Halloween Horrors: Sandy Wreaks Havoc on East Coast! Hormones Influence Political Views!Published 29 October, 2012
So, the last few days have seen horror appropriate for the Halloween season. On the one hand, there is the killer weather walloping those of us living on the Eastern Seaboard, and then on the other hand there is frenzied apoplexy over the publication of a paper suggesting that hormonal changes might influence political views.
I don’t know anything about hurricanes, so I’ll just address the other blood-curdling event. The excitement began with a post that appeared on “the chart” at CNN by Elizabeth Landau reporting on a paper in press in Psychological Science entitled “The Fluctuating Female Vote: Politics, Religion, and the Ovulatory Cycle” by Kristina Durante, Ashley Arsena, and Vladas Griskevicius. The story (the text of which can be read here, among other places) was taken down off of CNN, apparently in response to a flurry of criticism in the comments section of the post and elsewhere; in place of the article, there is now a short entry that reads in part that “some elements of the story did not meet the editorial standards of CNN.” Subsequently, others have chimed in with their views, including various members of The Blogging Class.
A scan of the comments on the page on which the story was posted indicates that people are, in a word, miffed. User Suzanne, whose views do not seem atypical, captures, I think, the flavor of the comments, writing:
This is offensive in its claims. CNN, stop covering useless information that perpetuates discrimination. The idea that any woman is “more religious” while ovulating is ridiculous. Feminine views on religion and politics don’t change every few weeks. Our values do not drift due to hormone surges. Printing this crap ought to be beneath you, but apparently, by CNN standards, this is worth covering. Shame on you.
So, yeah, a bunch of people are pretty angry.
It’s not just the commenting class that seems vexed. Huffpost ran a headline that reads, in part: “CNN Reports Women Voters Apparently Incapable Of Cognition.” The usually even-keeled Retraction Watch, even though they typically report on retractions of scientific papers, posted about this retraction from CNN. They comment:
Here’s how Durante explains this: When women are ovulating, they “feel sexier,” and therefore lean more toward liberal attitudes on abortion and marriage equality. Married women have the same hormones firing, but tend to take the opposite viewpoint on these issues, she says.
And then later in the piece:
Why estrogen would make single women feel sexy but turns married women off goes unexplained in the article — probably because the notion is, on its face, complete poppycock.
Notice, of course, they don’t quote Durante as saying that estrogen turns married women off: the claim is that the hormonal changes alter their political views differently, not their libido differently. The notion that they dismiss as “poppycock” might well be, but it also misrepresents the claim. (Can we expect a retraction from Retraction Watch?)
Anyway, what about the actual paper? The authors report two studies, the first of which is focused more narrowly on religiosity. They surveyed 275 women, and estimated where each of them was in her cycle using the “reverse cycle day” (RCD) method; they also asked about their relationship status and religiosity. They found that there was a difference between single women and women in relationships such that single women closer to ovulation as assessed by the RCD method were less religious while women in relationships showed the reverse effect. This effect was replicated in the second study (see below; also please note that it was the interaction terms here that were significant, which is what I’m referring to here.)
The study that seems to have elicited the howls, and was the focus of the CNN piece, was the second one. The researchers had 502 women take an online internet survey, roughly half of whom were single, the other half mated. They again estimated where each woman was in her cycle using the RCD method, and asked a series of questions surrounding religious, political, and social attitudes, as well as which presidential candidate, Obama or Romney, the subject would vote for if the election were today. They divided political attitudes into two types, “social” (e.g., abortion, stem cells) and “fiscal” (e.g., tax policy, wealth transfers).
I mention this last point because in one of the key analyses, they got a three-way interaction. (F(1,299) = 8.15, p = .005.) (See the Figure, but note that I’ve only shown one panel of Figure 2 here.) For social, but not political issues, reporting that one was nearer the ovulatory phase of one’s cycle was associated with more liberal attitudes for single women; for mated women, women reporting they were nearer the ovulatory phase of their cycle expressed more conservative political views.
In the key analysis on voting (using a logit), they found a significant interaction such that single women were more likely to vote for Obama when they reported they were nearer ovulation as opposed to further from it (87% versus 73%); for mated women, it was the reverse (40% versus 23%).
