Arnold Kling  

Boys, Girls, and Math

PRINT
The McCain-Stiglitz Axis... Low Collective Benefits...

A study by Janet Hyde, Sara Lindberg, Amy Ellis and Carolyn Williams is getting a lot of publicity (two non-academic friends brought it up in conversation lastnight). They find that boys and girls do not differ in their performance on math tests. Unfortunately, I cannot find the actual study, so the link goes to a news story.


Hyde and colleagues sifted through math scores from 7 million students from 10 states tested in accordance with the No Child Left Behind Act, as well as the Scholastic Aptitude Test or SAT, a standardized test used for college admissions.

"Contrary to widely held stereotypes by parents and teachers that boys are better at math, our data ... showed that girls have reached parity with boys on math performance," Hyde said in a telephone interview.

What about at the very high end of the scale?

"While we did find more boys than girls above the 99th percentile at a 2-to-1 ratio, still, 33 percent of those kids who are above the 99th percentile are girls," she said.

Hyde's explanation for the fact that boys score higher on average on the math SAT's is that more girls take the SAT, so that the test includes more girls who are lower in the ability distribution.

My view would be the reason that more girls take the SAT's is that there are more girls in the middle of the ability distribution to begin with. My casual reading of the literature and observation in the classroom is that boys tend to have more variance--more at the top, and more at the bottom. That is why, as college admissions have grown, the big increase has been among girls.

The idea that more girls just randomly decide to go to college strikes me as unlikely. I would tend to think that, holding ability equal, a boy would be about as likely to want to go to college as a girl. However, I am just speculating. I am not a professor of psychology.

UPDATE: This article, which ranks disciplines according to "political correctness" would suggest that the difference between my views and those of psychology professor are predictable. Thanks to Tyler Cowen for the pointer.


Comments and Sharing





COMMENTS (9 to date)
jb writes:

I think you're spot on. The median boy and girl may be equivalent at math, but more variance smushes out the bell curve and that's why there are more boys at the top of the range,and (most likely) more boys at the bottom of the range as well.

Did she reference the bottom of the range in her report?

burger flipper writes:

I thought that this was the basic thrust of the comments that got Summers booted. That the variance was higher with men, so you'd expect almost exclusively men 3 or 4 standard deviations from the mean--right where elite institutions claim to draw from.

steve hsu writes:

The PISA data doesn't have the SAT selection bias.

PISA math scores (international testing of hundreds of thousands of 15 year olds in dozens of countries; OECD averages below). M and F standard deviations (SD) show statistically significant differences (error in SD is about 1-2 points), leading to different populations in the tail.

M mean 503 SD 96
F mean 492 SD 90

M/F ratio at
1 in 1000 ability level (+3 SD): 5/1

This is what Larry Summers was talking about. To see the raw data follow this link: http://infoproc.blogspot.com/2008/06/asian-white-iq-variance-from-pisa.html

Effect already measurable at the *beginning* of kindergarten:
http://infoproc.blogspot.com/2007/11/gender-differences-in-extreme.html

callejohan writes:

The issue of boy/girl differences in math ability have been studied at the Center for Talented Youth (CTY) at Johns Hopkins. Since the results might be seen as controversial, they have not been widely publicized.

They identified students of both sexes with high math performance, and then tested the parents too. They found that for boys to do very well in math, it was sufficient that one of the parents was math gifted. But for girls, both parents had to be high math performers. Since this indicated a difference that would be consistent with genetic causes, the studies have been kept rather quiet.

But it would obviously explain why the upper tail of the distribution is predominantly male. It's the hormones, stupid -- or maybe that x and y thingie, who knows.

Both my daughters are CTY veterans with lots of math and both scored very high on SAT math -- but they could never compete with boys in high-end college math courses. They're very happy doing other things, however.

randy writes:

i have been watching this story develop, and finally after seeing a couple reports which did not mention that the average says nothing about the upper-end...i tracked down an article on sciencemag.org which had more details.

on that site they said that yes, in the upper end of math abilities white boys outnumber white girls 2 to 1, but among asians the girls outnumber the boys 2 to 1.

i've got my own pet theory as to why this is the case. ask me, if you really want to hear it.

loki on the run writes:

LuboŇ° Motl, an ex-Harvard University male :-) physicist had a posting on this.

A Faustian bargain indeed.

steve hsu writes:

Their claim that the F:M ratio is 2:1 at 99 percentile for Asian Americans is very strange. It's not that way in NE Asian countries like Japan, Korea, Taiwan (see below).

In the PISA data the M:F ratio among high achievers is greater than 1:1 for every country. There is variation in the ratio, which points to cultural factors, but it's never 1:1 and the male variance is always larger than the female variance.


PISA 2006 data available here:

http://spreadsheets.google.com/ccc?key=pxL1tXTYbfAAEdVd38vQJEw

Sheets 6.2b and 6.2c have the math results by gender.

Note the PISA data is the highest quality data set of this type -- the tested populations are nationally representative, tests are carefully translated into different languages, etc.

loki on the run writes:

callejohan, do you have any cites or links?

Lord writes:

Are they also going to show boys do as well as girls verbally, demonstrating the lack of any difference?

Comments for this entry have been closed
Return to top