In this paper, we estimate the effect of primary school teachers’ gender biases on boys’ and girls’ academic achievements during middle and high school and on the choice of advanced level courses in math and sciences during high school. For identification, we rely on the random assignments of teachers and students to classes in primary schools. Our results suggest that teachers’ biases favoring boys have an asymmetric effect by gender— positive effect on boys’ achievements and negative effect on girls’. Such gender biases also impact students’ enrollment in advanced level math courses in high school—boys positively and girls negatively. These results suggest that teachers’ biased behavior at early stage of schooling have long run implications for occupational choices and earnings at adulthood, because enrollment in advanced courses in math and science in high school is a prerequisite for post-secondary schooling in engineering, computer science and so on. This impact is heterogeneous, being larger for children from families where the father is more educated than the mother and larger on girls from low socioeconomic background.