Evergreen I

### A New Theoretical and Methodological Look at Gender Differences in Math Tests: Reasoning, Psychosocial Attitudes, and Multilevel Models (306505)

**Eiliana Montero-Rojas, University of Costa Rica*

*Tania Elena Moreira-Mora, Costa Rica Technological Institute*

*Vanessa Smith-Castro, University of Costa Rica*

*José Andrey Zamora-Araya, National University of Costa Rica*

**Keywords:** ambivalent sexism, gender differences, gender stereotypes, Math performance, multilevel models, reasoning abilities

Differences favoring males were explained for two standardized tests of mathematical context: the Math section of the University of Costa Rica’s Admission Test and the Math test from the secondary school’s exit exam. The sample was 487 students from ten public schools in the big metropolitan area of Costa Rica, 269 were women. Multilevel regression models were estimated using the scores in these tests as dependent variables. The reduced model only included sex as predictor, whereas the complete model included, additionally, an indicator of reasoning abilities, and the following scales: hostile sexism, benevolent sexism, attitude toward gender equity in Math, and, self-efficacy in Math. Results point out that gender differences are reduced considerably by controlling for reasoning abilities, perhaps due to a self-selection process by males. Benevolent sexism predicted performance in the tests in a lesser degree. There was also evidence of the moderating effect of the school for the relationship between sex and the score in the admission test, and, especially, for the means in the secondary school exit test. Math self-efficacy had also a considerable predictive power in the latter.