What Summers actually said

I've been following the Larry Summers controversy with some interest, but I didn't want to say anything about it myself until I had read his actual words. I finally got around to reading the transcript of the speech on the weekend, and to my less than complete surprise the whole thing seems rather overblown.

To summarise what I had heard second-hand, Larry Summers (Harvard University's President) was invited to address a conference on Diversifying the Science & Engineering Workforce, and upset many people with his words. I haven't been able to find out whether the conference was exclusively about gender disparity or about diversifying the workforce in a more general manner, but certainly gender disparity was a major theme, and it was what Summers chose to talk about. The way his remarks were presented second-hand gave the impression that he had the gall to stand up and lecture such a conference about the inferiority of women.

He actually said nothing of the sort, and I urge anyone who cares to go and read the transcript. It's not the easiest reading, because it is a faithful transcript of a speech rather than a script, but without reading it anything I or anyone else says will just be a quote out of context, and it is a speech that lends itself particularly well to quoting out of context. I will try to summarise what I saw in his speech:

  1. There are very large disparities in the representation of different groups in different professions.
  2. This includes the representation of various ethnic groups in various fields, as well as extreme under-representation of women in tenured positions in science and engineering at top universities and research institutions.
  3. One possible explanation for this is a difference in interests between men and women.
  4. Socialisation probably accounts for part of this difference, but biology also accounts for part.
  5. Nonetheless there is a real disparity in hiring, and part of it is definitely also attributable to discrimination, even if not overt discrimination.
  6. We have to look at this imbalance differently at the very top of a field from how we look at it in the field as a whole, because of some real differences in distribution between men and women [I'll come back to this below because it's important and non-obvious].
  7. There's a limit to how much any one institution can achieve alone, because the overall supply of women at the high echelons of science and technology is restricted.
  8. However, a lot of institutions (his own included) could do more to help support faculty with young children, and this would probably make them more attractive to women.
  9. Interruptions kill academic careers, and it is still the case that mothers are much more likely to interrupt their career for childrearing than fathers.
  10. We should avoid hiring quotas because they distort hiring unacceptably.

I've taken selectively from the speech, and not entirely in order, because it really wasn't a well-ordered and structured argument. I would definitely encourage you to read the thing yourself rather than just accepting my summary as gospel.

Anyway, I want to come back to a couple of these points. I'll start with #6 because I think it's particularly easy to misunderstand, and in the speech he got sidetracked with some foolish statistical handwaving that obfuscated the point. Last week Pinker also spoke about this, and he explained it rather better. The thing is that any one measure of 'intelligence' or 'aptitude' is inherently flawed, and often the only thing that a particular measure can be shown to reliably predict is performance on that measure. This makes any individual measure useless, and when a broad basket of 'intelligence measures' are compared, we get the neither shocking nor particularly illuminating result that men have a better mean score on some, and women on others. However, across a great diversity of such measures, as well as measures of behaviour, there is a consistent difference in standard deviations [note: I don't have a proper reference to give for this, but it was cited by Pinker who does generally know what he's talking about]. Specifically, men have a larger standard deviation than women. This means that even if the mean of the two groups is exactly the same (and in the absence of any serious evidence to the contrary, we should assume it is so for aptitude at science), the distrubitions won't be. At either extreme, there will be more men.

This shouldn't have any consequences at all for the overall proportion of men and women in any field, because the distribution is symmetrical and there will be the same number of men as women with better-than-average ability. However, when it comes to the pinnacle of a field—such as tenured professors at Harvard, which was what Summers was talking about—it can be expected to make a significant difference to the talent pool available. This means that genuinely equal and fair representation would not be split 50-50, and the same is probably true of other positions of extreme prominence, like heads of state, CEOs and so on.

The thing is that this is not an endorsement of the extremely imbalanced status quo, and it only would be if it were backed up with a careful statistical study that demonstrates the difference in talent availability really being as extreme as the disparity in hiring. I don't believe it is so, it's certainly not a mainstream position to argue that it is, Pinker clearly argued that it isn't, and I don't see any evidence that Summers thinks it is. In fact, in the same controversial speech Summers did address several sources of discrimination that he believes affect women trying to get into science and engineering academia, and described these as something that absolutely, vigorously needs to be combated. There is a world of difference between arguing that an enforced 50/50 division wouldn't be fair, and arguing that there is no unfairness in the status quo.


