Sexism is derailing mathematicians from an early age

A study of students in Israel by Victor Lavy and Edith Sand has discovered a surprising result…or maybe not so surprising to you, but I was rather shocked. Math teachers score girls' performance lower when they know their identities.

In math, the girls outscored the boys in the exam graded anonymously, but the boys outscored the girls when graded by teachers who knew their names. The effect was not the same for tests on other subjects, like English and Hebrew. The researchers concluded that in math and science, the teachers overestimated the boys’ abilities and underestimated the girls’, and that this had long-term effects on students’ attitudes toward the subjects.

For example, when the same students reached junior high and high school, the economists analyzed their performance on national exams. The boys who had been encouraged when they were younger performed significantly better.

They also tracked the advanced math and science courses that students chose to take in high school. After controlling for other factors that might affect their choices, they concluded that the girls who had been discouraged by their elementary schoolteachers were much less likely than the boys to take advanced courses.

But…math. Isn't that one of those incredibly objective disciplines in which questions all have a right answer and a best method, and there's no wiggle room for adjusting a score? Just like all of science -- there's no subjectivity here at all.

No, when you're evaluating how well students think, there's always lots of room for taking student knowledge into account. I teach genetics, and it's a good example: I grade exams with my nice brief key by my side, and when students come up with the same answers I do, it's easy and fast. But when they don't, I have to look much more closely. Did they just make an arithmetic error in the last step? Did they understand the basic concepts, but just fail to integrate them all? Did they demonstrate a complete lack of comprehension of basic Mendelian principles? I have to see some sign of understanding in the work to make the effort to track through the problem more carefully, and it would be tempting to, for instance, know that this student did poorly on their last exam, so it's not worth the effort to try and figure out what dumb mistake they made this time.

(I take steps to avoid that trap: I grade papers anonymously, not looking at the name on the first page.)

But I have a hard time imagining taking a negative attitude towards a math problem on the basis of the solver's sex. Apparently it's common enough that it actively skews assessments downward, though.

I'm familiar with the Swedish study that showed a pervasive bias against women scientists on the job market, but it's clear the problem goes much deeper: women are being discouraged from going into math as early as middle school.

The paper also tried to puzzle out what was going on with these teachers, and found some other interesting correlations.

Older and single teachers seem to favor boys over girls: the coefficient of a dummy indicator of being older than 50 years old is positive and significant (0.206, SE=0.104), and so is the estimate of the indicator for single teachers (0.315, SE=0.202). The estimated coefficient for teachers from Europe-North America origin is negatively and significantly correlated with teachers’ biases (-0.204, SE=0.113). The other individual characteristics that we examined are being married (positive but insignificant) and the number of children and the proportion of daughters, both of which have negative coefficient but not significantly different from zero.

So older teachers are more biased in favor of boys; there's hope that that effect will diminish as a newer generation of teachers takes over. I don't think we can insist that teachers get married.

I also wondered about the effect of the teacher's sex on this problem. Buried deep in the paper is an interesting revelation: they couldn't look at that because all of the teachers in their sample were women. We learn two things from that, of course: that women can propagate sexist attitudes (no surprise), and that teaching is a deeply gendered profession. The gender distribution in the teaching profession also has to be sending a message to girls and boys.

Here's the authors' conclusion.

We also find that favoritism of boys among math and science teachers has an especially large and positive effect on boys math test score and on their successfully completion of advance math and science studies in high school; the respective effect on girls is negative and statistically significant. The estimates of the direct-subject effect in math are of special interest because of the considerable gender gap in math achievements and its impact on future labor market outcomes. Moreover, since this gap in math achievement partly results from teachers' stereotypical biases against girls in mathematics, eliminating these biases will go a long way toward reducing the math achievements gender gap, and it will also decrease the gender gap in enrollment in advanced math studies. The impact on the various end of high school matriculation outcomes carries meaningful economic consequences because these high stakes outcomes affect sharply the quantity and quality of postsecondary schooling and impact earnings at adulthood as well.

Another message we should take away from this: teaching is important. All you primary school teachers out there are shaping society as a whole.