Impact of Female Peer Composition on Gender Norm Perceptions in Secondary School in Uruguay

Abstract

Gender norms affect girls’ educational achievements. Yet, how norms are formed and what drives their changes are still open questions. This article examines peer effects on students’ gender norm perceptions and skills formation. The study uses a Uruguayan nationally representative survey of ninth grade students and exploits the quasi-random variation in the proportion of female peers across classes within schools for causal identification. Results show that a higher exposure to female peers leads to more progressive gender norms. Furthermore, these effects are driven mostly by male students. Female students are also impacted by peers’ gender composition and significantly reduce the time devoted to domestic work and improve mathematics scores when exposed to more female peers. No effects on language test were found for either gender. These results suggest that short peer interactions in secondary school contexts may have substantial effects on reducing gender stereotypes and change gendered behaviors among students.

HIGHLIGHTS

• In Uruguay, higher exposure to female peers in the class leads to more progressive gender perceptions by male students.

• Class sex composition also reduces time devoted to domestic work and improves mathematics performance of female students.

• Even one-year peer interactions in secondary school may have substantial effects on increasing gender-egalitarian attitudes and behaviors.

KEYWORDS:

• Gender norms
• gender inequalities
• peer effects
• developing countries

JEL Codes:

• I24
• J16
• Z13

ACKNOWLEDGMENTS

I want to thank María Inés Berniell, Marisa Bucheli, Cecilia Parada, Rodrigo Oliviera, and the Feminist Economics associate editor, editor, and three anonymous referees for their substantial comments. I am also grateful to all participants at the UNU-WIDER development conference 2022, 26th Annual LACEA Meeting 2021, 12th NIP seminar (Uruguay) 2021, 13th International Scientific Conference 2021 “Applied Economics Conference: Labour, Health, Education and Welfare,” and the Australian Gender Economics Workshop 2022 for their excellent technical suggestions and useful comments. I also acknowledge UNU-WIDER's support through the PhD fellowship program.

Notes

1 See Izaguirre and Di Capua (2020) for many countries in Latin America, McEwan (2003) for Chile, and de Melo (2014) and Alan, Ertac, and Mumcu (2018) for Uruguay.

2 For instance, PISA surveys only provide information for a sample of students within a classroom, thus challenging the identification of peer effects.

3 There is 52.6 percent agreement with the statement that “Being a housewife is just as fulfilling as working for pay” (Galván et al. 2023).

4 Students entering technical schools must choose a field of study based on their interests (arts, manual works, logistics, robotics, among many others). This paper will focus on regular schools only, as the sex composition of technical classes could be mediated by other gendered preferences.

5 Fourteen percent of students in the estimation sample attend extended time schools (eight hours a day), and the remaining 86 percent regular time schools with morning and afternoon groups.

6 A similar process of class formation is identified by Ammermueller and Pischke (2009) in 94 percent of the schools in their sample of European primary schools.

7 The database is publicly available https://www.ineed.edu.uy/aristas-2018-tercero-de-educacion-media/https://www.ineed.edu.uy/aristas-2018-tercero-de-educacion-media/.

8 The sampling frame is made up of 523 schools and 48,372 students. This includes all public and private urban schools in the country with at least one ninth grade student.

9 I exclude students in technical schools (23.1 percent), without information on age (4.7 percent), in classes in which the number of students is lower than the first percentile of the number of students per class distribution (0.6 percent), without information on control variables (6.7 percent), and in schools with only one classroom (4.5 percent).

10 All estimations are computed using survey weights calculated by INEEd.

11 The household wealth index provided by INEEd that considers information on household composition, health attention, dwelling characteristics, and comfort; parents' educational level; lecture habits in the family; educational expectation; and parental supervision, among others.

12 Differences are statistically significant at conventional levels, with the exceptions of the proportion of only child and live with both parents.

13 The same procedure was followed for the other peer variables.

14 Cronbach Alpha Coefficient is 0.8065 for the original variables, and 0.7778 for the standardize variables. In both cases, the reliability statistic is well above the threshold suggesting internal consistency (0.7).

15 Note that while X includes student own gender, the vector of peer controls does not include peers’ gender composition as that is the treatment variable denoted as ShareFem. An alternative specification could also include peers’ gender norms perception to account for other potential channel of peer effects. Results hold significant for all items except for Wages which loses significance. Estimations available upon request.

16 Previous studies also rely on school fixed effects for causal identification. See McEwan (2003) for Chile, Vigdor and Nechyba (2008) for North Carolina, and Ammermueller and Pischke (2009) for six European countries (France, Germany, Iceland, the Netherlands, Norway, and Sweden).

17 A challenge also faced by previous studies (Ammermueller and Pischke 2009).

18 This gauge is similar to Lavy and Schlosser (2011), who find that 89 percent of the schools had a standard deviation that fell within the empirical 90 percent confidence interval.

19 For own gender regression, I also include the share of female students in the school to control for the inherently negative relationship between the share of female peers and own gender for women students (Guryan, Kroft, and Notowidigdo 2009).

20 Results of estimating the same regression including a quadratic term point to a nonlinear relationship between female peers and gender norms. The estimated coefficient of the square is negative and significant meaning that the relationship is positive but decreases with the proportion of female peers. Unfortunately, the structure of the dataset does not allow to explore nonlinearities further.

21 This exercise follows the methodological approach proposed by Vigdor and Nechyba (2008).

22 Even though gender gaps in mathematics scores are negligible according to Aristas’ data, female students are relatively better in language and male students in mathematics.

Additional information

Notes on contributors

Martina Querejeta

Martina Querejeta is a PhD student in economics at the National University of La Plata, Argentina. She completed a Bachelor’s degree and a Master’s degree in Economics at the University of Republic, Uruguay. She has previously worked as a research assistant in several projects regarding gender and development issues, and has also experience in the implementation of public policies in Uruguay working in the Ministry of Social Development. She is a member of the Interdisciplinary Center for Development Studies. Her key research interests include social policies, labor economics, and gender.