The Bell Curve Flattened

The Bell Curve Flattened

The Bell Curve Flattened

Jan. 18 1997 3:30 AM

The Bell Curve Flattened

Subsequent research has seriously undercut the claims of the controversial best seller.

(Continued from Page 1)

Having conditioned its audience to view IQ as all-important, The Bell Curve then manipulates statistics in a way that makes IQ look bigger, and everything else smaller, in determining Americans' life-chances.


The basic tool of statistical social science in general, and of The Bell Curve in particular, is regression analysis, a technique used to assign weights to various factors (called "independent variables") in determining a final outcome (called the "dependent variable"). The original statistical work in The Bell Curve consists of regression analyses on a database called the National Longitudinal Study of Youth. The authors claim to demonstrate that high IQ is more predictive of economic success than any other factor, and that low IQ is more predictive of poverty and social breakdown. Virtually all the early commentators on The Bell Curve were unable to assess the merits of the regression analysis. "I am not a scientist. I know nothing about psychometrics," wrote Leon Wieseltier (who was otherwise quite critical) in a typical disclaimer.

But by now the statistics have been gone over by professionals, who have come up with different results. The key points of their critique of The Bell Curve are as follows:

What Herrnstein and Murray used to measure IQ is actually a measure of education as well as intelligence. All the people tracked in the National Longitudinal Study of Youth took the Armed Forces Qualifying Test, which Herrnstein and Murray treat as a good measure of intelligence. Because the material covered in the test includes subjects like trigonometry, many academic critics of The Bell Curve have objected to its use as a measure only of IQ and not at all of academic achievement. Herrnstein and Murray concede in the footnotes that scores tend to rise with the subjects' education--but they seriously underestimate the magnitude of this rise, as shows. And they resist the obvious inference that the test scores are measuring something other than intelligence.

Most of The Bell Curve's analysis is devoted to proving that IQ has more predictive power than parental "socio-economic status." But Herrnstein and Murray's method of figuring socioeconomic status seems designed to low-ball its influence, as explains.

Herrnstein and Murray begin their discussion of the National Longitudinal Study of Youth data by announcing that they aren't going to analyze the effect of education, because education is too much a result of IQ. It's not an independent variable. (Of course, according to their theory, socioeconomic status is also a result of IQ, but somehow, that doesn't stop them.) Therefore, what you'd most want to know from a policy standpoint--how much education can increase opportunity--isn't dealt with in the book, except in two obscure footnotes. Both would seem to support the liberal, pro-education position that Herrnstein and Murray say is futile. One footnote shows education increasing IQ year by year. The other shows a higher correlation between college degree and family income than between IQ and family income.

One of The Bell Curve's theoretical linchpins is the high heritability of IQ. Herrnstein and Murray, sounding like the souls of caution, write that "half a century of work, now amounting to hundreds of empirical and theoretical studies, permits a broad conclusion that the genetic component of IQ is unlikely to be smaller than 40 per cent or higher than 80 per cent. ... For purposes of this discussion, we will adopt a middling estimate of 60 per cent heritability." This now looks seriously overstated. Michael Daniels, Bernie Devlin, and Kathryn Roeder of Carnegie Mellon University took the same studies on which Herrnstein and Murray based their estimate, and subjected them to a computer meta-analysis ("a powerful method of statistical analysis"--The Bell Curve). Their paper, which has not yet been published, says: "In brief, studies of IQ, and our reanalyses of them, suggest a narrow-sense heritability of 34 per cent and a broad-sense heritability of 46 per cent. [The difference between broad and narrow is too technical to explain in this limited space.] This is a far cry from Herrnstein and Murray's maximum value of 80 per cent or their middling value of 60 per cent. Consequently, Herrnstein and Murray give the impression that IQ is highly 'heritable,' but it is not."

If the purpose of the whole exercise is to figure out what our social policies should be, then, "Which is more predictive, IQ or socioeconomic status?" isn't the essential question anyway. Making it the essential question avoids the issue of whether IQ is really so massively predictive that it drowns out everything else. (Herrnstein and Murray mostly leave the evidence for this, their central contention, to footnotes. The figures they offer are far from dispositive.)

The chapter of The Bell Curve on policies that might be able to overcome the fate of a low IQ focuses mainly on whether early-childhood programs like Head Start (most of which aren't run with raising IQ as their primary goal) can raise IQ significantly over the long term, and sorrowfully concludes that they can't. What the book doesn't discuss is whether public schools--by far the biggest government social program--can raise IQ, or earnings after you control for IQ. As James Heckman of the University of Chicago wrote in the Journal of Political Economy, " Evidence of a genetic component to skills has no bearing on the efficacy of any social policy. ... The relevant issue is the cost effectiveness of the intervention." (As an example of where the kind of analysis Herrnstein and Murray didn't do can lead, a new study by Jay Girotto and Paul Peterson of Harvard shows that students who raise their grades and take harder courses can increase their IQ scores by an average of eight points during the first three years of high school.)

At the beginning of The Bell Curve, Herrnstein and Murray declare that "the concept of intelligence has taken on a much higher place in the pantheon of human virtues than it deserves." And they claim that their view of IQ tests is "squarely in the middle of the scientific road." They end by expressing the hope that we can "be a society that makes good on the fundamental promise of the American tradition: the opportunity for everyone, not just the lucky ones, to live a satisfying life." Throughout, Herrnstein and Murray consistently present themselves as fair- (or even liberal-) minded technicians who have, with great caution, followed the evidence where it leads--which, unfortunately, is to a few unassailable if unpleasant scientific truths that it is their reluctant duty to report.

In fact, The Bell Curve is a relentless brief for the conservative position in psychometrics and social policy. For all its talk of reflecting a consensus, the sources it draws upon are heavily skewed to the right. Herrnstein and Murray used quasi-nutty studies that support their position (as Charles Lane demonstrated in the New York Review of Books), and ignore mainstream studies that contradict it (as Richard Nisbett showed in the New Republic). The data in The Bell Curve are consistently massaged to produce conservative conclusions; not once is a finding that contradicts the main thesis reported in the text. (shows how Herrnstein and Murray have made the convergence in black-white IQ scores, which they claim to find "encouraging," look smaller than it actually is.) The Bell Curve's air of strict scientism doesn't preclude the use of lightly sourced or unsourced assertions, such as the statement that the median IQ of all black Africans is 75, or that "intermarriage among people in the top few percentiles of intelligence may be increasing far more rapidly than suspected" (no footnote). Though they piously claim not to be doing so, Herrnstein and Murray leave readers with the distinct impression that IQ is the cause of economic success and failure, and that genetic difference explains the black-white IQ gap.