Researchers at Princeton presented subjects with photographs of candidates for public office, for whom all the details of their bios, party affiliations, and experiences were randomized. Subjects were shown pairs of candidates and given time to read all about their qualifications. They were then asked which they would vote for.
Time and again, the candidates whose photographs were judged by a separate group as “looking more competent” won these mock elections irrespective of actual qualifications.
This experiment was repeated using actual photographs of real candidates for national office. The candidates whose photos were rated as “looking more competent” after the fact had won 72 percent of actual Senate races.
The experiment was repeated again with real candidates before the actual elections. Again, those who “looked more competent” before the election ended up winning 72 percent of actual Senate races.