The ghost of Larry Summers (I know! And he isn’t even dead yet!) has risen again, with John Tierney of the NY Times “daring” to consider the notion that maybe women aren’t as mathy as men. There’s a lot to object to in his story, from the title (Sorry, John, but it isn’t daring to promote a stereotype at all) to the feeble caveat at the end, where he says he willing to consider “possible social bias against women” in the sciences. “Possible”? Really? Say it ain’t so, John!
But no, let’s cut straight to the heart of the issue. The problem here is sneaky sleight of hand.
Here’s what everyone in society, in academia, and in the sciences wants: we want to employ the best people with the best aptitude for the job, and with the greatest possibility of success. The universities want to hire people who are really, really good at science. Agreed?
Now here’s the problem: there is no clear marker or metric for success in science. It’s a complicated task, with lots of variables and lots of different strategies for doing well. It’s not like looking for the person who runs the 100 meter dash the fastest, in which we could just line up the applicants, fire a starting gun, and give the job to the first person who crosses the finishing line. So what do we do? We use proxy metrics.
The best proxies are measurements that most closely approximate performance in science. We look at publication records, grants awarded, recommendations of colleagues, the sort of thing we’d expect our new scientist to continue doing. It’s not perfect — maybe the applicant is a neurotic living on the edge who’s about to break down, or maybe they have an abrasive personality that will affect the performance of other faculty — but it’s a good start. It’s what most committees should evaluate most highly in the hiring process.
There are other proxies, too. Did they get good grades in their college courses? That indicates some discipline. In their teaching, did they get good student evaluations? Student evals are fraught with problems, but an unbroken record of negatives is a warning sign. Do they score well in IQ tests, SATs, GREs? That’s a proxy, too. It would indicate that they’re pretty smart, which is an extremely important property if you’re going to be a scientist.
All of those things are still just proxies for the constellation of properties you want in a scientific colleague. We have to balance them to get an idea of the potential of an applicant: it would be insane to hire someone with no experience, no publications, and no grants just because they got straight As in high school and college. But for some reason, in this tedious argument about the suitability of women to do science, all that gets mentioned is a gender difference in performance on standardized tests.
Even if we concede a genuine gender difference in performance on standardized math tests that is independent of social factors (which I don’t yet), gender is a proxy for intelligence (and a very poor proxy, too), which is a proxy for scientific aptitude. We’re getting pretty damned far from actual substance of the job requirements.
So I have two proposals, both of which still use the handy shortcut of a simple numeric proxy which the advocates of these ideas favor beyond all reason, but additionally, get away from these inflammatory, socially loaded issues of gender and race (let’s not even get into that one, but skin color is another proxy used to estimate intelligence). It might help defuse the tension that talking about judging people on their sex always causes if we simply used a different proxy.
Let’s just use a different indirect metric; I suggest wealth. We already know that this one works out surprisingly well, as this chart shows.
Obviously, rich people are inherently smarter than poor people. Tierney points out that the right tail of the SAT math test distribution has about a 3:1 boy:girl difference; I wouldn’t be at all surprised to learn, though, that the rich:poor difference is even greater.
Tierney has a wonderful quote in his article. The sex ratio at the right end of the distribution hasn’t been changing much, so he reports that
The Duke researchers report in Intelligence, “Our data clearly show that there are sex differences in cognitive abilities in the extreme right tail, with some favoring males and some favoring females.”
The researchers say it’s impossible to predict how long these math and science gender gaps will last. But given the gaps’ stability for two decades, the researchers conclude, “Thus, sex differences in abilities in the extreme right tail should not be dismissed as no longer part of the explanation for the dearth of women in math-intensive fields of science.”
[By the way, that double-negative in the sentence is hopelessly confusing — it should mean that sex differences should be dismissed as part of the explanation, but in the context they’re saying exactly the opposite. Must have been written by a man, with their poorer verbal skills.]
By the same reasoning, we can also argue that wealth differences in abilities should not be dismissed, since they tend to be perpetuated over many generations. We can just stop wasting time and money trying to educate poor children or correcting the inequities of poverty in our schools, because the data clearly says that it’s highly unlikely that any of them will succeed in science.
So here’s my specific proposal: every scientist should report on their CV the approximate amount of money their parents were making while they were attending college. It’s a simple, single number with a wide range, allowing us to easily place everyone on a scale of potential performance. If you come from parents on the left side of that chart, you are less likely to be a competent scientist, and you should admit that fact; if you’re on the right side, employers ought to be able to use the information that you’ve had definite advantages and a leg up on the job.
Wait — we’re still using a proxy for a proxy. Let’s cut straight to it and use SAT/GRE scores directly. Forget everything else, let male and female faculty report their scores right on the CV, and we’ll sort them out for matters of tenure, promotion, rank, etc. right from the value being argued over.
You see, there’s a shifty little game that proponents of gender discrimination are playing. They argue that high SAT scores are indicative of success in science, and then they say that males tend to have higher math SAT scores, and therefore it is OK to encourage more men in the higher ranks of science careers…but they never get around to saying what their SAT scores were. Larry Summers could smugly lecture to a bunch of accomplished women about how men and women were different and having testicles helps you do science, but his message really was “I have an intellectual edge over you because some men are incredibly smart, and I am a man”, which is a logical fallacy. Even if we accept his premise, we don’t know that any individual man is smarter than any individual woman — unless we get full disclosure. It’s as if I went up before a WNBA team and lectured them on how men were on average taller and stronger than women, and therefore play a better game of basketball, and didn’t have to do a little one-on-one on the court — where I’d be humiliated despite my membership in the testosterone club.
If these scores are really so important, let’s go for it and rank scientists work by their math SAT scores. The NIH can use it to prescreen grant applications — those from scientists with scores below 750 go in a pity pile for funding if there’s left-over money, those with scores over 750 get priority ranked by the usual methods, with the math SAT used as tie-breaker for applications on the edge of the funding level. We’ll resolve scientific debates that way, too: for instance, isn’t the one thing you need to know to figure out what side of the group selection debate you should be on is the relative SAT scores of David Sloan Wilson and George C. Williams? Why aren’t these numbers available?
The real advantage, though, is amusement. Suddenly all the men who had been arguing that being in the elite top 0.001% of was so essential to great scientific success, but who are not themselves quite that high, would find themselves arguing that science is an enterprise with many parameters and no single simple number can encapsulate the entirety of the process, and say, shouldn’t you all be looking at my publication record, my grants, my contributions to scholarly discussions?
Lest you think I’m being self-serving here, I will admit that under proposal 1, I’d have to get demoted — my parents’ economic status was way, way over to the left. I’d do much better under proposal 2, because I’ve always done phenomenally well on standardized tests. Either way, I don’t care, and if either of my schemes were actually implemented I’d be arguing against them, anyway.
The problem is fundamentally one of hitch-hiking on others’ reputations. We get these waves of articles touting the statistical superiority of males because some people want their club, the Men’s Club, to have that prestige of being better than the Women’s Club, despite the fact that their individual performance may not be better than the performance of individuals in that other, ‘inferior’ group. “Men” is a proud and meaningless association of human beings — it is a granfalloon. Seriously, the fact that Stephen Hawking happens to be in Club XY with me does not in any way bestow upon me the intellectual luster of Hawking. Nor are Carolyn Porco, Lisa Randall, Shirley Ann Jackson, or Pardis Sabeti somehow less likely to succeed in science because they don’t have a Hawking-like penis. Yet somehow we end up going around and around this irrelevant argument about the statistics of a granfalloon all the time.