“Every year I teach dozens of students at the University of Birmingham. Most of the students on the gender and sexuality courses are women. I guess this is because the boys don’t think that gender applies to them: that it’s a subject for girls.”
-Louise Brown
You know the stereotype, perpetrated throughout the United States (and well beyond) for generations: girls aren’t as good at math as boys are. For a long time, people pointed towards the long list of (almost exclusively male) mathematicians and scientists as support for this idea.
Never mind the fact that women had been disenfranchised from these careers for centuries. From Barbie dolls to Harvard presidents, the gender disparity in mathematics and the sciences is often attributed to a hypothesized difference in intrinsic aptitude, even today.
Over the past generation, however, standardized tests in the United States have seen that gender gap completely disappear. First among elementary and middle schoolers, then among high schoolers, and today, male and female students achieve identical average math scores on the SATs.
Despite the fact that there are known social, institutional and economic gender inequities, even while the gender disparity has been progressively disappearing for older girls and boys over time, there is still a marked gender inequality at the highest career levels. Regardless of how much the inequality has lessened over time, the idea that this is somehow due to an inherent female inferiority persists. There are still people, even today, who steadfastly believe that there are more male mathematicians and physics professors (among other fields) not because women are being treated and judged differently or unfairly, but because men are naturally superior to women at this.
(Don’t think that’s true? Go read the comments on my last article on gender in science, from a mere eight months ago.)
But you know how prejudices and confirmation biases work: if you think things are a certain way for a certain reason, then when your reasoning is shown to be incorrect because your premise is flawed, what do you do? Do you question your conclusions, or do you just find a new explanation that brings you to that same conclusion? Most recently, the argument goes something like, “even though men and women are equal on average in math ability, men have a greater variance in their abilities. So there are more very dumb men, but also more very smart men, and those are the ones who become scientists, etc.” And then a statistic like this gets thrown up.
“A-HA,” you hear, “69% of the perfect SAT Math scores were achieved by boys, more than twice as many as the girls!” Never mind that in the 1970s, the disparity was 93%-7%, and that girls’ performances have been steadily rising. Clearly, girls just can’t possibly be as inherently gifted or talented at math as boys, and that’s why there are more male physics and math professors. (And feel free to apply this flawless logic to whatever field you feel like.)
Now, I can tell you all about my personal experience at all the different levels of education. From the elite math students in elementary and middle school to the best math students in the nation at the high school level, to math and physics majors at elite universities, at the undergraduate and graduate levels, onto the faculties at those same colleges and universities, I have lots of stories about achievement, adversity, and gender-specific obstacles that only women face. But this isn’t that story of the anecdotes I’ve accumulated over a lifetime in that environment; this is about something far larger than my accumulated experience. This is science!
Last month, a new study came out, spanning 31 countries and measuring the mathematical performance of well over 100,000 students, on the standardized TIMSS exam. Here are the results.
(This and all subsequent images from Kane & Mertz, 2011. If you don’t like fooling around with someone else’s interpretation, I’ve uploaded the full paper, by Jonathan M. Kane and Janet E. Mertz, here.)
As you can see, between each country, there’s practically no substantial difference between the mean scores of male/female students. In some countries, boys do better on average, in some countries girls do better on average, but across all countries, there’s no statistically significant difference in the average scores of girls and boys.
But you’ll also notice there’s something called the “Gender Equity Index” on the x-axis. What’s that? It’s a weighted measure of the gap between men and women when it comes to things like economic equality, education, and political/economic empowerment. Sweden is highest, with a score of 89, while Yemen is lowest, with a score of 31. (The U.S. comes in 24th, at 74.) 100 would be actual gender equality; no country in the world has it yet.
The gender gap index is almost the same, but also includes health and survival (the U.S. does slightly worse here), and is on a scale from 0-1 instead of 0-100. As you may well have guessed, countries that have higher levels of equality do, in general, see things like greater female representation on the International Mathematical Olympiad teams.
But the hypothesis, remember, was that males have a greater variance than females when it comes to math ability, and that’s why there are so few females at the high end of the career spectrum in math and science. Is this true, that males have a greater variance than females in their math ability, and if so, is the variance significant enough that it could account for this gender disparity? Well, let’s see what the data has to say about that.
Well, there are some countries, like the Czech Republic, that show absolutely no difference between boys (blue) and girls (red), not in average and not in variance. So if males do have an inherent variance greater than females, it isn’t exhibited everywhere.
There are some countries, like Bahrain, where boys do in fact have a larger variance in their math achievement than girls do, but also have a lower average. In the case of Bahrain, the variance only makes up the boys’ failings at the high end; there are the same percent of girls achieving top scores as there are boys at the high end. So even where boys do show a greater variance in their scores, it doesn’t always translate into greater achievement at the high end.
There are even a few countries, like Tunisia, where the inverse of this hypothesis is true! In Tunisia girls’ math scores are lower on average than the boys, but exhibit a larger variance! Here, too, at the high end, there are more girls than boys. If there is something going on where boys have an intrinsically higher variance than girls, and this results in more top male achievers at the highest scores, we aren’t seeing it across the board.
But the point was not to hand-pick countries; I could have just as easily chosen individual countries where the data supports the stated hypothesis. The point is to look at all the available data, complete with gender inequities and all, and draw the best conclusions we can based on that. So, what are the overall findings?
Overall, although there are many countries where there is virtually no difference in variance between boys and girls, boys actually do show a slightly greater (by a few percent) variance than girls in performance. (For example, the U.S. shows an 8% greater variance for boys.) But these differences in variance between countries are much greater than the difference in variance between boys and girls in any individual country! The question, then, becomes whether that difference in variance could explain the gender gap in math and science among men and women?
The authors tried to control for many different factors, and did their best to separate out what effects on test scores these different factors could possibly have. What did they find? In their own words,
None of our findings suggest that an innate biological difference between the sexes is the primary reason for a gender gap in math performance at any level. Rather, these major international studies strongly suggest that the math-gender gap, where it occurs, is due to sociocultural factors that differ among countries, and that these factors can be changed.
In fact, if there’s one thing that really drives this point home, it’s the graph of achievement in each country, regardless of gender, as a function of gender equity.
Wow! The percent of students scoring above 400 (low) and above 550 (high) rise dramatically, among both genders, when there’s greater equity among men and women! In other words, every step forward that a country takes towards eliminating the gender disparity in the economic, political, and educational realms leads to greater math achievement for both genders.
But I’ll give you the conclusions of the authors themselves:
In summary, we conclude that gender equity and other sociocultural factors, not national income, school type, or religion per se, are the primary determinants of mathematics performance at all levels for both boys and girls. Our findings are consistent with the gender stratified hypothesis, but not with the greater male variability, gap due to inequity, single-gender classroom, or Muslim culture hypotheses. At the individual level, this conclusion suggests that well-educated women who earn a good income are much better positioned than are poorly educated women who earn little or no money to ensure that the educational needs of their children of either gender with regard to learning mathematics are well met.
io9 also has a great writeup of this story, for those of you who want some further reading. I hope that if there’s one thing you take away from this, it’s the lesson of this last graph: the closer we get to gender equality, the more everyone benefits. Go read the paper yourself, and convince yourself that there are demonstrably far more significant factors than gender in determining math ability; the data is all in there, along with other “inherent-gender-ability” hypotheses that are also discredited. It’s time to put this sexist hypothesis for the achievement gap where it belongs, buried in shame in our past.
Because we all benefit from equality, and now you’ve seen the science that proves it.


