At the four major Norwegian research universities, eight out of ten of the permanent academic staff are members of a formal research group. The rate of membership varies between academic fields, with the highest percentages in medicine and natural sciences, and the lowest in the humanities where only six out of ten are members.
However, among the group members, only 26% conduct their research “to a large degree” in a formal research group, while 12% conduct their research “to a large degree” in an international network and 18% in an international research network in combination with informal or formal collaboration in the university (Table 3). Two out of ten report that they conduct their research to a large degree alone. Hence, being a member of a formal research group does not necessarily imply that most of their research is conducted within the group.
Among those who are group members, the relative importance of the research group differs much across fields (Table 4). Particularly in the humanities a large proportion of the staff conduct most of their research alone. In the humanities, only 10% of the group members conduct their research “to a large degree” in the group while 45% undertake their research “to a large degree” alone. In contrast, 42% of the members in medicine and health conduct their research “to a large degree” in a formal group, and only 7% conduct their research alone. We find the same difference between the two academic fields in how they are involved in international networks. While only 8% of the humanists conduct their research in an international network in combination with a group, 19% of the researchers in medicine and health do that.
Table 4 Percentages of group members who conduct their research in various ways to “a large degree”
The research group is clearly most important in the field of medicine and health, while undertaking research in an international research network, as the only way of conducting research or in combination with formal or informal collaboration in the institution, is most important in the natural sciences. The results indicate that the traditional field differences persist, even though formal research groups are introduced in all academic fields.
An important question is whether those who are not members of a formal research group differ from the group-members in ways of undertaking research. Are research networks an important alternative to group membership for these academic staff members? Among the non-members, the majority undertake their research alone (38%), while a lower share engage in an international research network (22%).
Researchers above 60 years of age undertake research in international networks to a lesser extent than their younger colleagues. There are generally very small gender differences in the ways of undertaking research. The differences between associate professors and professors are however greater. Figure 1 displays that professors report to a much higher degree than associate professors that they conduct their research in formal groups and in international networks, while associate professors more often research alone.
So far we have undertaken bivariate analyses. However, logistic regression can control for different variables at the same time. Hence, we do four different logistic regressions using each of the four different ways of conducting research “to a large degree” as dependent variables, controlling for gender, age, academic rank, type of group membership, and academic field (Table 5).
Table 5 Logistic regressions for academic staff who conduct their research in various ways “to a large degree”
Doing the regressions subsequently, introducing more and more variables into the model, a first striking result is how much of the variation the model explains increases when academic fields are introduced. This confirms our findings from the bivariate analyses that the importance and use of research groups differ much between the academic fields.
Table 5 shows that being a man has a significant effect on the probability to work alone. However, gender has no other significant effect. Furthermore, the younger the staff, the more they are involved in international research networks. Finally, being a professor increases the probability to work in an international network, and decreases the probability to work alone.
Although being a member of a research group does not necessarily mean that academic staff conduct their research to a great extent within the group, being a member increases the likelihood to do so. Being a member of a group also increases the probability to work in an international network in combination with a group, and decreases the probability to work alone.
Academic field is introduced into the model with the humanities as reference category. The probability to work alone is less for all other academic fields, and the probability to work in a formal research group is higher for all other fields. The probability to work in an international group in combination with a group is also higher for the social sciences, the natural sciences and medicine. These results are in line with the bivariate analyses.
Hence, rank, academic field, and whether or not academic staff are members of a research group seem to be the important factors which affect the way the academic staff conduct their research.
Research collaboration and publication productivity
One of the purposes of the formalization of research groups is to increase the publishing activity of academic staff. Table 6 displays that group members have more publication points (3.1) than non-members (2.6). However, this difference is only significant between researchers in the social sciences, probably due to the small size of the subgroups in the various fields.
Table 6 Number of publication points by different types of group membership
The most productive group members are those who “to a large degree” undertake their research in an international network in combination with formal or informal collaboration with colleagues within the institution. This finding confirms previous studies that have found that international collaboration increases the scientific performance of research groups (Barjak and Robinson 2007; Martin-Sempere et al. 2002; Van Raan 1998) and the productivity of individual researchers (Kyvik and Larsen 1994; Abramo et al. 2011).
