The ‘Glass Ceiling’ in Research: The Mystery Solved?

Editor’s Note: This article, originally published in French by the Observatory of University Ethics on March 4, 2022, was translated into English by the Observatory before being edited to align with Minding the Campus’s style guidelines. It is crossposted here with permission.

The expression “glass ceiling” is a polysemic metaphor, at least in its use.

The “glass ceiling” can refer to the discrimination that a demographic group experiences in accessing the highest positions in the socio-professional or political hierarchy. We can thus speak of a glass ceiling for women, if there is discrimination in access to positions or promotion. To the extent that we know the explanation for the blockage here, namely discrimination, this first meaning, in the use of the expression, seems distorted.

There is another meaning of the term that is more in line with the glass metaphor. The glass ceiling can refer to invisible barriers in access to positions of responsibility without it being possible to explain it well. From this point of view, it is only an observation of statistical disproportions or a lack of parity in high positions. The explanatory part then remains a mystery. Thus, we can read:

[…] the ‘glass ceiling.’ This expression, popularized by the American press in the 1980s, describes a situation where women have less access than men to the highest hierarchical positions, without it being possible to identify the reasons why they are not promoted or recruited at these levels. (Fremigacci, Gobillon, Meurs, & Roux, 2016).

The “Glass Ceiling”: A Possibly Inappropriate Metaphor

Regardless of the definition adopted above, the metaphorical glass ceiling may not be appropriate for characterizing attrition from a demographic group at the top, given that there is empirical evidence to explain the disparities by preferences or productivity.
Here we are interested in the glass ceiling mentioned in the gender disparities in the academic and research world. There is a statistical trend: a lower presence of women at the top of the hierarchy.
The glass ceiling at the University is well known in France, with women being a minority among lecturers—44 percent in 2017—and even more so among university professors—24 percent in 2017.¹

It is notable that this imbalance also results from current recruitments and not only from those of the past.² Finally, to talk about the “glass ceiling” is to make an implicit comparison between two statistics: that of the share of women in certain jobs—which is—and that of a reference population—which should be. This raises the question of the right standard to use to talk about the “glass ceiling”: should we take the share of women in the entire population, in the pool of candidates, in all candidates or even in all candidates, taking into account characteristics—age, seniority, scientific productivity?

Considering the situation of the past in a static way, at the end of 2006 at the National Center for Scientific Research (CNRS), we note: “At the end of 2006 it had 26 people, [78 percent] of whom were women, including 43 researchers, 11 percent of whom were women.

The proportion of women is very unequal depending on the disciplines:

1. 17 percent of researchers in mathematics,
2. 18 percent in physics,
3. 19 percent in engineering sciences,
4. 31 percent in chemistry,
5. 39 percent in life sciences,
6. 44 percent in human and social sciences.

For all disciplines, they were 38 percent of research officers and 22 percent among research directors. By comparison, at university, 40 percent of lecturers, 18 percent of professors. (Haton, 2014). Two questions therefore arise. Are there explanations for this under-representation of women in the academic literature? If so, are there other explanations than employer discrimination?

The “Glass Ceiling” and Its Halo of Explanations in Literature

The Ministry of Higher Education and Research studied parity among teacher-researchers (Tourbeaux, 2016). It notes that compared to the pool of candidates, there are relatively fewer actual female candidates compared to men, and suggests: “This situation could result from a phenomenon of self-censorship, that is indirect discrimination.” This example illustrates the ambiguity of the expression “indirect discrimination” which in reality refers here to self-censorship, that is, a decision by potential candidates. There is a shift from factors internal to women—what depends on them—to external factors—what does not depend on them.

