Although the educational attainment gap between boys and girls has been reduced in the last years, boys still outperform girls in mathematics starting from the end of primary school. This has lasting consequences for the availability of skilled workers in technical and innovative _elds because mathematic skills influence the choice of majors. Moreover, from the individual point of vue, mathematic skills yield high returns on the labour market. Although several possible determinants of the difference in mathematical performance between boys and girls have been investigated in the US literature, the test score gap in mathematics has remained largely unexplained as yet. Nevertheless there are reasons to believe that cultural factors play a role in the development of the test score gap, and that it cannot entirely be due to genetic differences in ability. Firstly, because the mathematics test score gap is non significant in kindergarten and develops during the primary school years, and because the gap seems to occur only in certain subgroups of the population. This paper contributes to the literature by providing new evidence on the sources of the gender test score gap in mathematics using detailed information on teachers and pupils from the complementary national PISA survey in Germany. Firstly, we investigate whether the share of female mathematics teachers affects the mathematics test score gap. The idea is that a higher share of female mathematics teachers may provide a positive role model for female pupils. In addition, female mathematics teachers may use didactical methods that are more adapted to girls and they could have less prejudice towards girls performance in mathematics than male teachers. In a second step, detailed information on parents and pupils perception of mathematics allows us to investigate the role of self-confidence and extrinsic motivation in the appearance of the test score gap in mathematics. Although girls on average face more favorable learning conditions with respect to the number of books at home, the type of secondary school, the time spent in kindergarten and even parental expectations, we find that they have less confidence in their math skills than boys. Moreover, boys believe more often than girls that they will need mathematics skills in their future jobs We do not find a significant effect of the share of female mathematics teachers in secondary school on the gender test score gap in mathematics. Based on these results, it seems ineffective to increase the share of female mathematics teachers to reduce the test score gap. To the contrary, the gender test score gap is reduced by more than half when parents have more than 500 books at home. Female pupils' lower extrinsic motivation and self-confidence in mathematics also seem to account for part of the test score gap. In order to go further in the exploration of the reasons behind the development of beliefs about differences in talent in mathematics between boys and girls, we believe future research should be done in cooperation with psychologists.

Keywords

mathematics test score gap, gender, share of female teachers, self-condence