December 7, 2005

Do black women have higher IQs than black men?

Thomas Sowell has been arguing for a number of years that black women score higher on IQ tests than black men:

Years ago, while doing research on education and IQ, I happened to be in the principal's office at a black school in Cincinnati, as he was preparing to open a large brown envelope containing the results of IQ tests that his students had taken. Before he opened the envelope, I offered to bet him that a large majority of the students with IQs over 110 would be girls. He was too smart to take the bet.

Studies had shown that females predominated among high-IQ blacks. One study of blacks whose IQs were 140 and up found that there were more than five times as many females as males at these levels. This is hard to explain by either heredity or environment, as those terms are usually defined, since black males and black females have the same ancestors and grow up in the same homes. Meanwhile, white males and white females have the same average IQs, with slightly more males at both the highest and lowest IQs.

This is just one of many unsolved mysteries that is likely to remain unsolved, because doing research on race and IQ has become taboo in many places. My own research was financed in part by a grant from a foundation that told me to remove any mention of IQ research from the activities listed in my project's application. They didn't care if I used their money for that purpose but they did not want it on the record that they had financed research into race and intelligence. Many schools and boards of education also did not want it on the record that they had cooperated by supplying data for any such research. Only when assured of complete anonymity would they let me into their records.

So, to test this, I asked Charles Murray to take a look at the most recent renorming of the U.S. military's IQ test that all applicants for enlistment must take, the Armed Forces Qualification Test or AFQT, and he graciously obliged, sending me some new and important data that I have not seen reported before.

(Why haven't many seen this elsewhere? The U.S. Armed Forces are both extremely committed to IQ testing of potential enlistees, which they began doing in 1917, and aware that the mainstream media blindly condemns IQ testing as racist pseudo-science. So, the military tries not to draw public attention to its enormous investment in IQ testing and research. The unfortunate side effect of the military's reticence is that this allows know-nothings to blither in the press about how IQ has been discredited.)

As Murray noted in his recent Commentary article (see Footnote 41), the previous renorming of the AFQT in 1980 now appears to have underestimated average black IQ, reporting an improbably large 18.6 point gap between the races, due to low-scoring blacks, especially black males, being more inclined to give up partway through the daunting 105 page pencil-and-paper test and not even trying to answer the remainder of the questions. To make up a fictitious illustration, a black guy who could have scored 75 if he'd toughed it out and tried to answer all the questions might get depressed by looking at so many questions he couldn't answer and give up on p. 57 and get a 65 instead.

To get around this, in the 1997 renorming of the AFQT, the test was given on a computer and if you missed a lot of the early questions, the computer would feed you easier questions to keep you from giving up. This also helps make the test more discerning by finetuning the questions more appropriate for your general level of intelligence.

This points out two sides to the old question of how much does the type of IQ test matter. In some ways, it doesn't matter that much. There's an old saying among psychometricians that "Life is an IQ test." La Griffe du Lion has shown that the roughly one standard deviation difference in IQ between blacks and whites shows up over and over again in the real world -- for example, in Florida in the famous Presidential balloting in 2000, blacks tended to botch up their ballots (e.g., by voting for two different candidates for President) and thus make them uncountable at a rate about a standard deviation higher than whites did, costing poor Al Gore the White House.

On the other hand, when attempting to make subtle comparisons such as between the sexes or to see if the racial gap is closing over time, then the fine points of the different tests matter. I only have one test to report here, but it's a big one, the AFQT, which the U.S. military has invested millions of dollars in validating over the last half century.

So, here is the data from the U.S. military's 1997 renorming of the AFQT on computers, as provided to me by Charles Murray, "using the 1997 cohort AFQT converted to an IQ metric, and employing population weights to reach nationally representative results." The renorming was done on the 1997 National Longitudinal Study of Youth, a large scale sample of people approximately aged 15-23, I believe. "White" means "non-Hispanic white."

AFQT IQ Black White Diff.
Male 88.4 102.7 14.3
Female 90.8 103.6 12.8
Diff (2.4) (0.8)




St. Dev.


Male 13.30 14.75
Female 13.58 13.30


So, at least on the AFQT, Sowell appears to be right: black women score 2.4 points higher than black men, while white women score only 0.8 points higher that white men.

Along these lines, several readers have sent me a new article about a gene variant that drives down IQ in males but not in females. (It was tested only on whites.)


Scientists in North Carolina say they have identified a gene that affects IQ, a finding that, if confirmed, would be a significant step toward understanding the genetic basis for intelligence. The new research could also have ethical implications because the effect of the gene appears to be quite dramatic: The scientists say that males who inherit a particular version of the gene have, on average, an IQ that is 20 points lower than males who don't.


It's also interesting to note that white males have significantly higher standard deviations than the other three groups. This greater variability among white males may be related to the continuing dominance of white males in Nobel Prizes and other measures of extreme right edge of the Bell Curve performance.

Perhaps the most striking aspect is the smaller white-black IQ gap than in 1980. This shrinkage is partly due to the new test format that blacks don't find so discouraging. But, as Murray said in his Commentary article, it may well also represent a genuine closing of the gap. Due to the apples vs. oranges differences in the AFQT methodologies between 1980 and 1997, we can't say for sure. But, this is something that should be tracked closely in the future.


My published articles are archived at iSteve.com -- Steve Sailer

4 comments:

Anonymous said...

The AFQT is not an IQ TEST smh. Its a test on what what you know. I have tooken the test already.

Anonymous said...

Its because black people are not as educated as whites... jeez....the iq test does not test common sense....it test education level. I mean it has math and english questions....i am certain if black people were better educated we would do a lot bette

Anonymous said...

Thats not true at all. There is no trivia on IQ tests. Everything on it can be figured out with pure problem solving skills and being able to recognise patterns.

Anonymous said...

Wow. Seems we have both sides of the famous bell curve here. "WE would do alot bette"? I too have "tooken it" (?) Proof comes in many ways. Wow... amazingly no test is given for those using the internet. Maybe we should. What was that famous Forrest Gump quote.... "“Stupid is as stupid does.” – Forrest Gump" Bet Forrest never posted to the internet.