Theoretic background:
Extensive research has shown that Science, Technology, Engineering, and Mathematics (STEM) disciplines are widely perceived as male-dominated fields. This persistent stereotype significantly contributes to the gender gap in mathematics at school, the underrepresentation of women in STEM majors at university, and the gender-based segmentation of the labor market.

Aim:
This paper seeks to deepen the understanding of factors encouraging the selection of STEM at University and their difference between boys and girls, in the subpopulation of students graduating from Scientific Lyceum in Italy. Specifically, the research addresses the following questions:
- To what extent does mathematical ability predict the choice of a STEM at university?
- Does mathematical ability influence male and female students differently in their decision to enrol into STEM studies?
- Can a higher cultural level in the family counteract gender stereotypes regarding academic choices?

Data:
We made use of a longitudinal population dataset obtained by joining together three official administrative/research sources of educational data: Ministry of University and Research , Ministry of Education and Italian National Institute for the Evaluation of the Educational System (INVALSI). Students enrolled for the first time at any public or private university in Italy in the academic year 2019/2020 were retrospectively matched with their data in the year 2018/2019 when they attended last year of high school (scientific lyceum), both in terms of proficiency and in terms of contextual characteristics.
The key variables in the final dataset include the macro-area of the university course (STEM vs. non-STEM), INVALSI math test scores, teacher-assigned math grades (as measures of proficiency), and the mother’s highest educational attainment (as a proxy for family cultural background). Additional contextual factors include geographical location, school type, and individual characteristics such as migratory background and history of grade repetition.

Analyses:
We classified the macro-area of discipline of the university course selected at enrolment into STEM vs. non-STEM following the official classification provided by the Ministry of University and Research. This was our outcome variable for two logistic regression models, stratified by gender, designed to predict choice of a STEM course at university, with specific attention to the role of math skills and socio-cultural factors at familial level. We compared the results in terms of ODDS ratios of the variables and we observed the different profile of determinants and the variation of magnitude for boys and girls.

Results:
Key findings highlight the central role of mathematical proficiency, especially as reflected in teacher-assigned grades, in influencing the decision to undertake STEM studies. This effect is notably stronger for male students.
Interestingly, maternal education exerts contrasting effects based on gender: while higher maternal education increases the likelihood of STEM enrolment for girls, it appears to reduce the same likelihood for boys. This may suggest that a higher family cultural background mitigates gender stereotypes in both directions: challenging both the stereotype that girls are less suited for STEM and the stereotype that boys should avoid fields associated with education, caregiving, and the humanities.
These findings should be further discussed to confirm their validity.

Keywords: STEM, gender, University, math skills, population, longitudinal study.