Today is apparent "disagree with patrissimo day" for me. Rather than argue more with his assertions that you're as bad as Michael Vick if you eat meat (1, 2, 3), I'd like to talk about his much more interesting post in which he talks about Roy Baumeister's talk to the APA:

Is there anything good about men?

Men go to extremes more than women. Stereotypes are sustained by confirmation bias. Want to think men are better than women? Then look at the top, the heroes, the inventors, the philanthropists, and so on. Want to think women are better than men? Then look at the bottom, the criminals, the junkies, the losers.

In an important sense, men really are better AND worse than women.

A pattern of more men at both extremes can create all sorts of misleading conclusions and other statistical mischief. To illustrate, let’s assume that men and women are on average exactly equal in every relevant respect, but more men at both extremes. If you then measure things that are bounded at one end, it screws up the data to make men and women seem significantly different.

Consider grade point average in college. Thanks to grade inflation, most students now get A’s and B’s, but a few range all the way down to F. With that kind of low ceiling, the high-achieving males cannot pull up the male average, but the loser males will pull it down. The result will be that women will get higher average grades than men — again despite no difference in average quality of work.

The opposite result comes with salaries. There is a minimum wage but no maximum. Hence the high-achieving men can pull the male average up while the low-achieving ones can’t pull it down. The result? Men will get higher average salaries than women, even if there is no average difference on any relevant input.

Today, sure enough, women get higher college grades but lower salaries than men. There is much discussion about what all this means and what should be done about it. But as you see, both facts could be just a statistical quirk stemming from male extremity.

This debate about why there's still inequity in various sectors of society is both fascinating and fraught with difficulty. On one side, we have the people who argue that differences in distribution are evidence of discrimination, either overt or unconscious, against women (call this the cultural explanation). On the other side, people seem to believe that any remaining differences are explained by evolutionary psychology or biological differences between the sexes (the biological explanation).

From a historical bayesian perspective, the cultural camp looks like clear favorites. Women (everywhere) and minorities (in Europe and China [note that Chinese are not minorities in China] and North America, at a minimum) have held much worse jobs and had many fewer opportunities over the last, say, 2000 years at least. Over that time, there have been many arguments about why that was the natural order of things, from phrenology to religious justification, with theories from the leading scientists of the day -- the explanations were endless. And from a modern perspective, it seems very clear that the differences in job/role distribution had little to do with different fundamental aptitudes and a huge amount to do with culture. One major piece of evidence for this is to look at women in our society now: they are succeeding at many things that would have been incomprehensible in ages past. Even things like teaching, which we think of women as being good at now, were forbidden to women in England only a century or two ago.

Historically, societal changes have been slow. In historical terms, overt discrimination happened very recently; the Civil Rights Act was enacted in 1964, and things like sexual harassment were not ruled illegal until the late 70's. There was overt, institutionalized gender and race discrimination less than 50 years ago in this country. It seems a little naive to think that those ingrained attitudes have been completely erased in the intervening few decades.

However, just because the discrimination argument has basically always been right historically, and the natural distribution argument has always been wrong, doesn't necessarily mean that the natural distribution argument has no truth to it. But the bar for the biological explanation should be significantly higher than the bar for the cultural explanation.

In any field of science, if a theory has been tested in many different situations and supported in those tests, then evidence challenging the theory needs to be very strong before scientists will reject the theory. This is because many or most experiments and studies have flaws of one kind or another, and if a theory is well-tested, it is more likely that new research has a flaw than it is that the theory is wrong. It may well be that the new evidence is correct and damning, but the bar is clearly higher, as it should be. I think the same should apply for biological explanations of differing roles for gender in society, where the tested theory is that of cultural explanations.

And on the other side, there is good evidence for cultural causes of things like gender differences in performance on a math test via an effect called stereotype threat, and studies that show the evaluation of average professorial resumes changed based on whether the name on the resume was male or female. If those are accurate, and at least stereotype threat is duplicated and supported in many repeat experiments, then cultural explanations account for at least some of the gap between men and women in performance.

So I am skeptical of claims, even intelligent-seeming claims, that the reason there are more men in top positions in our society is mostly because of biological effects like greater variation. It certainly sounds plausible, but there have been many, many plausible explanations for why women should accept worse roles and outcomes than men which have essentially all turned out to be bogus and self-serving. Plausibility is not enough in this debate, given the history.

There has long been a campaign, based in the bizarre feminist notion that women should be treated as well as men, that there should be "equal pay for equal work." It's a widely quoted statistic that women earn seventy three cents for every dollar that men earn.

The standard reply is that much of that is because of differing choices that women make -- that on average women choose less lucrative careers, major in less lucrative subjects, and generally devote less time to their careers and more time to actual, you know, life, compared to men.

What I've never seen is a good study that actually tried to analyze all of those different factors. The American Association of University Women just did such a study, where they took all of factors that are known to affect wages -- degree, time in the workforce, career choices, marital status, children, and so on, and ran a giant regression. All those factors explained most of the pay gap, but accounting for them, women made $.95 for every dollar that men made. The authors attribute that to gender discrimination.