Trackback URL for this entry is: What Summers did wrong
Excerpt: Having written all that I did in defence of Summers' argument, I still don't have a great deal of sympathy for him. He did make several very important mistakes, which are things he should know better than to do, and go some way towards explaining the d...
Weblog: Research in Progress
Tracked: March 21, 2005 12:32 PM What should we teach?
Excerpt: I tend to be one of those ‘glass-half-full’ kind of people. Maybe it is because of my fundamental sense of...
Weblog: Mano Singham's Web Journal
Tracked: March 22, 2005 08:57 AM


Having read the transcript, the thing that annoys me the most about this speech is that he is standing up and presenting these thoughts as though they are somehow much more special and valid because he has had them. And when some privileged white guy stands up and does that and starts talking about women in the sciences as though every thought (and he didn't present a whole lot here that was novel to me) were some huge revelation, it's going to annoy me right out of the gate.

And THEN he started in with the stupid anecdotes. The one about the number of female Harvard Business School grads who are still working full-time after 10 years (3 out of a class of 22) was more insulting to me- as though he were insinuating that women only go to Harvard Business School to find a man and have babies later. There are quite a few reasons (including "I sold my company for millions of dollars and retired early") why a Harvard Business School grad might not work full-time.

Posted: March 21, 2005 03:21 PM

This is a very useful post. I too attended the Pinker talk and have also been following the Summers controversy. I have been thinking about these things a lot but felt that I had no real expertise to contribute and felt that I had to read more. I am really glad that Eldan has started the ball rolling on the discussion.

There is a nice book, however, for those interested in the whole issue of gender-based representation in academia and the professions. It is "Why So Slow" by Virginia Valian. It does shed an interesting light on some of the reasons for the underrepresentation but does not get into the biology issue. (At least, not as far as I can remember, and my copy of the book has been borrowed so I cannot check that.)

Posted: March 22, 2005 08:52 AM

One thing I just thought of in rereading this: I believe that the overrepresentation of men in the upper bit of the IQ distribution is too minor to explain the dearth of women heads of state, simply because heads of state don't tend to be all that smart. I think I heard an estimate of 115 or 120 for average presidential IQs in the US (though I can't actually back it up with a citation). The overpresence of men doesn't start until probably 2-3 standard deviations out (130-145), if memory serves. This may be true of CEOs, etc. as well. I'm just not sure rising to leadership positions takes the kind of intelligence that shows up on IQ tests - at least, not in the same measure as does science.

I'm more prepared to buy the argument where science is concerned, but there again I suspect the average IQ in a Harvard physics department would be lower than you might think (though probably above 115). I agree with Becky that factors other than raw intelligence contribute to getting the science positions everyone salivates over. In my department (one of the better departments in the nation), all the people who get invited to give job talks are smart and do good research - the ones that get hired also have to give a good presentation of the research, and not be married to someone we don't like who also wants a job.

Posted: March 23, 2005 08:37 AM

Becky: all the things you point out definitely make a big difference, and I think the networking aspects of them are relevant here, because it's part of how any gross imbalance perpetuates itself. People tend to hang out more with people like themselves, and it's much easier to get into collaborations and get a feel for what the rest of your department is interested in if you're hanging out with them outside strictly academic settings like seminars.

Melinda: I reckon [and this is trying to mind-read someone, so take it with a pinch of salt] that Summers saw himself as presenting things which had been overlooked by the conference so far. As such I don't think it was without value, but I also just don't think he was the one to do it. Not because he's a "privileged white male"—I really don't think that's relevent to the general points, though it does make arguing by anecdote even weaker—but because he was talking so far out of his expertise. Had they invited, say, Pinker to speak, several of the same points would have been worth him mentioning, but I daresay Pinker would have done a better job of it.

Erin: you are right about where the differences in distribution start to count. However, my understanding is that this effect applies to most measures of ability, not only IQ. Take this with a pinch of salt again, because I'm reporting secondhand information without proper references. If it does turn out to be only IQ, then what I wrote in the post is irrelevant because IQ is such a limited measure.

Posted: March 23, 2005 09:22 AM

I should probably clarify my own comment: in the last sentence above I meant irrelevant to any field, not just to global politics.

Posted: March 23, 2005 09:59 AM

It seems that we agree- I think that if I were to phrase my initial comment better, I'd say that the talk smacks of Summers thinking he has authority to speak on this subject just because he is who he is, not because he has done significant research on the matter, or had to directly experience its consequences.

Posted: March 23, 2005 11:19 AM

Post a comment