Myth: The black/white IQ gap is 15 points and growing.

Fact: The black/white IQ gap is 7-10 points and shrinking.




Summary

Nationally representative IQ tests show a black/white IQ gap of 7 to 10 points. Academic achievement tests, for those who consider them valid measurements of IQ, show an average gap of about 10 points. In general, these tests show the gap is being reduced about two and a half points a decade. The Bell Curve flouts all this evidence on the basis of a single study, its own flawed analysis of the National Longitudinal Survey of Youth.



Argument

Among scientists, it has been common knowledge that both black and white IQs have been rising over the decades, with black IQs converging upwards towards whites. This fits nicely with America's rising standard of living, which has been rising somewhat faster for blacks, thanks to the Civil Rights movement and other anti-poverty measures.

The authors of The Bell Curve, then, face an uphill battle in trying to prove that black and white IQs are not converging, but diverging. They attribute this divergence to "dysgenesis," which supposedly results when dull people interbreed. In fact, they assert this downward trend in spite of the famous "Flynn Effect," which has been raising IQs for all people and all classes world-wide, about 3 points per decade.

In a second departure from scientific consensus, the authors also maintain that the IQ gap between blacks and whites is 15 points, an usually high figure. But this is not what four major IQ tests for children have found:

Black/white IQ gap in major IQ tests (1991) (1)

Ravens Standard Progressive Matrices       7 points
Kaufman Assessment Battery for Children    7
Stanford-Binet IV (two separate studies)  10

The authors of The Bell Curve note these numbers, but dismiss them. The Kaufman-ABC results, they claim, suffer from statistical problems. (2) Even granting them their objection, however, still leaves three major studies showing a 7 to 10-point gap, not a 15-point one.

Why do the authors discount both the Flynn Effect and these test results in their assessment of black IQ? Because of a single study: their analysis of the National Longitudinal Survey of Youth (NLSY). The NLSY has tracked the lives of 12,000 young people since 1979, at one point giving them an Armed Forces Qualification Test (AFQT). But The Bell Curve's analysis of the NLSY data is deeply flawed. First of all, the AFQT isn't even an IQ test. "This is an achievement test," said Halford Fairchild of the American Psychological Association said, in its authoritative review of The Bell Curve. "It shows the extent to which you've benefited from school. To assert it's a proxy for IQ is a big lie." (3)

Even if we accept the AFQT as an IQ test, however, the authors still commit serious errors analyzing it. One of the best-known critics of their methodology is psychologist Richard Nisbett, whose arguments follow here. (4) Herrnstein and Murray's comparison of black and white IQs did not follow the changing IQs of children over a period of time (the most reliable method), but differences of IQ between mothers and children at a fixed point in time (which introduces all sorts of statistical complications). Herrnstein and Murray found that the black/white IQ gap among mothers was 13.2 points, but among their children was 17.5 points. (5) This four-point increase forms the basis of their claim that the black/white IQ gap is growing.

However, the study itself is unrepresentative of the nation at large; the average IQ of the children was 92, significantly below the national average of 100. The reason is because the women in this long-term survey had not finished their childbearing years, so all the children in the study had been born to young mothers. Unfortunately, young motherhood is correlated to low IQ, which is why their children scored so far below the national average.

Herrnstein and Murray note this problem, and claim that sampling weights added to the NLSY make the survey sample "nationally representative." (6) But this leads to trouble when they attempt to analyze a growing IQ gap in children. Young motherhood is not unusual for black women, but for white mothers it suggests low socioeconomic status, which in turn suggests low IQ. Therefore, the IQs of the young white and black mothers were similar… although the IQs of their children were bound to diverge. Why? Because black mothers, having average black IQs, would have children with the same. But white mothers, with below-average white IQs, would have children with IQs that would tend to rebound back towards the higher white average. That's because trait variation is natural in children, and if white parents are already near the bottom, then their children, if they vary, have nowhere else to go but up. This is technically known as "regression towards the mean," and it is a common statistical deception. You can get the opposite "result" by comparing average black mothers with the smartest white mothers; the white children would tend to drop back down to the white average, having nowhere else to go but down. In this case, Herrnstein and Murray could have reported that IQ gap is shrinking, not growing!