Regarding sociologists, Alain Chenu and Olivier Martin (2016) conclude at the end of their analysis, on the glass ceiling among academics, that the stereotypes in force in the professional environment are responsible for the exclusion of women from positions of responsibility (Chenu & Martin, 2016):

The case analyzed therefore provides solid indications that inequalities between women and men in access to the sociology professorship do not in any way result from a relative inferiority in the skills of women who enter sociology. It seems that the main cause is a differentiation, especially noticeable beyond the age of 35, sometimes real and sometimes imputed, of commitments in the sphere of family life and the weight of stereotypes in force (including in a professional environment rather considered attentive to discrimination) which tend to exclude women from positions of responsibility. Breaking down these different effects more precisely would require empirical investigations that go beyond the scope of the sources used here.

However, the authors previously noted that the differential in promotion from lecturer to university professor seems to be explained by productivity, measured by the number of publications and citations:

Professors have an average H index of 6,9, meaning that to one decimal place, seven of their publications have been the subject of at least seven citations identified in Google Scholar. Among lecturers who were not promoted at the end of the fifteen years observed, the score is 3,7. Access to the professorship therefore seems to sanction better scientific productivity. Among lecturers who were not promoted, the productivity levels are almost exactly the same for women and men, and the same is true for those who became professors. Overall, it therefore seems that if men are more often promoted than women, it is because there is, among them, a greater proportion of individuals who manage to maintain a high level of publication during their career as lecturers. The lesser access of women to the body of professors therefore seems to sanction lower scientific productivity, and not result from discrimination which, with equal scientific records, would favor men. (Chenu & Martin, 2016)

Apart from the “glass ceiling” itself, there are a multitude of external factors—to the group supposed to be subject to them—which would constitute it, i.e. a sort of halo:

• The persistence of stereotypes in the academic world as well as a mainly masculine “habitus” (Bourdieu) according to which there would be a greater “naturalness” for men to be in a position at the university with regard to the qualities put forward—self-affirmation, technique, rationality, etc.
• The “Matilda effect,” which highlights a tendency to “invisibilize” or underestimate the work of women in the scientific field, which has a disqualifying effect on their contribution.
• Conversely, there is a “Matthew effect” according to which, in the scientific field, “the more you have, the more you receive.” It thus leads to an accumulation of advantages for the benefit of men.
• The ‘Old boys’ club’ effect: existence of networks of in-groups, a tendency to favour access, even informal, to certain groups which are rather masculine. Women can have access to them but the cost of entry is higher.
• The ‘Sticky floor’: household chores weigh more heavily on women, not only in their homes but also in academia—the notion of ‘academic housework.’
• The concept of university/research organization as a “greedy institution:” idea according to which the researcher must be “fully invested in his/her missions.” Hence the emergence of the feeling of being overwhelmed by one’s professional life which can also concern men but which, currently, weighs more on women—work/family conflict in particular. (Fusulier, Lhenry, & Hermann, 2018).

These explanations, which favor external factors—and not preferences constituting an internal factor—are qualitative. It would certainly be interesting to explore these avenues further using a quantitative methodology to test the existence of an effect and, if applicable, its magnitude.

In doing so, through testing methods, gender biases were able to be tested with randomized methods in the context of university recruitment in the United States, including with an estimate of the salary gap for identical profiles, in the material and life sciences (Moss-Racusin, Dovidio, Brescoll, Graham, & Handelsman, 2012). The authors highlight significant biases, with the male candidate John being perceived as being more competent than his counterpart Jennifer, benefiting from better tutoring and above all a higher salary—14 percent more in favor of men. Men and women who recruit would be affected by these biases. Note, however, that the sample is made up of only 127 academics.

In the United States, some studies show that sexist bias can appear in exchanges between economists on professional forums prior to recruitment, with inappropriate comments about physical appearance suggesting in the short term an evaluation of candidates not exclusively based on their professional skills and in the longer term a potential risk of sexual harassment (Wu, 2017).

Preferences and Productivity: Explanations for the Disparities?

Contrary to American studies showing biases against women, when it comes to recruitment in science in the United States, university psychologists Stephen Ceci and Wendy Williams show that women are twice as likely to be recruited as men with the same profile, while recruiters in economics—a highly mathematical discipline—show no gender preference (Williams & Ceci, 2015). By using the testing method, recruiters are unaware that they are receiving fictitious CVs, which makes it possible to gauge real practices.