The bulk of the objections to this work seems to stem from the CNN article, rather than the article itself. As can be seen in the hyperbole of the HuffPost headline I mentioned above, the trope seems to be that the suggestion is that women (but implicitly not men) are swayed by their hormones on important matters such as voting, and people find this objectionable and offensive. More about that in a moment.
But first, a trickle of criticism of the work itself has begun, such as a post by Scicurious, discussing the work. Scicurious is very concerned about the method of assessing ovulation. As she indicates, this was done with a proxy measure, the self-report RCD method. This means that this measurement has some error. That’s true. But it’s also true that every measurement has error. (Yes, even if one does an assay, those measurements will have error. Less, sure, but all measurements have error.) The researchers could have brought women into a lab to do this work, but in that case I’m guessing that they could have run far fewer subjects. There are always tradeoffs in study design, and here they chose a larger N over a better measurement. They are not the first researchers to do so, and won’t be the last. (One might argue that measurement error makes it harder to get the predicted effect – especially a three-way interaction – but, still, it’s true that the technique is an indirect assay of ovulation. This argument would be more persuasive if it could explain why measurement error would yield the systematic results observed, which I do not believe it can.)
Second, Scicurious worries about the extent to which the single/mated groups differ. Yes, the single women were, on average, for instance, four years younger than their counterparts in relationships and were, unsurprisingly, less likely to have children. True, they didn’t put these into the reported analyses, but, of course, if it turns out that, for example, the findings are driven by age, then it seems to me that this pattern too would require an explanation.
Finally, yes, it’s a cross-sectional design, and that carries with it all the advantages and disadvantages of cross-sectional designs. Getting the same women at two different points would have been just delightful. Every design choice carries tradeoffs, but pointing out that there are advantages to longitudinal design is not, to my way of thinking, fatal to the research.
In any case, like all studies, this one has limitations. To my eye, none of the criticisms point to an alternative explanation for the data. Is the argument that the data are due to chance, and it’s all spurious? Maybe, but then it seems a bit weird that they replicate the religiosity effect in the first study so nicely. If it’s all just noise, then why do they see the same results in two datasets? But, sure, any finding could be due to chance, and of course shortcomings ought to be borne in mind.
Anyway, my sense is that the reason people got so irate on CNN’s web site had little to do with the issues of measurement error using the reverse cycle day method or whether they should have used age as a covariate in the general linear model. My sense is that the objection has to do with the perception that this work is endorsing or perpetuating a stereotype in which women, but not men, are influenced by their hormones, limiting, or undermining, their rationality.
I reject such a view. First of all, because this is a study on voting behavior, there’s no right answer. The claim here isn’t that hormones are affecting rationality – getting the right answer – but rather preferences. (Not only that, but the claim is that these preference shifts are adaptively rational, but that’s somewhat beside the point.) Second, the study didn’t compare women and men’s behavior because the question at stake was about women. I studied the paper fairly carefully (though I haven’t seen the Supplemental Materials), and I didn’t see any claim that hormones don’t affect men’s behavior and decision making (which would, in the words of Retraction Watch, be poppycock). Hormonal changes affect both men’s and women’s behavior. (As support for the idea that there is a link between hormones and behavior, I offer as evidence pretty much every single issue of the journal Hormones and Behavior.)
The facile riposte to the worry that science perpetuates stereotypes is to say that we should face the truth, whatever it may be. In this case, the finding doesn’t seem to me to be all that unpleasant. So women’s attitudes change in systematic, textured ways as a function of their cycle. So? My sports team loyalty shifts in systematic ways as a function of which city I’m in and how much I’ve been drinking. (I turn into a Dolphins fan after a few nostalgia-inducing beers; I’m a hometown Eagles fan when I’m sober.) I don’t feel that inconsistency in attitudes, preferences or fandom is such a horrible thing. I might be biased on this particular issue, but such inconsistencies are part and parcel of the fact that the human design is complicated, and part of that complexity is that we all dance to the tune played by our hormones, and still more distally by the forces – weather, location, time – that conduct our hormonal orchestra.
Yes, the design here has some limitations and, yes, maybe some people would prefer it if we could all think of our ideologies as constant as the North Star. But I hardly think it brings shame on CNN to report that they might not be.