There has been extensive psychological research into the fact that people are more receptive to others who are perceived as being similar. A new study from Stanford by Jeremy Bailenson, Shanto Iyengar, Nick Yee and Nathan Collins has applied this finding to the political sphere (pdf link).
The research, published with the telling title "Facial Similarity between Voters and Candidates Causes Influence," concludes that there is a significant impact of, well, facial similarity between voters and candidates. The researchers presented subjects with pictures of various political candidates (from the 2004 and 2008 Presidential elections, as well as the 2006 Florida gubernatorial race), controlling for political views and familiarity with the candidates. Subjects were shown pictures of the candidates that had been slightly morphed to resemble either themselves or other subjects and asked to indicate their preference for the candidate.
The study showed that subjects felt more favorably about candidates who resembled themselves, all else equal. However, the effect was not significant if the subjects were strongly partisan or very familiar with the candidates (i.e. if they had already made up their minds). Independents, people less familiar with the candidate, and people with generally weaker political feelings were swayed by facial similarity. I do not have doubts about the ability to control for political preference, since 1) the researchers say the impact with linear with their measure of political views, meaning it could be controlled effectively and 2) the dependent measures included sliding scales rather than binary outcomes, which allow a gradation of opinions to be expressed (including minute divergences on account of subliminal factors).
No doubt the headline regarding this study will be that people vote for people who look like them - but I don't think that's the real takeaway at all. People were only impressed by the facial similarity if they had not yet formed an opinion about the candidate, and even then the effect was relatively small (though quite significant). This holds much more of a policy impact with regard to a candidate's introduction to a new population than how to market him or her to an established base.
Will future political ads feature digitally manipulated faces? I hope not; Being forced to try every local dish to prove you're "one of us" is so much more fun than subliminal messages.