The United States Academy of Sciences, in its 2009 report, shows through more than 300 pages of statistics, polls, and surveys that there is no significant gender bias that women are victims of in science—physical sciences, mathematics, computer science, engineering, life sciences, earth sciences—with men and women having the same career opportunities (United States Academy of Sciences, 2009).

In their article Are Women Less Productive in Research? A Critical Review of Empirical Studies and an Econometric Analysis on Panel Data for CNRS and University Physicists in France, economists Jacques Mairesse and Michele Pezzoni write that there has been a broad consensus in academic research for thirty years about the gender gap in productivity among scientists, but that the reasons for this differential remain a mystery subject to debate: “The lower productivity of women in science is an old research question, examined in the last thirty years by economists and sociologists of science who refer to it as the ‘gender productivity gap’ or ‘gender productivity bias,’ or more simply the ‘productivity mystery'” (Cole and Zuckerman [1984]). There is a broad consensus in the literature that measured productivity is lower for women scientists than for their male counterparts, in virtually all disciplines and regardless of the productivity indicator adopted. However, the reasons why this is so are still a matter of debate, remaining largely an enigma.³ (Mairesse & Pezzoni, 2015). The two economists give their estimate of this productivity differential, estimating that “the productivity of female physicists in terms of publication is much lower, by around a third on average, compared to that of their male colleagues.”

These results are found in the United States, including in feminized disciplines such as psychology, with 34 percent fewer publications for women among assistant professors and 27 percent among full professors, with no gap found among associate professors (Eagly & Miller, 2016).

Despite the fact that Mairesse and Pezzoni seem to blame unequal chances of promotion and family commitments, it is important to keep in mind that work has been carried out on these chances of promotion, precisely to inquire about possible discrimination in the promotions in question. Indeed, they qualify their statement as follows:

We conclude, however, that this difference in productivity disappears for the CNRS and that it is even reversed for universities when we take into account several factors, notably those linked to unequal chances of promotion and to notable discontinuities in publications, which may reflect strong family commitments.

Sociologist Anne Revillard comments on the work relating to the promotions of female economists in higher education in France (Bosquet, Combes, & Garcia-Panalosa, 2014): “However, from the perspective of a more global analysis of gender inequalities, the observation that differences in scientific productivity contribute 40 to 43 percent of the gap between women and men in promotions acquired (Table 1) would deserve to be endogenized” (Revillard, 2014).

Anne Revillard adds:

Several major conclusions emerge from the exploitation of this database. First of all, the probability of access to PU and DR positions differs significantly according to gender: among all teacher-researchers, the probability of occupying a PU position is 39,9% for men and 17,6% for women, a gap of 22,3 points, and at the CNRS, the probability of occupying a DR position is 44,8% for men and 17,9% for women, a gap of 26,9 points. The breakdown of this differential in the positions achieved shows that the age differences⁴ and publication score explain 70% of the gap in the case of the CNRS and 80% in the case of the University.

Clément Bosquet, Pierre-Philippe Combes and Cecilia Garcia-Penalosa, in their article entitled Why do women occupy fewer positions of responsibility? An analysis of university promotions in economics shows that the promotion differential for economists at the University and at the CNRS is explained by fewer female candidates, fewer publications by women and a younger age. They find absolutely no discrimination in access to positions and conclude: “It seems relatively clear, however, that a total absence of discrimination at the time of choosing the person promoted among the candidates would not be enough to significantly increase the promotion rate of women.”

