The National Merit Scholarship is a competition for recognition and scholarships. About 0.5% of each class become semifinalists, and about 95% of those semifinalists become finalists. The threshold score varies by state, adjusted so that half a percent of students in each state qualify – which means that the threshold scores varies somewhat by state.
Some finalists get a small scholarship from the organization itself ($2000 over four years) but more scholarships, with more money, are available from some colleges, also to children of employees in some corporations, etc. Free rides are possible.
The test is not quite a national IQ test with prizes, since the winning threshold varies by state – but it’s pretty close to one, probably closer than anything else. The threshold is about 2.58 stds above average – which would be about 139 on an IQ test.
There are those who don’t like this test, probably mostly because they don’t like the overall results, some because they personally did poorly, but there are other reasons. It is certainly the case that members of influential classes in this country are probably chosen more by test scores than they were once upon a time, while they seem to be crazier and less competent than they were back in the day: you have to wonder. Not that our political class is a mandarinate yet: you can find Senators with combined SAT scores of 800, Governors with an ACT score of 18.
One objection is that very few members of minority groups become finalists. Patrick Hayashi, a retired senior University of California official who had overseen admissions at UC Berkeley for ten years, asserted that not one of the hundreds of National Merit Scholars who came to the campus during those years was black or Hispanic. He estimated that “the percent of National Merit Scholars who are black, Hispanic, and American Indian is close to zero and that the absolute number of poor students from these groups is also close to zero”.
To Hayashi, this proves that there is something wrong with the test, but of course it is exactly what you should expect, since the fraction of a group with a significantly lower mean of some quantitative trait that exceeds a high threshold (in height, IQ, whatever) is very much smaller than in a group with a significantly higher mean, as I have pointed out before. Consider black Americans. If their mean IQ is one std lower than whites, while the width of their distribution is lower (12 points instead of 15) – then for them the NMSPQRT threshold is 4.475 standard deviations. Instead of 1 in 200, the fraction of winners is less than 1 in 200,000.
This is too simple: IQ can’t be exactly Gaussian, blacks in the US are not perfectly homogeneous, etc. But it does show the trend: such high scores are much, much, much rarer in groups with low mean scores. It’s also the case that any black kid with such a high score would get a far better offer from Harvard.