The under-representation of women in science (especially STEM fields) has been a long-standing concern (Ginther and Kahn 2004, Ceci and Williams 2014). While numerous supply-side explanations have been offered (e.g. gender differences in competitiveness, or incompatibilities of a scientific career with family obligations), there is also the possibility that stereotype threats lead to the undervaluation of work by female researchers (Reuben et al. 2014).
Identifying biases against female researchers is especially challenging at the junior level, as research ability (e.g. the quality of a job-market paper) is not easily measurable. At a later stage, when publications and citations become a meaningful measure of research quality, gender differences in research recognition can be more readily identified.
We investigate whether there are gender differences in becoming a member of one of the most prestigious academies in the US – the American Academy of Arts and Sciences (AAAS) or the National Academy of Sciences (NAS) – from 1960 to 2019. Being elected a member is an extraordinary honour, as this award is granted to the most impactful and established researchers in a field. We select three disciplines that are similar in terms of the main dimensions of research output (publications and citations are key) but differ in the degree of feminisation: psychology, a discipline with a sizeable share of female researchers; mathematics, a clearly masculine field; and economics, which lies in between these two disciplines. In all three disciplines, elected members have a strong publication record, and we compare election probabilities for established male and female researchers with comparable curricula vitae (CVs).
One might argue that obtaining a strong CV is more difficult for female than for male researchers because of different hurdles in the profession (see e.g. Card et al 2020, Sarsons et al. 2021, and Hengel 2022). If so, this will affect the interpretation of results, a point to which we return later.
Construction of the data set
For each of the three disciplines, we construct a sample of active researchers, using a five-step procedure. The sample of active researchers (which we call a ‘risk set’ as they are at risk of getting elected) is based on all authors who publish in the most important journals of each field. How do we identify these journals?
Step 1: Identify high-impact journal in each field
To get at the most important journals in each field, we use a revealed preference approach: we analyse the journals in which the elected members published. More precisely, we take the subset of all AAAS and NAS members with a Google Scholar page, identify their 20 most cited publications, and tabulate the journals in this list. This provides us a list of 13–16 journals in each discipline. Reassuringly, all commonly described top journals in the disciplines are there.
Step 2: Download information on publishing authors
We download title and author information for all articles published in the relevant (13–16) journals in each discipline, from the inception of the journal until 2019. We eliminate non-regular articles (e.g. notes and letters to editors) and retain all articles since 1930 (to allow for a 30-year period between publishing and getting elected).
Step 3: Build dynamic CVs for all active researchers
For each active researcher, we link author information in the relevant journals in the field (using a combination of first, middle, and last name). A researcher enters the database the first year she publishes in one of the relevant journals; subsequently, information on publication and citations is added on a yearly basis. In the end, for a given author in a given discipline, we know for every year (since entering the database) how many publications she has in each of the journals, and how many cites each publication received. To measure research productivity in year t, we take the cumulative publications in each of the journals up to year t, and the cumulative citations to previously published articles in each of the relevant journals up to year t. To provide a concrete example: an entry would be David E. Card in the year 2007: five (cumulative) publications in the AER, five (cumulative) publications in Econometrica, 301 citations for the AER publications, and 410 citations for the Econometrica publications.
Step 4: Assign gender to the authors
First, we use a gender coding protocol developed in Card et al. (2020). Second, if the coding protocol was inconclusive, a team of research assistants manually looks up the author’s gender. Priority is given to more prominent authors.
Step 5: For all the active researchers, merge yearly information on membership in the AAAS and NAS
As such, we know for all years in which the researchers are active whether or not they became member of one of the academies.
Election prospects for women in psychology, economics, and mathematics
For each of the three disciplines, we estimate a logistic hazard model for the event of being selected as member of the AAAS and NAS. The risk set consists of all active researchers who were not already selected. We control for research quality with accumulated publications and citations, and are interested in the coefficient estimate of the gender dummy. Conditional on research quality (and year fixed effects), does the gender of a candidate predict election success?
As the importance of the journals may vary over time, we estimate the model for three different periods separately: 1960–1979, 1980–1999, and 2000–2019.
