Are women less persistent? Evidence from submissions to a nationwide meeting of economics

ABSTRACT

Female under-representation in high-profile career positions has relevant impacts on firms’ outcomes, research topics, and public policies. In the academic profession, women’s participation decreases as they evolve in their careers. To understand the lack of women in economics in Brazilian academia, we investigate the decision to submit papers to the largest conference in the country (Brazilian Meeting of Economics, or ANPEC Meetings), an important achievement in the profession. We explore a novel panel dataset of researchers and match them with web-scraped data of their résumés to test gender differences in the probability of submitting an article one year after having a paper (same or new) rejected in the previous year. Our findings suggest that women desist 2.9% points more than men when facing rejection. We also find evidence that younger women give up more and that the quality of the undergraduate program relates to the gender gap in the likelihood of desisting. Finally, we argue that more competitive women may self-select into higher-quality institutions.

KEYWORDS:

• Female underrepresentation
• competitive behaviour
• academic conferences

JEL CLASSIFICATION:

• J15
• J16
• C23
• A11

Acknowledgement

We thank Roberto Meurer and Eliane Rêgo for kindly providing the administrative data of ANPEC submissions. We also thank the Department of Economics at the University of São Paulo, FIPE and FAPESP (grant no. 19/16952-7) for their financial support. We are grateful to all participants in the regular seminars of the Brazilian Women in Economics (EconomistAs), the Family and Gender Economics Study Group (GeFam) and the Regional and Urban Economics Lab at the University of Sao Paulo (NEREUS) for their contributions. Finally, we thank Fabiana Pereira, Luiza Karpavicius, Pedro Feijo, and Victoria Klarosk for their excellent research assistance. The remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ These statistics are provided by the Brazilian Women in Economics (BWE) research group, that conducts an annual survey focused on gender differences in the academic career.

² In the United States, the CSWEP Annual Report collects data from 250 U.S. departments since 1972. It is published in one of the volumes of the American Economic Review: Papers & Proceedings (Lundberg 2017).

³ The Lattes CV is a nationwide platform that contains the Brazilian professors’ and researchers’ academic curricula.

⁴ The quality of the institution is measured by a national evaluation in Brazil (CAPES evaluation), which we will detail in the following sections.

⁵ See Niederle and Vesterlund (2011) for a complete survey, and Flory, Leibbrandt, and List (2015), and Flory et al. (2018) for studies showing that women prefer to cooperate than to compete in the United States and Malawi, respectively.

⁶ Booth et al. (2019), in turn, carry out an experiment with individuals from different birth cohorts in China to select individuals who grew up with different social norms and different political and economic regimes (the communist era, especially during the Cultural Revolution, and after the economic reform and establishment of a market economy). They emphasize the impact of propaganda and indoctrination in gender equality during the communist period, in addition to the greater appreciation of the work of all, on women’s behaviour.

⁷ The authors also estimate the likelihood of girls who did not make it to the second round at the Dutch maths Olympics to try again next year and show that it decreases significantly when compared to men under the same conditions.

⁸ Valentova et al. (2017) find that, in 2013 and 2014, women scientists were more often represented among holders of the lowest productivity levels (level 2), while male scientists were most often found at higher levels (1A and 1B) in “Engineering, Exact Sciences, Earth Sciences” and “Life Sciences”. Although they did not find this imbalance in humanities and social sciences, the behaviour was similar to that in the exact sciences in economics. Of the 207 scholarships awarded in Economics by CNPq (National Council for Scientific and Technological Development), only 29 were to women. Lundberg (2017) shows that in economics departments in the U.S., the higher the career position, the lower the percentage of women. Tenreyro (2017) reports a similar share of women in economics departments in the U.K. and CWEN (2017) in Canada.

⁹ Chari and Goldsmith-Pinkham (2017) analyzes the representation of women economists in the programs of the NBER Summer Institute in the period 2001–2016, an annual conference – highly competitive – promoted by economists affiliated with the NBER (National Bureau of Economic Research). Hospido and Sanz (2021) investigate the gender difference in articles accepted for presentation at the European Economic Association Annual Congress (2015–2017), the Spanish Economic Association Annual Meeting (2012–2017), and the Spring Meeting of Young Economists (2018). The authors conclude that female-authored papers were 3.3% points (or 6.8%) less likely to be accepted than male-authored papers.