There's a long, interesting section about how to address the gap. The suggestions are:

• Promote careers in science and mathematics to girls. This means making them more female-friendly, not just trying to convince skeptical girls that math is fun. It turns out that taking advanced math classes in high school makes you more likely to take engineering and math classes in college, so that might be an easy step to take.
  
• Encourage women to negotiate for better jobs and pay. Men negotiate much more aggressively for better pay, have higher expectations, and are more likely to apply to and get jobs that they are not very qualified for. However, women are viewed negatively in the workplace if they are too assertive, so it's not just a matter of telling women to negotiate more -- the cultural context has to allow for it in a way that it does not right now. Negotiate hard and get fewer promotions and worse relations with coworkers, or settle for a clerical job and lower pay instead of a managerial job. Nice choice.
  
• Part time work and allowances for motherhood. I didn't know this, but apparently when men have kids they respond by working *more* hours. And they compensation goes up by more than the hours explains. Women, on the other hand, work fewer hours, and get paid less than the drop in hours would explain.
  
• Rethink using hours worked as a measure of productivity. This is a good idea in general -- I would much rather be paid for what I accomplish than the hours it takes to accomplish it, and my employer really should agree too. That would allow women to work with more flexibility and over fewer hours, which is good because women in general are less willing to just put in hours.
  

Anyway, fascinating study. While it shows that the pure gender gap is a much-more-manageable 5% instead of ~30%, it also argues convincingly (to me) that there are structural reasons that women take lower-paying jobs and careers.

A Nevada judge said in a child pornography sentencing ruling:

"These kinds of offenses are problems with impulse control," said Carson City District Judge Bill Maddox prior to sentencing. "When I say that, it's my understanding that most men are sexually attracted to young women. When I say young women I don't just mean women that ... you should be attracted to. I mean women from the time they're 1 all the way up until they're 100."

He's sort of right, in that men can be attracted to inappropriately young girls normally... but 1? I don't think that's normal. It seems to me that there is a big line that can be drawn at pubescence -- when girls develop secondary sexual characteristics which are evolutionarily designed to attract males, you might expect the males to be attracted.

But babies and toddlers? I don't think that's a case of just poor impulse control. I think it suggests that there's an underlying problem with the person in question, coupled with problems with impulse control.

At any rate, I sure don't envy the shitstorm of criticism that Mr. Maddox is about to find himself in.

Hat tip to Feministing (great logo) via Salon's Broadsheet.

What's wrong with our society exhibit A: when a fashion show organizer put healthy women (read: not anorexic) on stage instead of the standard "needs a sandwich" fare, a fashion industry cried out in disgust, including Vogue Australia's decision to delete the whole show from their website. It's not like the women were unattractive, either. Just not skeletal. You can see for yourself.

Maybe the designer in question should just move to Spain.

Exhibit B: Victorian shops, whatever those are, are selling padded bras for girls as young as 6. A spokesman spouted, "It is more about hiding what you have got than showing it off. It is certainly not there to make children look like they have breasts." Er, yeah. I'm sure that's what it is.

Girls outperformed boys in the new writing portion of the SAT. The article talks a bit about how amazing it is since girls haven't outperformed boys in any part of the SAT since the 70's, but the essay portion is new, so it's not really comparable to previous SATs.

In related news, SAT scores dropped significantly. This is even less newsworthy, since if you read the fine print:

Comparing only scores from students' first tests, math scores actually rose one point and critical reading scores fell three points.

So there was basically no change. Why the apparent drop? Because with the new writing portion, people who'd taken the test before decided not to retake it, and on average when you retake it your score goes up by 30 points. So fewer retakes = lower scores. (My score went up 140 points on the second try, possibly because I had a bet on the score the second time around.)

The really interesting report is about stereotype threat in test scores.

Strange but true: Women score much lower on math tests if they are first asked unrelated questions about gender issues. The phenomenon is called "stereotype threat" -- a kind of performance anxiety discovered in 1995 when psychologists found that black students at Stanford University did significantly worse on intelligence tests if they were first asked to identify their race on the test form.

Since then, dozens of other experiments have confirmed that subtly cuing women or minorities to think subconsciously about their sex or race causes them do poorly in areas where the stereotype suggests they are weak.

This is pretty well confirmed, and really surprising. Just asking someone what gender or race they are can affect their scores? But what's nice is that the effect goes both ways. When a group of students was asked to think about how smart they were before an exam, girls did much better.

The gender gap closed dramatically, as women's scores improved while men's stayed the same. "There was no significant difference between men and women," McGlone said. "With a pretty simple manipulation, we could significantly reduce this gap," which suggests that "there might be things that make all of these biological factors go away."

It's also strange and interesting to me that the men were much less affected by these cues. This suggests to me that women are underperforming their cognitive ability on these tests, evidently because of social gender stereotypes (though there could be something else going on as well), and when you try to reverse the effect, women perform closer to their potential. Men, on the other hand, perform closer to their ability, either because they are competitive, are less fucked up by society, or for some other reason.

