It’s good to be king

Innate ability may explain gender gaps

In 1999, the Committee on Women Faculty at MIT released a report claiming that there was significant gender bias at MIT. Women made up a minority of the Institute’s professorship, and on average were paid less and allotted less lab space; the report alleged this was due to a “subtle but pervasive bias” against women at the Institute. In response, the administration began a concerted effort to recruit more women and increase the pay of female professors. They succeeded, though at the cost of convincing many that women were being given an unfair advantage.

Today, the committee is gearing up for another push. A new report claims that while significant progress has been made, the subtle bias still exists, and more work (read: policy favoritism) is still needed to correct for the prejudices of MIT staff.

At face value, the activism is not unwarranted. Even if pay and lab space have equalized, women still make up less than 20 percent of the faculty of the Schools of Engineering and Science. Perhaps there is even a case to be made for bias at the admissions level: in the School of Engineering as a whole, male students outnumber females by more than 2-to-1, and in some areas, the skew is even greater; male mathematics graduate students outnumber females by almost 4-to-1, and male graduate physics students outnumber females by almost 4.5-to-1.

And yet, there is something deeply suspicious about the “subtle bias” that is supposedly responsible for keeping women down. It’s difficult to point to any hard evidence supporting the claim that prejudice is the root cause of gender gaps at MIT, and the soft evidence is mixed — for every woman with an anecdote about how the Institute has not given her what she deserves, there is a man with the opposite perception. Moreover, even if the soft evidence were not mixed, the plural of “anecdote” is still not “data” — empirically unsubstantiated claims of systemic bias are not a compelling basis upon which to make radical policy changes. And Veldman can draw all the trend lines he wishes between 1970 and today — the fact that Susan B. Anthony had to fight for suffrage does not mean that Susan Hockfield has been denied the opportunity to reach her full potential.

What then of the skewed gender ratios? Aren’t they evidence themselves of bias?

Perhaps. But that is just one hypothesis. And before we launch into a second round of increasing favoritism shown to women (while adamantly claiming that we are doing anything but), let’s entertain at least one other hypothesis, that the distribution of men and women in the upper echelons of science and engineering is not due to an invisible-and-unspoken conspiracy, but instead is due to something simpler: there are simply more smart men than there are smart women.

I hear the sounds of an angry mob already.

Let’s start by taking for granted that men and women, on average, are virtually identical in intelligence, even as it applies to areas like math or science. This is a pretty safe position — most of the literature puts men no worse than women, and a decent chunk even puts them ahead, claiming a point or two of higher IQ for men.

Let’s also take for granted one other point, that measured IQ scores are a very strong predictor of outcomes in life. The Bell Curve by Herrnstein and Murray is the magnum opus of this position, but a hundred other studies could serve in lieu of it.

Now let’s add in one final factor: suppose the variability of male intelligence is higher than that of women. This idea has quite a bit of support in the literature, and since the point I am trying to make is highly controversial (it cost Larry H. Summers ’75 his job), here is a laundry list of supporting studies: Deary, Thorpe, Wilson, Starr, and Whalley (2003) looked at a 1932 IQ test administered to nearly all 11-year-olds in Scotland and found no statistically significant difference in mean IQ, but a statistically significant difference in variability in male children. Hedges and Nowell (1995) performed a metastudy of national IQ tests from over a 32-year period and found similar results. Irwing and Lynn (2005) found that men outnumber women 2-to-1 at the 125 IQ level, and 5.5-to-1 at 155. Machin and Pekkarinen (2008) found higher variance in math and reading scores of male children in the Organisation for Economic Co-operation and Development (OECD). Deary, Irwing, Der, and Bates (2007) studied sibling pairs and their scores on the Armed Services Vocational Aptitude Battery (ASVAB) and found higher male variance. Ardin and Plomin (2006) tracked roughly 10,000 twins from age 2 to 10 and measured a higher variation in intelligence in male twins from 3 years old on.

