I read the study that was the basis of the link Dude provided. I found it less than a convincing debunking of higher male variance in mathematical ability. They actually go so far as to plot a histogram of international gender Variance Ratios (VR) that peaks at 1.16, with an inter-country variance of only 0.0054 (I calculated myself from the data. The authors claim a "large" variance without actually stating it.), and go on to argue that the natural variance ratio is 1.0, and the explanation for anything else is bias. The amusing thing is that they implicate bias in exaggerating the natural variance ratio in just about every country tested, and refuse to accept that bias could possibly diminish this ratio in the handful of countries that show a smaller ratio (which could also be caused by sampling error). Of course, they also discount the idea that different countries with different populations could have different biological variance ratios. In effect, they assume that many groups of humans are genetically similar in order to (attempt to) disprove that 2 groups of humans are genetically different.
If you calculated your inter-country variance based on the information from Table 2, your data set is incomplete. The purpose of Table 2 is to demonstrate whether consistency exists among different tests given at different times within the same country. If a country did not participate in enough different testing cycles to shed any light on this, its data was left out of Table 2.
When the authors then comment on the "large variance" you're taking issue with, they do so like this (with figure 1A being the histogram in question):
In fact, the VRs calculated using the 2007 TIMSS
eighth-grade data set studied in detail here varied
widely among countries, ranging all the way from 0.91 to 1.52 (Figure 1A).
We can look at the 2007 TIMSS 8th-grade column on Table 2 and find Tunisia on the low end with 0.91. We cannot find a high of 1.52... the highest figure provided is 1.31 for Taiwan. The study states that 8th graders in 52 countries participated in the 2007 TIMSS, yet exactly half of those results are reported in Table 2.
I'd agree with the statement that 0.91 to 1.31 is a wide variance, and 0.91 to 1.52 even more so.
Additionally, the study was based on knowledge tests rather than ability tests like IQ
Since the study is out to determine how mathematical performance differs among the sexes amid different cultures, that was the right choice.
Some more interesting data which came from other studies and were mentioned in the intro to this one:
This gender-stratified
hypothesis is consistent with several recent findings.
For example, Hyde and collaborators ([20],
[25]) reported that girls have now reached parity
with boys in mean mathematics performance in
the United States, even in high school, where a
significant gap in mean performance existed in the
1970s. Likewise, both Brody and Mills ([3]) and Wai
et al. ([51]) noted a drop in nonrandom samples of
students under thirteen years of age, from 13:1 in
the 1970s down to approximately 3:1 by the 1990s
in the ratio of U.S. boys to girls scoring above 700
on the quantitative section of the college-entrance
SAT examination. The percentage of Ph.D.’s in the
mathematical sciences awarded to U.S. citizens
who are women has increased from 6 percent in
the 1960s to 30 percent in the past decade ([4],
[9]). Sociocultural, legal, and educational changes
that took place during this time span may account
for these dramatic improvements in mathematics
performance and participation by U.S. females.
So, what biological revolution occurred in the US between the sexes from the 70s to today that can explain these results? Answer: none. What social revolution can explain them? Answer: feminism.
