There’s no evidence that voters choose presidential candidates based on their looks

Jonathan Chait writes that the most important aspect of a presidential candidate is “political talent”:

Republicans have generally understood that an agenda tilted toward the desires of the powerful requires a skilled frontman who can pitch Middle America. Favorite character types include jocks, movie stars, folksy Texans and war heroes. . . . [But the frontrunners for the 2012 Republican nomination] make Michael Dukakis look like John F. Kennedy. They are qualified enough to serve as president, but wildly unqualified to run for president. . . . [Mitch] Daniels’s drawbacks begin — but by no means end — with his lack of height, hair and charisma. . . . [Jeb Bush] suffers from an inherent branding challenge [because of his last name]. . . . [Chris] Christie . . . doesn’t cut a trim figure and who specializes in verbally abusing his constituents. . . . [Haley] Barbour is the comic embodiment of his party’s most negative stereotypes. A Barbour nomination would be the rough equivalent of the Democrats’ nominating Howard Dean, if Dean also happened to be a draft-dodging transsexual owner of a vegan food co-op.

Chait continues:

The impulse to envision one of these figures as a frontman represents a category error. These are the kind of people you want advising the president behind the scenes; these are not the people you put in front of the camera. The presidential candidate is the star of a television show about a tall, attractive person who can be seen donning hard hats, nodding at the advice of military commanders and gazing off into the future.

Geddit? Mike Dukakis was short, ethnic-looking, and didn’t look good in a tank. (He did his military service in peacetime.) And did I mention that his middle name was Stanley? Who would vote for such a jerk?

All I can say is that Dukakis performed about as well in 1988 as would be predicted from the economy at the time. Here’s a graph based on Doug Hibbs’s model:


Sorry, but I don’t think the Democrats would’ve won the 1988 presidential election even if they’d had Burt Reynolds at the top of the ticket. And, remember, George H. W. Bush was widely considered to be a wimp and a poser until he up and won the election. Conversely, had Dukakis won (which he probably would’ve, had the economy been slumping that year), I think we’d be hearing about how he was a savvy low-key cool dude.

Let me go on a bit more about the 1988 election.

Suppose it’s true, as Chait believes, that Americans want their Presidents to look like Clint Eastwood rather than Danny De Vito. How come Dukakis was way ahead in the polls at the start of the general election campaign? The starting point is when people have the least information, when they’re the most superficial. It was by the end of the campaign, at which point voters focused more on party and ideology (see this article) and learned more about the candidates’ ideologies and issue positions, that they decided to go for the preppie from Connecticut over the wimp from Massachusetts.

To political scientists, this perspective–presidential elections turn on issues and the economy, not on charisma superficial perceptions of the candidates–is not new. Steven Rosenstone made the argument in his classic 1983 book, Forecasting Presidential Elections. And the political-science view of presidential campaigns has been gaining ground among knowledgeable reporters as well. Unfortunately it hasn’t made its way to the New York Times Magazine yet, but give it time.

My goal is not to mock

Chait is making an understandable error. He’s close to the action and focuses on the details of the candidates. And candidate effects are complicated. His article concludes:

In an old “Simpsons” episode, the unlikable brainiac Artie Ziff is elected prom king. “Instead of voting for some athletic hero or a pretty boy, you have elected me, your intellectual superior, as your king,” he says. “Good for you!” It’s funny because it hardly ever happens in real life.

Fortunately (or unfortunately), politics is not like the school prom. In the general election for president, the candidates are well-financed, are clearly distinguishable in ideology, and there are only two of them–thus none of the instability, associated with strategic voting, that we see in the primaries.

I don’t know what’s gonna happen in 2012, but political science research suggests that the Republicans could nominate a goofy short guy with glasses, or a rude fat guy, or whatever, and it wouldn’t make much of a difference. (Haley Barbour is a different story: a conservative from Mississippi could be far enough from the national mainstream to get hurt on ideology. But even then we’re talking a percentage point or two.)

It really irritates me when pundits trivialize politics and insult the voters. I’m sure Chait means well and, yes, I know that most voters don’t know anything about the federal budget, probably half of them can’t find Miami on a map, etc. But there’s no evidence that people vote based on candidates’ looks. Certainly not in presidential elections where the stakes are high and their party identification is clear.

If you want to rail at the mistakes voters make and the problems with our political system . . . fine, go for it! There’s a lot to complain about. But please please please please please don’t slam the voters for something they don’t do.

Here’s a rule of thumb. When thinking about “the voters,” think a bit about yourself. Do you vote based on a candidate’s looks? No. So why are you so so so sure that the ordinary undecided voters is doing so? Sure, many undecideds know less about politics than you do. But if they went out to vote, they might have some preferences. To think that they’re voting based on looks is just silly. And, more to the point, it’s not borne out by the data.

Check out Hibbs’s graph above.

Chait also pulls out this line:

A series of experiments has shown that subjects, even young children, can reliably pick the winners of races based solely on candidate photos.

No no no no no! As I wrote a couple years ago about a study that claimed an impressive 70% accuracy in predicting winners based on their looks:

It’s a funny result: at first it seems impressive–70% accuracy!–but then again it’s not so impressive given that you can predict something on the order of 90% of races just based on incumbency and the partisan preferences of the voters in the states and districts. I can’t be sure what’s happening here, but one possibility is that the more serious candidates (the ones we know are going to win anyway) are more attractive.