There are, however, clear differences across the fields. Humanists who work to “a large degree” alone have significantly more publication points than their colleagues who work to “a large degree” in an international network. Members in the social sciences who work to “a large degree” in an international network have significantly more publication points than social scientists who work “to a large degree” in a formal research group. Members in the natural sciences who conduct their research in an international network, and researchers in medicine who work in formal groups or international networks, have significantly more publication points than their colleagues who work alone. There are no such significant effects between researchers in technology, though this could be because of the small size of the subgroups (reaching from 7 to 80).
The number of publication points also varies between other variables. Men have more publication points than women. Professors are more productive than associate professors. Younger researchers tend to have more publication points than older researchers, but the difference is not significant.
As mentioned, 10% of the members do not have any publication points. Another way of investigating the difference in productivity between members and non-members is hence to see if the share of non-publishers is greater among the non-members. Only 7% of the members are non-publishers, in contrast to 20% of the non-members. Furthermore, the share of non-publishers is greatest among the researchers who work alone “to a great extent” (20%), while only 1% of the researchers that work in an international network in combination with a group “to a large extent” do not have publication points. The share of non-publishers among the researchers working in a formal research group “to a large extent” is 5%.
The share of non-publishers increases with age, from the group of researchers under 40 years of age where only 6% are non-publishers to the group of researchers over 60 where 15% are non-publishers. We find a similar difference between professors among which only 7% do not have publication points while 15% of the associate professors are non-publishers. There are no gender differences in the share of researchers that have and have not publication points.
Research collaboration and publication quality
While the number of publication points is primarily an indicator of publication activity, the percentage of publications in prestigious journals (Level 2 publications) can be regarded as an indicator of the quality of publications (Table 7). Group members have a greater share of their publications at Level 2 (29%) than non-members (23%). However, when 43% of the researchers have no publication at Level 2 and 21% have 50%, the average share of publication at Level 2 usually become a share of publication that few researchers have. Having publications at Level 2 or not might be a better quantitative measure of the scientific output.
Table 7 Percentage of researchers that have publications at Level 2 by different types of group membership
Table 7 displays that while five out of ten of the non-members have publications at Level 2, six out of ten of the group members have publications on this level. We find an even greater difference between researchers who work in an international network “to a large degree” in combination with a group (eight out of ten have publications at Level 2), and those who primarily research alone “to a large degree” (four out of ten).
The share of researchers having publications at Level 2 varies between background variables. While 52% of the females have publications at this level, 40% of the males have. The youngest researchers are also the group where most researchers have publications at Level 2. The greatest difference is between professors where 68% of the staff have publications at Level 2, in contrast to only 42% of the associate professors.
Regression analyses
Table 8 displays the results of the regression analyses with researchers without and with publication points (model 1), publication points (the logarithm, model 2) and researchers without and with publications at Level 2 (model 3) as dependent variables. In these models, we control for membership in a research group, being a leader of a research group, gender, rank, age, academic field, and how group members conduct their research.Footnote 3 We use the humanities as reference category for the same reason as in Table 4. We only include working alone and in an international network in combination with a group since these ways are the only two that have significant effects on the publication output in these models, and including only these two types of conducting research contributes to enough researchers in the reference category.
Table 8 OLS Regression analyses
The results in model 1 indicate that being younger than 60 years of age, being a professor instead of an associate professor, and being a researcher in the natural sciences or medicine instead of in the humanities increases the probability of having publication points. Being a man or a woman has no significant effect. Being a member of a group instead of a non-member increases the probability to have publication points, and being a leader of a group increases the probability of having publication points even more. Finally, conducting research “to a large degree” in an international network in combination with a group has the strongest effect on publication output.
Model 2 displays which factors have a significant effect on the number of publication points, where we have excluded the researchers with no publications. This is an OLS-regression where the dependent variable is the logarithm of publication points. Being a researcher in all fields but the humanities decreases the number of publications. This reflects the bias of the publication points, where researchers in the humanities have more publication points due to fewer co-authors. Also, being a researcher under 50 years of age instead of over 60 increases the number of publication points, and so does being a professor instead of an associate professor.
In model 3, being a leader of a research group, being younger than 40 years of age, being a professor, researcher in the natural sciences or medicine, and doing research in an international network increase the possibility of having publication points at Level 2. However, although collaboration in international networks has a significant effect, the age effect and the rank effect have stronger impact on the probability of having publications at Level 2.