Co-authorship networks and scientific performance: an empirical analysis using the generalized extreme value distribution

Abstract

This paper aims to explore the effects of collaborative behaviour on scholar scientific performance. Individual network measures related to scholar centrality as well as attitude to collaborate with others are derived from co-authorship networks in a given scientific community (i.e. Italian academic statisticians). Co-authorship information have been collected from three data sources of national-based, discipline-based, and international-based high-impact publications. Both network and individual covariates are used to model individual h-index by generalized extreme value distribution. Results show a positive association between performance and actors' central position in the network. Having a large number of co-authors and occupying central positions are likely to positively affect scientific performance.

Keywords:

• collaboration networks
• scientific performance
• generalized extreme value distribution
• h-index
• network centrality

Acknowledgments

We acknowledge that this work was partially supported by the FRA 2013 project of the University of Trieste titled ‘Esiti del percorso universitario e ingresso nel mondo del lavoro: basi di dati integrate, selezione di variabili e inferenza causale’.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Detailed descriptions are reported in official documents ‘declaratorie’ published in Italian Ministry of University and Research (MIUR) website: http://hubmiur.pubblica.istruzione.it/web/universita/docenti-e-ricercatori.

2. Because of retirements not fulfilled by adequate hirings of young researchers, they dropped to 680 at the end of December 2013. For further details, see http://cercauniversita.cineca.it/php5/docenti/cerca.php.

3. Papers published before 1989 in WoS were not available for the query process due to license restriction in 2010 at our universities. On the contrary, in CIS and PRIN all the included publications authored by an Italian academic statisticians in our population have been considered.

4. For instance, for Italian academic statisticians the largest individual h-index is equal to 28, observed for a statistician in the Statistics for E&T Research subfield.