Bosquet et al. highlight “that if women are less promoted than men in the French academic world, it is above all because they are less often candidates, even with identical observable characteristics, and in particular publications” and “that women did not have a lower probability of success than men” in the competitive examinations for the Agrégation du Supérieur in economic sciences between 1992 and 2008 and in the competitive examinations for Research Director at the CNRS between 1996 and 2008. The entire gap in promotion by gender is explained by productivity—the number and quality of publications—or 43 percent of the explanation, by the choice of applying—fewer female applications—or 20% of the explanation and 37 percent by age—a certain seniority being required to apply. If we consider that age, a proxy for seniority, is a productive characteristic, then 80% of the promotion gap is due to productivity, with only 20% strictly related to the choice to apply. It is also remarkable that the authors, testing the hypothesis of family constraints likely to dissuade women due to geographical mobility, conclude that this explanation is rejected by the data: “women apply even less than men to the CNRS competition than to the University, although then no move is imposed.”

This result, which is nevertheless clear regarding the futility of the impact of geographical mobility on the passage of the competitive examination for promotion, appears in a completely different way in another academic publication by summarizing it. It is then a question of the dissuasive impact of geographical mobility on female applications while the publication by Bosquet et al. tested and rejected this hypothesis: “Bosquet et al. (2014) question the rarity of women among university professors and research directors at the CNRS. Using panel data for economic sciences and relying on the particularities of the recruitment methods of university professors in this discipline (recruitment by competitive examination of aggregation), they put forward the idea that the under-representation of women in this administrative body is probably linked to differences in the job offer. Women tend to apply less for promotions than their male counterparts, perhaps due to a lack of self-confidence or family constraints (promotion very often involves geographical mobility)” (Fremigacci, Gobillon, Meurs, & Roux, 2016).

The European Central Bank has examined the differential in the promotion of economists within its ranks. The explanations for the differential in promotion are the lower propensity of women to compete, the fact that they apply less, despite a greater chance of being selected if they do apply.⁵ (Hospido, Laevenn, & Lamo, 2019).

Regarding the doctoral supervision bonus, in 2015 it appears that there are relatively fewer women among the candidates compared to the pool of teacher-researchers (Tourbeaux, 2017). Among the lecturers in the normal class, the pool includes 44 percent women compared to 34 percent of candidates for the doctoral supervision bonus and 33 percent of women among the laureates. This trend towards fewer female candidates compared to the pool is found at all grades of lecturer and almost all those of university professor.

However, the doctoral supervision bonus amounts to 4 euros per year, which is likely to result in a substantial pay gap.⁶

As Antoine Petit writes in his letter to female employees of the CNRS: “Studies show that, overall, the rate of self-censorship is higher among women than among men. The proportion of female candidates in competitions at all levels is very often lower than that of the corresponding pool. However, it is difficult to promote people who do not apply!” (Petit, 2019).

Initiatives Against the “Glass Ceiling” at the CNRS

Here are some general data on the gender distribution of staff at the CNRS:

In 2013, the CNRS employed 32 staff:

– 24 permanent staff, including 955 percent women
– 7 non-permanent staff, including 965 percent women

Among the permanent staff, we count:

– 11 researchers, 204 percent of whom are women
– 13 engineers and technicians (IT) including 751 percent women: – 50,4 engineers, including 9 percent women – 702 technicians including 44,5 percent women

While parity has generally been achieved among engineers and technicians—with the exception of research engineers—it has not yet been achieved among researchers. (“Parity in CNRS professions 2013”, published in December 2014, the contents of which can be consulted interactively here).

So, logically, provided that there is no difference in aptitude between men and women, we should find the same proportion among CNRS medal winners.

Let’s look at the numbers based on the CNRS source, or more precisely on the institution’s website. The data collection consists of counting silver and bronze medalists, with perceived gender identity based on the photographic appearance and first name of the winners. (For the silver medalists, here is the source. For the bronze medalists, here is the source.)

In 2016: Total silver medalists: 16 Women: 8 (50%)   Total bronze medalists: 40 Women: 22 (55%)

In 2017: Total silver medalists: 20 Women: 10 (50%)   Total bronze medalists: 40 Women: 26 (65%)

In 2018: Total silver medalists: 20 Women: 9 (45%)   Total bronze medalists: 42 Women: 21 (50%)

2019: Total silver medalists: 20 Women: 11 (55%) Total bronze medalists: 43 Women: 21 (49%)

2020: Total silver medalists: 22 Women: 11 (50%) Total bronze medalists: 46 Women: 27 (59%)

It is clear that instead of rewarding women according to their share in the overall research workforce, i.e. 33%, the CNRS is trying to respect gender parity, despite a non-parity pool. This result suggests that if there is discrimination, it does not necessarily go in the direction usually decried.