There is, as with anything, an evolutionary argument as to why relatively higher variability of male intelligence (or male ability in general) would be advantageous. A woman who finds herself three standard deviations above her peers in intelligence can’t take much advantage of that from a procreative standpoint — no matter how many resources she is able to collect, her reproductive success is still limited by the human gestation period and a high risk of death in childbirth. But a man who finds himself well-ahead of his peers can go to town — Genghis Khan took such advantage of his alpha male dominance that there is a 0.5 percent chance you are related to him. It’s good to be the queen … but it’s even better to be the king, and this favors a relatively riskier evolutionary strategy when it comes to male offspring.

What is the consequence of equal average intelligence between the sexes, a strong linkage between intelligence and outcomes, and a higher variability in intelligence for men? For an institution like MIT, which presumably selects its professors from the top 1 percent of the population, the relative flatness of the male ability distribution means two things: firstly, men are going to outnumber women, with the degree of skew increasing as standards are raised. And secondly, within any elite group, the average male intelligence is likely to be higher than the average female intelligence.

In many ways, the IQ variability hypothesis does a better job of explaining the world than does the “subtle bias” position. Claiming that prejudice is responsible for our observed world is a delicate balancing act — one needs to show that U.S. politics, corporate boardrooms, and high technology centers are all vulnerable to invisible bigotry, but conversely that in those axes in which women outperform men, whether it is their lower rates of high school drop out, incarceration, or mental retardation, the bias is gone and the observed outcomes are somehow due to some unrelated factors. Conversely, higher male IQ variability explains nearly the whole range of outcomes; men significantly outnumber women in the tails of life’s outcomes because, quite simply, there are more men in the tails of the ability distribution.

Before we embark on a second crusade for gender equality at MIT, we should have a thorough debate about the causes of our gender gap, and this time, we shouldn’t just shout down those like Larry Summers who dare to conjecture about alternate theories. Perhaps the Committee on Women Faculty is right, and MIT is not a meritocracy. But what if they are wrong? The consequence for adopting increasing favoritism towards women will not just be the deterioration of our meritocracy, but the institutionalization of a real bias and an increasing sense that the women at MIT have not earned their place.

Anon over 12 years ago

I am impressed that the Tech published this; I was afraid that college campuses were too politically correct to even discuss things that liberals do not like.

One more piece of evidence which I didn't see mentioned here is that in school special education classes, boys significantly outnumber girls, which is more evidence against bias and for the greater male variability.

Kathy Barker over 12 years ago

The prejudice is the valuation of certain kind of intelligence: good science (and a better world) require more than the intelligence defined by particular tests and Genghis Khan techniques and outcomes.

It isn't news that male and females brains and ways of thinking are different from each other, nor that there are other cultural, genetic, environmental basis for difference. This brings diversity of approach and creativity to an endeavor, and when intelligence is defined as narrowly as this author wants it, many intelligent and creative people will be excluded, as theybhave been historically.

The world cries out for solutions, and the dominant definitions not only of intelligence but success and progress and happiness must be rethought. Sorry, Genghis, you don't get to keep it all.

Anonymous over 12 years ago

This article if a catastrophe of question-begging and faulty reasoning. Just a few of its many, may errors:

1. IQ is not "intelligence."

2. The Bell Curve did not analyze data from a proper IQ test, nor was it peer reviewed. It is not in any way 'the magnum opus' on the predictors of "outcomes in life," incidentally a term this piece's author fails even to attempt to define.

3. The assertion that MIT "selects its professors from the top 1 percent of the population" begs the question: top 1 percent of what? "Intelligence"? Drive? Connectedness? Salespersonship? Probability of success in research career? Existing scholarship? Teaching ability? Ethics and integrity? Potential long-term contributions to the community of science and academia? That the author fails even to consider these questions reveals his poverty of imagination.

3. Finally, the argument itself is internally inconsistent. Suppose we grant the author's convenient assumption of greater variation in intelligence among men among candidates for faculty positions at MIT. This might account for a portion of the over-representation of men in the faculty as a whole. However: it would have no bearing on the disparate treatment of women who ARE faculty (they've already cleared the bar of admission). If anything, given historic biases, the women who manage to get hired as MIT faculty are likely to be smarter and more talented than their male peers.