Figure 1 Share female of new members and active publishers
Notes. In this figure, we plot the share female among new members in AAAS and NAS, all active publishers, publishers with five or more publications, and publishers with 15 or more publications. The new members series excludes researchers that are not matched to the active publisher data set. AAAS and NAS shares are a moving average that include two leads, the index year, and four lags. In 2020 and 2021, we use, respectively, 1 and 0 leads. Female shares for publishers are normal averages.
Psychology is the discipline with the highest number of female researchers. The share of women among the active publishers was close to 20% in the 1960s, and has steadily increased to 50% as of today. Relative to the increase in the pipeline of female researchers in the field, how has the representation of women among elected members developed? Figure 1 (Figure 2 in Card et al. 2023) shows that initially, the share of newly elected females in the academies was lower than the share of female researchers (this is true for all active researchers as well as highly productive researchers with more than five publications). Over time, the share of female members rose faster than the share of female researchers. The graph therefore suggests that preferences for electing female members may have changed. The regressions confirm this conclusion. Before the 1980s, the coefficient of the female dummy was negative for both academies, albeit insignificantly. The negative coefficient turned positive afterwards, and significantly so for the last 20 years.
Economics has been a male-dominated field for a long time. As can be seen in Figure 1, the share of female economists remained well below 10% until at least the 1990s, and only recently increased to 20% as of today. In contrast to this relatively slow increase in the pipeline of female economists, there is a steep recent increase in the share of newly elected female members. The regressions confirm that for a male and female researcher with a comparable CV, election prospects for women have become much more positive in the last 20 years.
Mathematics is even more masculine than economics. The pipeline of female researchers has increased over time, but the level is still low. In the 1960s, the share of female researchers was around 5%, and it is now around 10%. Did the evolution of elected female members develop in parallel? Actually no (see Figure 1). While initially the share of newly elected female members lagged behind the share of women in the pipeline, this pattern changed, and strikingly so in the last ten years. The regression results point in the same direction: historically, excellent female researchers faced a higher bar to get elected, but over time, the academies’ election preferences have become more favourable towards female researchers.
We investigate for three different disciplines whether there are gender differences in research recognition, as measured by becoming a member of one of the most prestigious academies in the US: the AAAS and the NAS. Even though economics, psychology and mathematics are very different (also in terms of feminisation), trends in election patterns over time are surprisingly similar. Initially, qualified female researchers in all three disciplines seem to have been held to a higher bar. Starting in the 1980s, a positive preference for female members emerged, but the magnitudes were small and not always statistically significant. However, there is consistent evidence that over the last 20 years, female researchers with similar publications and citations than men are more likely to be selected. There are different mechanisms that could account for this positive effect. First, women may face various hurdles that make obtaining a given CV more difficult, and voters may take this into account. Second, academy members may try to redress the under-representation of women in the past by increasing the share of women among newly elected members. And last, the academies could be aiming to achieve a share of female members that reflects broadly the share of women among active researchers.
Card, D, S DellaVigna, P Funk and N Iriberri (2020), “Are Referees and Editors in Economics Gender Neutral?”, Quarterly Journal of Economics 135(1): 269–327.
Card, D, S DellaVigna, P Funk and N Iriberri (2023), “Gender Gaps at the Academies”, PNAS 120(4): e2212421120.
Ceci, S J, D K Ginther, S Kahn and W M Williams (2014), “Women in Academic Science: A Changing Landscape,” Psychological Science in the Public Interest 15(3): 75–141.
Ginther, D K and S Kahn (2004), “Women in Economics: Moving Up or Falling Off the Academic Career Ladder?”, Journal of Economic Perspectives 18(3): 193–214.
Hengel, E (2022), “Publishing while female”, Economic Journal 132(648): 2951–2991.
Lundberg, S and J Stearns (2019), “Women in Economics: Stalled Progress”, Journal of Economic Perspectives 33(1): 3–22.
Reuben, E, P Sapienza and L Zingales (2014), “How Stereotypes Impair Women’s Careers in Science”, Proceedings of the National Academy of Sciences 111(12): 4403–4408.
Sarsons, H, K Gërxhani, E Reuben and A Schram (2021), “Gender Differences in Recognition for Group Work”, Journal of Political Economy 129(1): 101–147.