### Bulgaria 21 : USA 3

*by Christina Hulbe*

It's a bit difficult to see where articles about science fit in with the political and economic stories these days. That fogginess, combined with a busier than expected end of summer has kept me silent here for a while. Well, a grant proposal I finished up last week and a research paper about diversity and mathematical talent that I saw referenced by Phila at echidne's place inspired me to write this post. The topic may not quite fit what I was originally invited here to write about but if you care about the future of the scientific endeavor in the U.S. then you should probably care about diversity in science and math.

The grant proposal (really a pre-proposal, a subset of applicants will be invited back for the full proposal) was to a National Science Foundation program that requires significant programmatic dedication to improving diversity in science and engineering. To meet the requirement, my colleagues and I outlined a set of programs involving geoscience and mathematics for students and their teachers from elementary school onward, with emphasis on traditionally underrepresented groups.

Our interest is not purely altruistic or philosophical. We want to find the best students possible to populate our research programs and to become the next generation of research scientists. If we are missing out on entire groups of candidates because they don't have access to the right pre-collegiate and collegiate training, we're not going to accomplish that goal (link to an NSF report). In the past 28 years, only 2% of the approximately 20,000 PhDs awarded in science and technology fields have gone to traditionally underserved minorities, yet the U.S. Census Bureau expects that over 41% of the U.S. workforce will be composed of ethnic minorities by 2050.

Just as I was wrapping up the proposal, I saw Phila's reference to *Cross-Cultural Analysis of Students with Exceptional Talent in Mathematical Problem Solving*, a research article by Titu Andreescu and colleagues, just published in the Notices of the American Mathematical Society. The journal article is here and a press release is here. I couldn't wait to read it.

Andreescu and colleagues are interested in how countries identify and develop the kind of math talent that leads to PhDs, professorships in mathematics, and important research accomplishments. In a nutshell, countries in which skill in mathematics is valued and nurtured produce more high-achieving mathematicians than do countries where such skill is not valued. This cultural effect is clearly amplified for girls. This runs counter to arguments about gendered abilities made by some men (for example, Lawrence Summers' famous (infamous?) remarks here and Charles Murray's notions via paid access here).

Andreescu and colleagues use data from elite U.S. and international pre-collegiate and collegiate mathematical competitions to perform their analysis. These data are preferable to SAT scores and similar data, sometimes used to argue that females are simply not so good at math, because the competitions are designed to test creative mathematical thinking and insight (with essay-style proofs and so on), qualities that indicate exceptional talent.

I'll consider just the International Mathematical Olympiad (IMO), a pre-collegiate competition. Participating countries send up to 6 competitors each year. The United States tends to stay in the top 15 (of about 95) but our teams include relatively few girls (3 different girls have made 8 appearances since 1995) and in recent years, about half of our team members have been first or second generation immigrants, many of whom began their math training elsewhere. Eastern European and Asian teams include much larger percentages of girls and consistently out-perform teams from other regions of the world.

The relative representation of girls on IMO teams varies over time. This is important because together with variations from country to country, variations over time indicate that female underrepresentation in elite-level mathematics does not arise from anything intrinsic to females; if it did, female representation on these teams would be fairly constant over time. Here are data for few countries to illustrate the differences (from Table 6 of Andreescu et al.):

% girls on IMO team in three time periods: 1)pre-1988 2)1988-1997 and 3)1998-2008

South Korea

1) -

2) 3

3) 11

Bulgaria

1) 10

2) 5

3) 12

East Germany

1) 7

2) 11

3) -

West Germany

1) 0

2) 0

3) -

Russia (USSR)

1) 3

2) 20

3) 5

Serbia & Montenegro (various political incarnations)

1) -

2) 8

3) 24

USA

1) 0

2) 0

3) 8

Some of these totals represent multiple appearances by the same competitor. For example, the US sent only 3 different girls to the IMO from 1995 to 2008.

Bulgaria is a tiny country (0.11% of the world population) yet it far outpaces more populous countries when it comes to placing girls on its IMO team and consistently ranks in the top 15. Andreescu and colleagues calculate a median team rank of 5.5 for Bulgaria over the period 1995 to 2008, with 21 different girls competing. Russia, with a larger population (2.1% of the world population) also sends relatively large numbers of girls to the IMO and their median team rank over that period is 2.5, with 15 different girls competing. The girls are clearly performing well (they wouldn't be on the teams if they didn't). Just for reference, the US (4.5% of the world population) has a median rank of 3 over this time period and has sent only 3 different girls to the competition. Check out the paper for complete data.

The authors conclude "some countries routinely identify and nurture both boys and girls with profound mathematical ability to become world-class mathematical problem solvers; others, including the USA, only rarely identify girls of this caliber."

The word *identify* is important. The large variance in female participation across countries indicates that low representation in any given country is not necessarily a result of innate ability but is definitely a result of the selection process, which involves both educational opportunity and socio-cultural bias. Quoting Andreescu and colleagues again, "In a gender-neutral society, the real percentage [of girls with exceptional math talent] could be significantly higher; however, we currently lack ways to measure it." So that 24% in Serbia & Montenegro is very likely an underestimate of the true percentage of girls (and women) with profound mathematical ability.

So how does this circle back to the selfish interest in finding the best possible students for math and science graduate programs? In the US, boys and girls perform equally well in math through elementary school. Girls appear to fall behind in middle school but still earn nearly half of all bachelor degrees in mathematics. By the doctoral level, something has changed, women earn only about a quarter of math PhDs in the US (data from 1966 to 2004 here). What Andreescu and colleagues' study suggests is that we can't fall back on differences in ability to explain this underrepresentation.