Matthew Atkinson, Ryan Enos, and Seth Hill sent along this paper:

Recent research finds that inferences from candidate faces predict aggregate vote margins. Many have concluded this to mean that voters choose the candidate with the better face. We implement a survey with participant evaluations of over 167,000 candidate face pairings. Through regression analysis using individual- and district-level vote data we find that the face-vote correlation is explained by a relationship between candidate faces, incumbency, and district partisanship. We argue that the face-vote correlation is not just the product of simple voter reactions to faces, but also of party and candidate behavior that affects which candidates compete in which contests.

This is great stuff. They’re talking about a 2005 article by Alexander Todorov, Anesu Mandisodza, Amir Goren, and Crystal Hall which found that people thought the faces of winning congressional candidates looked more “competent” than faces of losing candidates. I wrote about this about a year ago and expressed skepticism about the interpretation of those findings . . .

It seemed likely that the more competent-looking candidates were more likely to be the ones that were more credible candidates for political reasons that were not directly related to looks (incumbency, ability to raise money, etc). In short, I suspected that, even if the voters had no idea what the candidates looked like, the Todorov et al. findings could still occur. Well, Atkinson et al. didn’t just speculate, they went and did a bunch of analyses. They find,

Because many congressional contests in the United States are not competitive and because candidates with high competence are more likely to enter the contests in which they have a reasonable chance of success, we find high competence candidates defeating their usually low competence challengers in the majority of contests. This dynamic produces a high correlation between facial competence and election outcomes. That candidate faces are not distributed randomly across contests and that it is likely that parties and candidates are making decisions which affect the allocation of candidates to races suggests avenues for future research. . . . Some of the media attention surrounding the research by Todorov et al. (2005) was probably generated by the sense that the finding demonstrates that the voting public is uninformed. We have demonstrated that appearance plays a much smaller role in election outcomes than one might infer from a casual reading of Todorov et al. (2005) or its representation in the media.

I like the little things about the paper too, such as the graphical displays of inferences. (I don’t like the tables so much–I write entire empirical research articles with no tables at all–but I guess that adding graphs is the first, key, step. Removing the tables can come later.) And they make a good choice by modeling incumbent vote share rather than simply modeling the binary win/lose outcome, which would discard information. I also like that the authors explicitly discuss the media reports of the Todorov et al. research.

Finally, I feel a little awkward saying this, but I think they should refer to my blog entry from last year, since it’s the first publication that I know of that questioned the “faces decide elections” reasoning. Even though Atkinson, Enos, and Hill probably came up with their ideas on their own and only encountered my blog entry later (as noted above, they went far beyond my speculations and did actual research), it would still be appropriate to cite it as relevant early work.

More thoughts on the least-important part of what I wrote above

P.S. Henry Farrell linked to the above and pointed out, correctly in my opinion, that a typical blog entry such as mine (a link with some quick discussion and not much follow-through) falls somewhere between an offhand comment (which can be cited in acknowledgments) and a published article. Atkinson, Enos, and Hill were aware of my blog entry (that’s why they sent me their article) but I don’t really know how my thoughts fit into their work.

Let me emphasize that I liked the Atkinson, Enos, and Hill paper and very much appreciate that they sent it to me. If they want to cite my blog entry, that’s fine, and if they don’t, I bear them no ill-will. I agree with some of the commenters to Henry’s blog post, who said that, if Atkinson et al.’s work preceded my blog entry, then it’s a pretty minor point that somebody else (in this case, me) noted something similar. If Atkinson et al. came up with the idea independently and did all the work, then they clearly deserve all the credit.

I certainly wouldn’t want for a blog entry to have any kind of intimidating effect, where I’m implying that so-and-so should definitely cite me. It’s really up to the authors of the article to decide how a brief blog entry fits into the existing literature. To me it seemed relevant as the first “published” criticism of the Todorov et al. paper, but I’m completely ignorant of the literature so I’d trust Atkinson et al. much more than me on this point.