September 17, 2014 4 Comments
A friend last week sent around a link to the new article nicely summarized here which prompted much joking on the matter:
A new study from the American Journal of Political Scienceindicates that different political affiliations may actually correspond with different body odors.
The researchers, led by Brown University political scientist Rose McDermott, found that conservatives and liberals smell dissimilar. While the difference is small, it is apparently significant enough that we subconsciously prefer the scent of those who vote like we do. “It appears nature stacks the deck to make politically similar partners more attractive to each other in unconscious ways,” the researchers wrote.
The how is here. It’s not quite fair to say that someone smells conservative or liberal. More so, like attracts like:
They got 146 participants to rate the attractiveness of the body odor of unknown strong liberals and strong conservatives, without ever seeing the individuals whose smells they were evaluating. Based on that, they concluded that people find the smell of others with similar political opinions to be attractive, suggesting that one of the reasons why so many spouses share similar political views is because they were initially and subconsciously attracted to each other’s body odor.
“People could not predict the political ideology of others by smell if you asked them, but they differentially found the smell of those who aligned with them more attractive. So I believe smell conveys important information about long-term affinity in political ideology that becomes incorporated into a key component of subconscious attraction,” said Dr. Rose McDermott, lead author of the paper.
I actually looked through the original article and I’m going to have throw some cold water on this. The support for the hypothesis touted far and wide (lots of links to this story– okay, including me) was significant at p<.1, one-tailed.
In both models, we observe the hypothesized positive coefficient on the negative absolute difference in ideology scores (–Abs. Ideology Diff.), though in both cases the coefficient is less precisely estimated (t = 1.48 in Model 1 and t = 1.45 in Model 2), but still with one-sided p-values less than 0.1. In all cases, the substantive effect of ideological similarity is small, which is to be expected.
How in the world such marginal results got published in the 2nd most prestigious journal is beyond me. Peer review is supposed to be blind, but I strongly suspect that having prestigious scholars behind the research had something to do with it. “the coefficient is less precisely estimated”!? I’m going to have to try that next time I don’t get the statistical significance I want.
Honestly, if the best result I got was p<.1, one-tailed on my key hypothesis, I don’t think I’d send to any journal. And if I only had 146 subjects, the first thing I would do is get more to increase the statistical power in hopes of having more compelling statistical significance. Anyway, this is also a great example of how the seal of approval from a major peer-reviewed journal means lots of wide-spread attention when the group of 7 political scientists I discussed this with last week over lunch who had seen the actual article were all skeptical. At least makes for some good political humor.