More on the Larry Summers hypothesis

I want to make a few followup points regarding yesterday’s post about the gender gap in the sciences.

Just to recap, the “Larry Summers hypothesis” is the idea that differences in intrinsic intellectual ability can explain the underrepresentation of women in academic science.  More specifically,  the proposed explanation is that, even if there’s no difference in average ability, men tend to have intellectual ability (as measured by scores on various tests) that scatters more widely than women.  That means that men populate the very extreme highs and lows of the distribution much more than women.  If being a successful scientist requires being in the extreme high end of this distribution, then that might explain the gender gap.

I don’t think this is likely to be a significant part of the explanation, for reasons I tried to explain.  The biggest one is that I don’t think that success in a scientific career is sufficiently strongly correlated with intelligence (where the “intelligence” is defined to mean “the thing that the tests in question measure”).  To be more specific, I certainly don’t think that it’s strongly correlated with presence in the high-end tail of the intelligence distribution, which is where the gender differences supposedly lie.

If we were just trying to explain the preponderance of men among Nobel Prize winners, maybe a case could be made for this effect (although even then I’m not sure).  But in fact the gender gap shows up much earlier and grows at each step.  (The “leaky pipe” is the usual metaphor here.)  More men than women major in physics; the ratio skews further at the Ph.D. level, still further at the level of faculty jobs, and even further among tenured faculty.  If you think that all of this is because of effects in the top 1% tail of the intelligence distribution, I’m afraid you’re making  a flattering overestimate of the physics community’s intelligence.  We’re tolerably bright, but not that bright.

But there are a couple of things I want to make clear:

1. People sometimes use your attitude about the Summers hypothesis as a proxy for your attitude about all sorts of other things: if you disbelieve the Summers hypothesis, you must believe in the mind as a blank slate, with no room for intrinsic biological differences between the sexes.  Conversely, if you believe in the Summers hypothesis, (according to some) you must be a sexist.  Those attitudes are ridiculous.  In particular, although I think the Summers hypothesis is probably wrong, I think that that intrinsic cognitive differences in the sexes are quite likely to be real and may explain all sorts of other phenomena. I’m even generally sympathetic to the evolutionary psychology point of view, which is anathema to a lot of people who argue against the Summers hypothesis.

(Incidentally, if I may play armchair psychologist for a moment, it seems to me that the authors of that NOVA blog post, which got me started on this whole subject, are making this error: they saw Tierney as sympathetic to the Summers point of view and concluded that he must believe in the whole constellation of despised ideas that they associate with that point of view.  I can’t see any other reason they would have so egregiously misrepresented what he said.)

2. I don’t think that adherents of the Summers hypothesis are bad people, and I don’t think that the hypothesis should be ruled out of the bounds of polite discussion (as many people seem to).  It’s an a priori possible explanation of the observed data, which is either true or false.  It has the right to a hearing, and its probability of being correct can and should be judged on empirical grounds, like any other hypothesis.  Personally, I think it must be found wanting on those grounds.  But it should not be ruled out a priori because we don’t like its social or political implications.

3. It may seem to follow from point 2 that Larry Summers got a raw deal, but my sympathy for him is extremely limited.  If you’re a politician (and yes, a University president is a politician), then you should know better than to speak off the cuff about an extremely controversial topic about which you clearly have given very little thought.  He made a boneheaded move in raising the subject the way he did, and he got what was coming to him.  Moreover, as is often the case when politicians are brought down by gaffes, this one probably wouldn’t have brought him down if he didn’t have a history of alienating people.

5 Responses to “More on the Larry Summers hypothesis”

  1. “More men than women major in physics; the ratio skews further at the Ph.D. level, still further at the level of faculty jobs, and even further among tenured faculty. If you think that all of this is because of effects in the top 1% tail of the intelligence distribution, I'm afraid you're making a flattering overestimate of the physics community's intelligence.”

    A back-of-the-envelope calculation shows that less than 0.5% of people study physics. That’s rather far into the tail (remember that Eisenhower was shocked when an aide informed him that 50% of US citizens were below average in intelligence). Some drop out, some become teachers, some go into industry etc. Ability is not the only factor in determining who stays on in academia, but it is one, so among physicists in academia we are talking about probably the top 0.1%. The top 0.1% in terms of “physics ability”. The fact that there isn’t a standardised test for this doesn’t make it irrelevant.

    I think it is setting up a strawman by saying that the Summers hypothesis requires tenured physicists in academia or whatever to be the top 0.1% in terms of intelligence (or some other result of a standardised test). All that is necessary is that they are in the top 0.1% in terms of physics ability. The same applies to other areas where only a small percentage of the population are active, ability is essential and the position is attractive: cooks at high-class restaurants, formula-one racecar drivers, maybe even rock musicians.

    There are many more physicians in the general population than physicists in academia, and there are more women among physicians.

    In summary, I don’t see massive evidence against the Summers hypothesis.

    Another question is, even if men and women are, on average, the same in terms of general intelligence, physics ability etc, they are probably different on average in other areas. Why do more men play chess? That might be related to the ability to think spatially etc. But why are there so many more men in rock music? (Sure, the famous people are also the top 0.1% or whatever, but the bias is already there at lower levels, both in chess and in rock music.)

    What puzzles me about rock music is that there is no obvious characteristic which is typically male which seems to be important. (OK, the female equivalent of, say, Robert Plant might seem a bit far-fetched, but consider Janis Joplin.) Even if whatever it is is not easily definable, if it is a Summers effect we would expect it only at the top, but that doesn’t seem to be the case. Maybe the Summers effect applies at the top, and lower down the motivation is to be at the top, even if one can’t make it. In other words, role models. (I don’t think that the lack of role models in other branches explains much, though. I don’t think men go into physics because of male role models and I don’t think that women avoid physics because of the lack of female role models.)

  2. Ted Bunn says:

    “The top 0.1% in terms of "physics ability". The fact that there isn't a standardised test for this doesn't make it irrelevant.”

    If you define “physics ability” to mean “the undefined set of characteristics that lead people to become physicists,” then I agree that professional physicists are in the top tail of that distribution. In other words, if you define the hypothesis in question to be a tautology, then it’s true (like all tautologies). If, on the other hand, you want the hypothesis to have any explanatory power at all, then “physics ability” has to mean something else, and in fact something measurable in other ways (e.g., by a test).

    In any case, the original Summers hypothesis was in fact precisely the one you call a straw man, namely that mathematics ability as measured by tests was the relevant factor. I’ve explained why I think that that hypothesis is abundantly falsified.

  3. I think the truth probably lies somewhere in between a tautology and a measurable-by-test proxy. Assuming that physics ability plays a role in getting a physics job, then of course it is a tautology that professional physicists are in the high tail of the distribution. Ditto for any profession where only a small fraction of the population practice it, it is desirable and ability is important. In most of these cases, there is a dominance of men (composers, chefs, chess grand masters, astronauts, physics professors, formula-one racecar drivers, computer programmers), so the general idea that the distribution of male talents is broader and that this leads to a dominance in such fields is I think not completely absurd. On the other hand, in at least some cases there is a correlation with a measurable ability. (Also, the fact that there is no standard test today doesn’t rule out that a standard test is possible.)

    Of course, a broader distribution means that men also dominate the bottom of the barrel.

    Old statisticians never die—they just get broken down by age and sex. :-)

  4. being a computer programmer myself makes me very proud of my job,*.

  5. dfgfgf says:

    Bizarre this post is totaly unrelated to what I was looking google for, but it was listed at the first page. I guess your doing something right if Google likes you sufficient to position you on the first page of a non related search.

Leave a Reply