718 | 1,371 | 428 | 52 | 46,343 |
16 | 6,847 | 5,564 | 3,184 | 1,371 | 528 | 153 | 16 | 17,663 |
32 | 1,954 | 1,708 | 1,021 | 428 | 153 | 24 | 3 | 5,291 |
64 | 196 | 192 | 105 | 52 | 16 | 3 | 0 | 564 |
Sum | 27,8698 | 18,3664 | 10,1791 | 46,343 | 17,663 | 5,291 | 564 | N = 634,014 |
Matrix of N_{ij}^{S} (Psychology of Women Quarterly) with N = 4,324
i’(bin)/j’(bin) | 1 | 2 | 4 | 8 | Sum |
---|---|---|---|---|---|
1 | 2,228 | 499 | 186 | 143 | 3,056 |
2 | 499 | 154 | 53 | 52 | 758 |
4 | 186 | 53 | 34 | 24 | 297 |
8 | 143 | 52 | 24 | 12 | 231 |
Sum | 3056 | 758 | 297 | 231 | N = 4,342 |
Matrix of N_{ij}^{S} (mixture of 14 journals in women’s and gender studies) with N = 11,996
i’(bin)/j’(bin) | 1 | 2 | 4 | 8 | 16 | Sum |
---|---|---|---|---|---|---|
1 | 6,332 | 1,415 | 424 | 205 | 44 | 8,420 |
2 | 1,415 | 574 | 204 | 69 | 31 | 2,293 |
4 | 424 | 204 | 160 | 50 | 8 | 846 |
8 | 205 | 69 | 50 | 22 | 3 | 349 |
16 | 44 | 31 | 8 | 3 | 2 | 88 |
Sum | 8,420 | 2,293 | 846 | 349 | 88 | N = 11,996 |
Method of visualizing the original data
For visualizing the original data we use the sum of co-author pairs in a bin (cell_{i’j’}), i.e. N_{ij}^{S} directly in dependence on i’(bin) and j’(bin), (cf. Tables 5, 6 and 7). Because log 0 is not given, we are using the value “0” for presentation of N_{ij}^{S} in the tables (cf. Tables 5, 6 and 7) but not for regression analysis.
Method of visualization the three-dimensional collaboration patterns
As the next step in the logarithmic binning procedure: N_{ij}^{S} of a cell (cell_{i’j’}) has to be divided by the width of the bin: (∆i’·∆j’). In other words, the new value in a bin is simply the arithmetic average of all the points in the bin. This new value is called the average co-author pairs’ frequency N_{ij}’.
Using the log–log–log presentation after the logarithmic binning procedure, the sequence of log i’ (rows) is as follows: log i’(log i’ = 0, 0.301, 0.602, 0.903, 1.204, 1.505, 1.806, …); the same holds for log j’ (columns).
The mathematical function for describing the three-dimensional distribution of co-author pairs’ frequencies (N_{ij} or N’_{ij} after logarithmic binning) is a special case derived from Kretschmer’s mathematical model for the intensity function of interpersonal attraction (cf. “Appendix”).
For visualizing the theoretical patterns (Social Gestalts) we use the Function Plot of SYSTAT and the Scatterplot for the empirical patterns.
After the overlay of the empirical distribution and the theoretical pattern into a single frame the goodness-of-fit is highest in the case where the empirical values (dots) are directly placed on the points where two of the theoretical lines intersect. As the distance between the intersection points and the dots increases, the goodness-of-fit decreases.
Results
The results of the application of the model to three networks are shown in Fig. 2.
The journal PNAS (1980–1998)
Journal “Psychology of Women Quarterly” (1976–2011)
Mixed bibliographies of 14 journals in women’s and gender studies (1976–2011) are the source for this current study
All of these three-dimensional distributions are well-ordered with changing shapes. The shapes depend on the accentuation of either similarities or, vice versa, dissimilarities, cf. Fig. 2. Whereas the convex distribution in the first row obtained from the data of the journal PNAS (1980–1998) shows the accentuation of similarities of the co-authors regarding their productivities, the concave distribution in the second row obtained from the data of the journal “Psychology of Women Quarterly” (1976–2011) shows on the other hand the accentuation of dissimilarities of the co-authors. The shape of the distribution in the third row (source: Mixed Bibliographies of 14 Journals) falls in the middle between those of the PNAS and the journal “Psychology of Women Quarterly”.
Explanation: In convex distributions (as in PNAS) the co-author pairs’ frequencies between authors with the same number of publications are higher than those with different numbers of publications. Thus, accentuation of similarities is expressed by convex distributions.
On the contrary in concave distributions (like in the Psychology of Women Quarterly) the co-author pairs’ frequencies between authors with the same number of publications are lower than those with different numbers of publications. Consequently, accentuation of dissimilarities is expressed by concave distributions.
An example for the theoretical predictions for the places of the empirical values in the theoretical patterns is also shown in Fig. 2. The lines of the theoretical patterns are obtained from the mathematical model. The points where two of the lines intersect are the theoretical predictions for the empirical values at those coordinates. The goodness-of-fit is highest in the case where the empirical values correspond exactly to the points obtained from the theoretical Gestalt (under these conditions we obtain after regression analysis: R^{2} = 1). In Fig. 2 the empirical values are presented in form of dots.
Conclusion
- 1.
The very large percentage of women in co-operation is striking and most probably related to the subject matter of the journals.
- 2.
Female scientists are relatively overrepresented as first authors. This result confirms the above mentioned results obtained by bibliometric indicators of gender co-operation. In contrast to the women, however, from the second place on in the by-line men are relatively overrepresented.
- 3.
The three-dimensional collaboration patterns are well-ordered, however, the shape is different from the well-ordered shape of PNAS. Collaboration behaviour in gender studies is different from that in the natural sciences. The accentuation of similarities in productivity of co-authors is shown in PNAS but the accentuation of dissimilarities can be observed in gender studies, especially in the Journal “Psychology of Women Quarterly”
The results confirm our expectation that the strength of gender bias is related to the subject matter of journals, and that it is less expressed in the journals of gender studies.
It would be interesting to extend the analysis of three-dimensional collaboration patterns further, to see whether such a characterization continues to hold, what it might imply about the patterns of authorship in different fields, what those patterns might imply about the role of collaboration, and whether there are differences between females and males in collaboration patterns.
Acknowledgment
Part of this work by one of the authors (Kretschmer H) was supported by the 7th Framework Program by the European Commission, SIS-2010-1.3.3.1. Project full title: “Academic Careers Understood through Measurement and Norms “, Project acronym: ACUMEN.