Moreover, in the CNRS info of July 17, 2019, we have confirmation of what intuition suggests by examining these statistics: “Antoine Petit, President and CEO of the CNRS […] has already taken several measures since his arrival at the head of the organization such as parity in the allocation of CNRS medals or the promotion of female researchers in proportion to the percentage of promotable women.”⁷

Is the Case of Donna Strickland, Nobel Prize Winner in Physical Sciences, Emblematic?

The press welcomed the consecration of Donna Strickland, the third woman to be awarded the Nobel Prize in Physics in 2018 after Marie Curie in 1903 and Maria Goeppert-Mayer in 1963. However, it turns out that a controversy has broken out over the fact that Strickland did not have a Wikipedia page before receiving the Nobel Prize. It then becomes clear that the heart of the problem lies in the fact that she was only an assistant professor and not a university professor. But why? Was she discriminated against in her career advancement? Here is what she herself says (Crowe, 2018): As one BBC Newshour reporter put it: ‘Why don’t you have the rank of university professor, given your prestigious work?’ Strickland, who will receive a quarter of the Nobel Prize, or about $250, for his work on ultrashort pulses using an optical lattice to boost their intensity for lasers, replied: “I never applied.”⁸ And again: “Your rank as a lecturer and not as a university professor has attracted a lot of attention. What did you respond to it? I really feel sorry for the university because it’s hardly their fault. This is precisely what people do not know, namely that a university professor, although having a different title, does not necessarily get a higher salary and that I do not lose my position—if I do not apply to be a university professor. So I never filled out the application form. I’m completely responsible for it. I think people assume it’s because I’m a woman, and I’m being held back. I’m just lazy. I do what I want to do and it wasn’t worth it to me.”⁹

The Anticipation of Non-Existent Discrimination Is Likely to Demotivate Women.

Psychological traits have a significant effect on remuneration, with internal locus of control and the propensity to face challenges being conducive to higher incomes, results obtained from data on Russia (Linz & Semykina, 2005). Internal locus of control consists of attributing responsibility for what happens to oneself while external locus of control is a propensity to feel powerless in the face of external forces—culture, society, teachers, bad luck, etc.

It is therefore important to keep in mind that the dissemination of messages suggesting that discrimination in pay, promotions, or hiring is endemic when the data do not suggest this could have a counterproductive effect on women’s expectations in professional life. As some studies show, even in the absence of discrimination in promotions, women apply less, and it cannot be ruled out that they may anticipate, in some cases, non-existent discrimination (Bosquet, Combes, & Garcia-Panalosa, 2014) but anticipated discrimination undermines women’s ambition (Fisk & Overton, 2019).

In particular, some authors regret that the discourse on discrimination persists, although the under-representation of women in hard sciences no longer seems to be explained by discrimination.¹⁰ (Ceci, Kahn, Ginther, & Williams, 2014).

These explanations would benefit from having the echo they deserve. While we obviously find studies reporting recruitment or advancement bias among researchers—sometimes also in the opposite direction to that usually suspected—it seems that overall, in France, we can explain the disparities only by the choices made and scientific productivity, namely the fact of applying and publishing [(Mairesse & Pezzoni, 2015); (Bosquet, Combes, & Garcia-Panalosa, 2014)], a differential which itself is explained, at least in part, by family commitments. Overall, by looking closely at the data, it is clear that the mystery of the “glass ceiling” seems to have been elucidated.

Editor’s Note: The bibliography can be found in the original article posted here. 

Image by W.Scott McGill — Adobe Stock — Asset ID#: 140703731