Mobility is an important aspect of a researcher's life, affecting the career of the scientist in many ways: through the change of host institutions, new career opportunities are chased, positions with higher prestige acquired, stronger collaborations can be created, and novel projects are started. As a result, the collaboration network, the productivity, and the research impact of the studies of the researcher are possibly affected every time a researcher changes their host institution. For all these reasons, mobility - both international and domestic - is rightly considered a very important part of the research career.

At first glance, the outcome of a movement may seem only beneficial. However, this is not always true. The downside to the creation of new opportunities is that old collaborations may collapse and old projects be abandoned. It is thus interesting to study the impact of academic migrations on the collaboration network of a researcher, also considering whether it varies across different career stages.

To this aim, we use a dataset containing details for the scientific publications (extracted from the Scopus repository - www.scopus.com) of 84,141 mobile authors that have moved 561,389 times. The key element of our analysis is the author's ego network, which is defined as the set of co-authors (alters) of the given author (ego) while being at a certain institution.

For studying the effects of a host institution change on the ego network, we focus on the analysis of a set of characteristics such as the change in the ego network size, and the similarity between the ego networks before and after the movement. We first consider an overall exploratory analysis, where we study all movements together, and analyse (i) the percentage change of the ego network size, (ii) the overlap between the ego networks across each movement, (iii) how much of the old network “survives" in the new one and (iv) to what extent the new network is built on old collaborations. We then drill down the analysis by separating movements that result in an increase of ego network size from those that result in a reduction. Finally, we separately analyse the same figures based on the different career stages of researchers.

We find that, on average, after a movement, the collaboration network tends to expand. However, the overlap between the old and new networks is not particularly high (Jaccard similarity between 10%-20%) and this result holds true both when the network size increases and when it decreases. Also, on average, approximately 30% of the old collaborations carry over to the new network, and they amount to approximately 20% of the latter. This turnover is related to the cognitive effort required to nurture collaborations: one cannot just add new relationships because time needs to be invested for their maintenance. In order to bring in new collaborations, one needs to replace old ones. The third finding is that, when the network size increases after a movement, still not all old collaborations are maintained (Figure 1). On average, only 36% of old collaborations are carried over to the new network. When the network shrinks, this fraction is even smaller (around 14%) but the old collaborations relatively weigh much more in the new network (around 36%). Movements tend, on average, to be more disruptive as the career progresses, partly because the networks after a movement are larger for senior people but also because former collaborations tend to be curtailed more.

Figure 1: Distribution of the changes in size (size_ratio) and composition of collaboration networks (similarity net_sim between old and new, fraction net_dep of old collaborations in the new network, fraction net_kept of old collaborations that keep collaborating after a movement) when the network size increases (blue) and decreases (red) after a movement.

Principal reference:

Dynamics of Scientific Collaboration Networks Due to Academic Migrations. P. Paraskevopoulos, C. Boldrini, A. Passarella, M. Conti. International Conference on Social Informatics (SocInfo) 2020, pp. 283-296.

Written by: Laura Pollacci