Thus: discrepancies in lab space, disparity in salaries, etc. are (contra the pathetic hand-waving of this author) decisive evidence of sex-based discrimination and bias in the treatment of MIT's faculty members.

In sum, this article - combined with the author's sex - reflects incredibly poorly on the men of MIT. I trust the administration, and the editorial staff of the Tech online, will take this into account in future hiring and admissions decisions.

Keith Yost over 12 years ago

1) IQ is a test that reveals intelligence. IQ test results and intelligence are correlated-- and even if they weren't, IQ test results are strongly correlated with outcomes in life (wages, lifespan, educational achievement, arrest record, etc etc), which, even if there was no causative effect, would be enough to erase claims of sexism (and replace them with claims of IQ-ism). This is a mathematical relationship, no one is claiming that IQ scores are perfectly correlated with some omnific intelligence factor.

2) The commenter's assertions are incorrect. I don't know another way to put it. The Bell Curve used one of the best sources of data we have, the NSLY, and it doesn't matter whether or not it was an "IQ" or an ASVAB test-- so long as it is a test whose results are strongly correlated with intelligence, you can do statistics with it. It was also a book, not an article-- it wouldn't get peer reviewed no matter how rigorous its research.

3, First one) Let's say MIT selects from the top 1 based holistically all of those categories, but intelligence has a causative effect on all of them (i.e. a more intelligent individual is more likely to have a higher probability of success in their research career teaching ability etc etc. The author "failed to consider" those questions because they were irrelevant-- whether MIT selects only the top 1 or selects probabilistically from the top n (the end effect of throwing in other, non-intelligence considerations) only affects the sex ratios that come out. In short, you have failed to consider what your claim means mathematically, and so even if I granted you the point, you would not have made your case.

3, Second one): Not only would higher variability of male intelligence mean that more men get in than women, it would also mean that the admitted men would be more intelligent than the admitted women, on average. Draw out the normal curves, super-imposed on one another, and draw a line anywhere to the right of the mean: the section to the right of that line will have a higher average for the flatter curve. Again-- you have failed to grasp the math in its entirety, and your understanding has suffered as a result.

Anonymous over 12 years ago

1. "even if there was no causative effect..."

If IQ differences are not causative, it doesn't make sense to posit them as the cause of sex disparities at MIT.

2. NLSY is fine for what it is, not what you want it to be. As statement of fact: NLSY did not measure IQ. Arguments around IQ therefore cannot rely on the Bell Curve (again, not peer reviewed; that it is not an article is a red herring.)

"so long as it is a test whose results are strongly correlated with intelligence, you can do statistics with it"

Nonsense. Buying cigarettes is strongly correlated with lung cancer. That doesn't mean buying cigarettes causes lung cancer. (Hint: it's smoking that is the cause).

What you really want to claim here is that the NLSY tests are a VALID index of intelligence, so that we can use analyses of those data to make causal statements about intelligence and outcomes. To do this, you need to explain what you mean by intelligence, and support your assertion with facts.

3. "Let's say.... a more intelligent individual is more likely to have a higher probability of success in their research career teaching ability etc"

This is an example of assuming what you hope to prove. You need evidence.

4. Incorrect. You again are assuming what you want to prove (unbiased hiring).

Again: suppose we grant the specious assumption of a difference between sexes in the marginal distributions of intelligence across all people (i.e. longer tails among men).

It does NOT follow that there must be similar differences in the conditional distribution of intelligence among men and women who happen to apply to MIT and are subsequently hired.

This is trivial: even if we assume that "very smart" men outnumber "very smart" women 100 to 1 in the general populace, it remains possible that EACH AND EVERY female faculty member at MIT is more intelligent than EACH AND EVERY male faculty member.

You can work this out by basic theory, but common sense is sufficient. You could create such a faculty from scratch, given the power to choose.

Besides yourself, however, there are many more insidious mechanisms that might lead to most women at MIT being smarter than most of its men - even IF there are more "really smart" men elsewhere.

In rethinking this, you might consider the obvious potential causes of such a discrepancy: namely discriminatory practices in hiring (a historical fact).