![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
The present article examines whether French female researchers face a glass ceiling, an invisible barrier to promotion. Using an original database from the National Institute for Agricultural Research (INRA), we estimate duration models for promotions. The methodology used allows us to take into account censored observations and unobserved heterogeneity. Our results show a significant gender effect that does not contradict the glass-ceiling hypothesis. In addition, factors that boost promotion seem to be radically different according to gender and we present evidence that promotion strategies are different for males and females.
Acknowledgements
The authors are grateful to all the participants in the research project ‘Female careers in academia’, especially Christine Musselin and Vincent Mangematin. We would also like to thank Stephen Bazen and Bart Cockx, for their helpful comments on earlier drafts of the article, as well as the INRA, for providing access to the data and financial support. Any errors are our own.
Notes
1In recent years, a small number of scientists have been recruited directly as senior researchers. As this type of career path does not constitute promotion within INRA, it was not analysed in this study.
2In order to obtain an ‘HDR’ diploma, also known as a ‘Habilitation’, a candidate must prepare a written document outlining his or her contribution to science. As for a PhD, this document must then be defended in front of a jury.
3The impact factor is complete for all publications from the early 1980s onwards, but it may be more imprecise for earlier publications. This must then be taken into account in the econometric results and comments. We therefore introduced a cross variable between COH1 and PUB_SCORE, in order to treat the potential collinearity.
4Various specifications for distributions of durations (log-normal, weibull, etc.) were tested. The results were not affected by the specifications chosen.