I’m not saying that the study that David Brooks Jonathan Chait is citing is wrong, exactly, but I don’t think it provides evidence that Mitch Daniels would be dead meat in the presidential election. What matters is the economy.

Why this annoys me so much

There’s some political science research on the importance of the fundamentals in presidential elections. But that’s pretty obscure stuff. You can’t very well expect a political pundit to be reading back issues of the British Journal of Political Science (yep, that’s where our article appeared, even though it was all about the U.S. We first submitted it to the American Political Science Review but they rejected it. Too many graphs and not enough tables, I think.) So, sure, I can’t blame Chait for not being up on the research consensus. And, as I wrote above, I’m sure he means well. In this case he’s trying to give Republicans the advice to nominate a pretty-boy rather than someone serious. OK, fine. But making a mistake that simultaneously (1) insults the voters, (2) mocks those political insiders who favor substance over style, and (3) brings up the old politics-as-high-school analogy, but this time in all seriousness . . . that’s just annoying.

Chait’s jumping the snark on this one, and I’m sad to see it in my local paper. Especially when so much real politics is going on right now!

13 thoughts on “There’s no evidence that voters choose presidential candidates based on their looks

  1. There's a different angle: in high-level US politics, there aren't any bad looking people. All the candidates you mention are good looking, charismatic, well-dressed and on the best possible behaviour. Any comparison on looks becomes basically pointless.

    Why would this be? A presidential candidate – or candidate candidate – is not facing their first race or popularity contest. They're all in the race because they've won numerous contests earlier on in their careers; public elections; internal party contests; competition for donors and endorsements and whatever. The people showing up in your presidential elections are already so thoroughly filtered there is very little to distinguish them as far as looks or behaviour is concerned.

  2. It's the curse of Freakonomics: writers and readers love cute stories, whether or not they have any factual basis. Even true for me: I thought Freak was a good read, and it sure generated interest in economics.

  3. And 4) by mentioning the owner of a co-op gives the impression that he knows nothing of any type of business but bog-standard capitalism.

    I'm European and so, by US political standards, a raving socialist, but how can I trust any of his statements?

  4. gelman: could you please how you would go about showing that the 70% effect is not meaningful? i re-read your post from earlier, and i still don't understand what they could have done to convince you. i imagine you will say something like: fit a multi-level model including incumbency and economic growth, but i can't see exactly what you'd do, and i'm curious. much obliged….

  5. Merijn makes an excellent point and I think you miss the point here somewhat. Voters don't ask themselves, "hmm, who is the better looking of the two candidates". No, this goes on subconsciously without realizing they trust better looking people more, or believe they are more intelligent. There is the story (don't know if it's true) that those who heard the Nixon-Kennedy debate on the radio generally felt Nixon won, but those who watched it on TV felt the opposite.

  6. Joshua:

    What those studies found, basically is that candidates who win elections have faces that look like those of more serious politicians. For most elections, it's pretty clear ahead of time who will win and who will lose, for reasons such as party ID that have nothing to do with looks. It is interesting that more serious candidates look more seriously–this is a real finding, it didn't have to be this way–but it says very little if anything about voter decisions.

    Particularly in the general election for president, when the two candidates are clearly distinct in ideology and party identification, I don't see looks as being important.

    In answer to your question: if I were studying the impact of appearance on voting, I'd do a project on the Ansolabehere-Snyder scale, looking at lots of elections, not just general elections for president.


    Given the evidence of Hibbs, Rosenstone, etc., I think the burden of proof is on the people who want to show that candidates' looks do matter. Regarding the Nixon-Kennedy debate, see here.

  7. A similar question has been studied using European data by Poutvaara and coauthors. They find a small but significant positive effect of beauty on electoral success. Authors call it a causal effect.

    I would be very surprised if the same effect wasn't present in US elections, even though finding it might prove more difficult.

  8. Someone should do a study asking voters to choose between pictures of candidates taken years before the race they were about to compete in. Perhaps ones taken after their previous victory. Maybe focus on open elections without an incumbent.

    I'm hoping the next president has no hair on top of his head and plenty on his face. Maybe an old Amish dude. Those Amish must be doing something right.

  9. thanks gelman, two follow-up questions:

    1) what is the Ansolabehere-Snyder scale? a quick google search yielded nothing obvious.

    2) perhaps one could control for "seriousness" or incumbency by only comparing people with similar scores on those metrics? perhaps this is the same as the AS scale?

    thanks again

  10. Joshua:

    By Ansolabehere-Snyder scale, I meant not a scale of measurement but their scale of operations. When studying the incumbency advantage, Ansolabehere and Snyder looked at lots and lots of elections including many statewide offices as well as Congress. Basically, I meant that they did a comprehensive analysis of all sorts of races.

  11. Trivia note: Artie Ziff was voiced by Jon Lovitz (another Jonathan), who also took care of the Dukakis impressions on SNL. Probably not a coincidence that Chait thought of that Simpsons reference while he was thinking about a caricature of Michael Dukakis.

Comments are closed.