This sort of result is why I think that psych tests that show systematic cognitive differences between men and women are often suspect, at least as they are used to try and show biological differences. If really minor social cues like these can cause big swings on a cognitive test, then pretty much any test that doesn't explicitly control for those things is likely to have a significant bias, and always in the direction of the social stereotype. So you'd get results that confirm stereotypes, like that women are worse at math than men. At the very least, the gap would be larger in these tests than biology itself could explain.

Lawrence Summers, president of Harvard, has resigned.

He'll be remembered for his comments that the lack of women faculty at Harvard in math and hard sciences could be explained by innate differences between the genders, but at least according to the article, that's not why he's stepping down:

The resignation caps an increasingly rancorous standoff between Summers and disaffected Harvard staff, and came a week before faculty members were to vote in the second no-confidence motion against Summers in 11 months.

...

[T]he abrupt resignation of arts and sciences dean William Kirby on January 27 deepened opposition against him. Several faculty have accused Summers of pushing Kirby out and called for his resignation at a faculty meeting this month.

Apparently mostly women are being considered to succeed him. This should make everybody happy. The camp that says that he was sexist and deserved to be fired for his comments should prefer a woman in place instead; and the camp that thinks that women are better at nurturing and communicating while men are better at math and science should be happy, since president of a university is all about the former and not at all about the latter.

Really, what grounds for objection could there be?

It turns out men report more sexual partners than women in surveys, in this case web-based.

The women reported on average 8.6 lifetime sexual partners. The men claimed 31.9.

Rather than let it go at that, Brown and his colleagues later in the survey asked the participants to rate the truthfulness of their response. About 5 percent said they lied. In addition, more than 10 percent said they knew their answer wasn't accurate.

"They gave an answer and then two minutes later admitted they had lied about the answer," Brown said.

But there's more to the discrepancy. Men and women use different methods to calculate their past dalliances.

Women rely on a raw count, a method Brown says is known to result in underestimation.

"They tend to say, 'I just know,' and if you ask them to explain how they know, they say, 'Well, there was John, Tom, etc.'"

Men also rely on a flawed strategy.

"Men are twice as likely to use rough approximation to answer the question," Brown said. "And rough approximation is a strategy known to produce over-estimation."

Mostly what this says to me is that survey-based research is much less reliable than other things. People make stuff up even in anonymous settings where there's no incentive to.

The difference in estimation methods is interesting too, but it seems unlikely to me to account for the vast discrepancy. If your true rate is 10 or 12 (based on the women's rate assuming a pretty big undercount), are you really going to report 31 through misestimation?

I may have blogged about this before, but I want to make sure I have a link to the paper, so here it is again.

This fascinating paper studies the impact of evoking stereotypes on cognitive performance.

The researchers did one of a series of studies in which they gave students tasks intended to remind them of stereotypes of groups that they belonged to -- asians are good at math, females bad, elderly wise or senile, etc) caused performance on related cognitive tasks to change to become more in line with the stereotype.

In other words, if you remind asian girls that asians are known to be good at math, they'll do better on a math test than a control group. If you remind them that girls are known to be bad at math, then they'll do worse than the control. Even more bizarrely, if you just remind them that they're asian, there will be a (smaller) boost; similarly, simply reminding them that they're girls causes a drop in math performance.

The human brain is a strange, strange machine.

patrissimo just posted about a fascinating debate on the role is science in studying gender issues in science, inspired by the recent much maligned comments by Lawrence Summers, in which he appeared to claim that part of the reason that women are underrepresented on Harvard math and science faculties is due to innate ability or preference.

Part of the debate talked about whether this was even a reasonable topic for study. In my view, it clearly is something that should be looked into deeply. I don't think there was disagreement there among the debaters or anyone who might read this blog. But they disagreed on a lot of other things.

What I found most telling was the description of an experiment on evaluating scientists. A bunch of psychology faculty were given CVs of prospective faculty, one of whom was incredible, and one of whom was solidly successful. Some of these resumes were labeled by a male name, and some by a female name. The profs were asked to rate how good a candidate the person would be.

There was no difference in evaluations of the amazing candidate. If you really shine, nobody cares if you're male or female. But:

What about the average successful vita, though: that is to say, the kind of vita that professors most often must evaluate? In that case, there were differences. The male was rated as having higher research productivity. These psychologists, Steve's and my colleagues, looked at the same number of publications and thought, "good productivity" when the name was male, and "less good productivity" when the name was female. Same thing for teaching experience. The very same list of courses was seen as good teaching experience when the name was male, and less good teaching experience when the name was female. In answer to the question would they hire the candidate, 70% said yes for the male, 45% for the female. If the decision were made by majority rule, the male would get hired and the female would not.

This sort of thing makes me think that regardless of what people find about innate differences between men and women, the composition of the Harvard faculty will not be explained solely through those differences. More than just unconcious discrimination like in this experiment is the fact that, if it is believed that such unfairnesses exist, then women will be less likely to try to get into positions for which males are preferred. If you're smart enough to make it in academia, you're smart enough to make it in some other field where you won't have extra barriers to come up against. So there's a feedback effect that means fewer qualified applicants and even fewer jobs to women.

I wish Pinker had directly responded to the above experiment. Instead, they debated on whether the SAT was a good measure, and on innate preferences of women, and other things that weren't as interesting.