Thanks to this statistical anomaly, Herrnstein and Murray found a four-point increase in the black/white IQ gap in their particular sample. On this basis, they then argued that the gap is growing everywhere. But this is certainly not the case, and the mainstream studies remain the best indicator of the shrinking black/white IQ gap.

Other 15-point claims

Herrnstein and Murray cite numerous other studies in their book that produce a broad range of estimates for the black/white IQ gap. These range from no gap to a gap of 30 points (an absurd finding on its face, since this would mean that blacks are mentally retarded).

To put everything in perspective, the authors provide a chart summarizing the findings of 156 studies. (7) The results form a bell curve of their own: the vast majority of studies have found a difference of about 15 points, while only a few studies have found significantly smaller or greater differences. From this chart, one might reasonably conclude that the gap is 15 points.

But a closer look at the chart reveals some disturbing facts. It was compiled in large part by Arthur Jensen, Frank McGurk, R. Travis Osborne and Audrey Shuey -- all recipients of grants from the Pioneer Fund. This is the neo-Nazi organization that advocates eugenic policies -- namely, the phasing out of black people. This alone does not refute the chart, of course, but one would feel more comfortable with its objectivity if its compilers weren't funded by overt racists.

And a look at the footnotes reveals that these fears are well-founded. To be included in the chart, a study had to meet "basic requirements of interpretability." These requirements included several that many would view as fair: for example, that the studies had to include both a black and a white sample, that the sample sizes had to be larger than 50 for each group, etc. But the authors then report that socioeconomic status posed "a special problem" for them. They write:

"If a study explicitly matched subjects by SES, it was excluded. If it simply drew its samples from a low-SES area, it was included, even though some degree of matching had occurred." (8)

It is well-known that matching blacks and whites for socioeconomic status will see a large reduction in the IQ gap. Researchers justifiably dispense with such matching if they want to learn the raw differences. But including samples from "low-SES areas" creates great potential for mischief. Inner city schools are disproportionately filled with poor black children, but relatively richer whites are bussed into their schools, where they take the same IQ tests. Furthermore, an IQ test given to a broad region like a city would include inner-city black children and suburban white children. Both study samples would yield unnaturally large IQ gaps.

The only way to minimize statistical errors (and deception) is to consider nationally representative samples. As our first chart indicates, such tests indicate a 7 to 10-point gap. For those who accept academic achievement tests as valid proxies for IQ tests, the gap in 1990 was about 10 points, after having been reduced about 2.5 points per decade since 1970. (9)

Return to Overview

Endnotes:


1. K.R. Vincent, "Black/white IQ differences: Does age make a difference?" Journal of Clinical Psychology 47, 1991, pp. 266-270.

2. Richard Herrnstein and Charles Murray, The Bell Curve, (New York: Simon & Schuster, 1994), pp. 289-90.

3. Marilyn Elias, "Experts find fault with 'Bell Curve'" USA Today. (No date given; probably Fall 1995.)

4. Richard Nisbett, "Race, IQ and Scientism," pp. 50-52 in Steven Fraser, ed., The Bell Curve Wars (New York: HarperCollins, 1995).

5. Herrnstein and Murray, p. 356.

6. Ibid., p. 355.

7. Ibid., pp. 276-77.

8. Ibid., p. 717.

9. Nisbett, p. 49. The academic tests used for this statistic are listed in the chart below, which has been compiled from his written account:

Point reduction in black/white IQ gap (1970-1990)*

Test                                         Reduction
------------------------------------------------------
National Assessment of Education Progress    4.2 points
Scholastic Aptitude Test                     5.0
National High School Studies                 4.5
American College Test                        4.4
Graduate Record Examination                  5.7

*Total reductions have been extrapolated from different-length
studies to represent two decade period. Original lengths: NAEP,
1970-1990; SAT, 1976-1993; NHSS, 1973-1982; ACT, 1980-1991; GRE,
1980-1988.