¹⁰ The share of women in finance and macro and international fields is much lower than in applied microeconomics.

¹¹ In Portuguese “Associação Nacional dos Centros de Pós-graduação em Economia – ANPEC”.

¹² In Portuguese “Sociedade Brasileira de Econometria – SBE”.

¹³ Until 2012 submissions could be made in 12 different areas. In 2013 Political Economy became a separate area. In order to make the comparison over time possible we follow the old classification.

¹⁴ Appendix Table A1 presents the number of article submissions by authors that appear only once in the database by year. There were few observations in the years 2007 and 2008, and, therefore, we excluded them from our analysis.

¹⁵ As a double-check, we also perform the same procedure using data from the Annual List of Social Information (RAIS) submitted by companies to the Brazilian government. The adherence in the gender classification of RAIS and TSE is 99.2%.

¹⁶ We use the names of all candidates for the elections between 2008 and 2016. When we could not find the name in the TSE data, or the probability of being a woman (man) was less than 90%, we manually assigned the gender using internet searches of the authors.

¹⁷ Abrevaya and Hamermesh (2012), Hoekstra and Street (2021) and Card et al. (2020) use similar procedures to identify gender.

¹⁸ Lattes CV serves as an indicator of the participation of individuals in the Brazilian academic market.

¹⁹ Table 2 compares statistics for the full sample and the Lattes CV sample (matched with the Lattes database). Table 3 displays the female representation in the areas in the complete sample and the one used for the estimates (of the individuals in which it was possible to make the exact matching with the Lattes database). The samples used in our estimates are representative of the total sample, as well as the representation of women in each area.

²⁰ Most of our variables refer to the first year of our sample (2009) or are constructed for the panel (e.g. number of published articles per year).

²¹ There are authors who submitted more than one article per year. Hence, we created variables for the number of articles each author submitted per year and the number of accepted articles.

²² We use as control variables the accumulated number of publications, conference participation and advisories.

²³ As a robustness test, we estimate Equation (1)(1) $\begin{array}{rl}& submissio{n}_{it}={\beta }_0+{\beta }_{1}rejectio{n}_{it-1}+{\beta }_{2}rejectio{n}_{it-1}\times \\ & woma{n}_i+{\beta }_{3}acceptanc{e}_{it-1}+{\beta }_{4}acceptanc{e}_{it-1}\\ & \times woma{n}_i+X_{it}\theta +{\alpha }_i+{\delta }_t+{\text{isin}}_{it}\end{array}$(1) including rejection and acceptation in other lagged periods (t-2, t-3 and t-4). In Appendix Tables A2 and A3, we see that rejection in all previous periods decrease the probability of submissions. Also, the higher effect of rejection on submissions for women is only statistically significant considering rejections in t-1.

²⁴ We also estimate a variation of Equation (1)(1) $\begin{array}{rl}& submissio{n}_{it}={\beta }_0+{\beta }_{1}rejectio{n}_{it-1}+{\beta }_{2}rejectio{n}_{it-1}\times \\ & woma{n}_i+{\beta }_{3}acceptanc{e}_{it-1}+{\beta }_{4}acceptanc{e}_{it-1}\\ & \times woma{n}_i+X_{it}\theta +{\alpha }_i+{\delta }_t+{\text{isin}}_{it}\end{array}$(1) but considering the submission in t-1 as covariate. Results are displayed in Appendix Table A4. We observe that the probability of submitting an article in the subsequent period is always positive but higher for men than for women. The results from Table 4 show that the paper rejection drives the gender differences in the previous period: the acceptance in t-1 does not influence the probability of submitting in t, while the rejection does impact.

²⁵ For individuals who submitted more than one article per year in different areas, we consider the submission area as the mode of the research areas in the year. If there is more than one mode, we use the smallest mode.

²⁶ The fields with the highest female representation among referees are Industrial Economics and Technology (37.5% of female referees), and Applied Microeconomics (28% female referees), where we do not observe a gender gap in the likelihood of desisting. Meanwhile, Public Economics, Growth and Development, and Microeconomics & Quantitative Methods present low percentages of female referees (4%), and for these research fields, we do not find a statistically significant gender gap.

Additional information

Funding

The work was supported by the São Paulo Research Foundation (FAPESP) [2019/16952-7].