Based on Alexander Todorov and colleagues’ study of faces and election outcomes.
Election Prediction Test
Can appearances predict an election?
Can you tell who won an election just by looking at the candidates? Each pair shows two opponents from a U.S. Senate race. Choose the one who looks more competent, and at the end you’ll see how often the face you picked belonged to the winner.
There are 28 pairs and no time limit. If you recognize either candidate, choose I recognize someone: that pair is left out of your score.
The Dark Core Faces Test looks at what you read in a face, the American Politics Test at where you stand on U.S. issues, and the Social vs. Logical Brain Test at how context changes the way you check a rule.
Pair 1 of 28
How to answer
Look at both photographs and select Person A or Person B, then press Next. There is no time limit: the photographs stay visible while you decide. Select I recognize someone if either face is familiar.
For example, if the person on the left looks more competent, select Person A. Use Back to look again at an earlier pair and change your answer if you wish.
Start test
Who looks more competent?
BACK NEXT
First impressions and election outcomes
Imagine seeing two candidates you know nothing about: no names, no parties, no policies. One of them may still strike you at once as the more capable. Alexander Todorov and his colleagues at Princeton asked whether impressions like these line up with real elections. They showed about 40 people the faces of the winners and runners-up of U.S. Senate and House races and asked who looked more competent. When their picks were pooled, the candidate most of them judged more competent won 71.6% of the Senate races of 2000, 2002 and 2004, and 66.8% of the House races.
That 71.6% is a crowd’s score, not one person’s. Pooling many people’s picks cancels out each person’s misses; on their own, people pick the winner about 56% of the time, where chance is 50% (Olivola & Todorov, 2010). In a further experiment, people saw each pair for just one second and still picked the winners more often than chance. Quick impressions of competence, it seems, can play a part in how people vote, even though a face says little about how well someone would do the job.
Who else has tried it?
Researchers have put the same question to other judges. Korean students in Seoul picked the winners of U.S. governor and Senate races 62% of the time from the candidates’ faces, slightly more often than American students did (60%) (Na et al., 2015), though a later study in California found only about 54% for Korean judges and 55% for American ones (Lin et al., 2017). Swiss children aged 5 to 13, after a computer game about sailing from Troy to Ithaca, chose which of two faces they would want as the captain of their boat; their picks matched the winners of French parliamentary run-offs 64% of the time (Antonakis & Dalgas, 2009). Three rhesus monkeys, watched with an eye tracker, looked longer at the losers than at the winners; counting the face a monkey looked at less as its pick, they matched the winner 54.6% of the time (Jiang et al., 2024, a preprint not yet published in a journal). A computer trained on people’s impressions of faces picked the winner in 65.5% of U.S. Senate races (Joo, Steen & Zhu, 2015). And students who saw each face for only 1/30 of a second still judged the winner the more competent of the two 55% of the time (Mattes et al., 2010).
Two findings suggest the face itself does less than it seems. Students shown the photographs with the face cut out, leaving only hair, clothes and background, picked the winner 65% of the time, against 52% for students shown the face alone (Spezio et al., 2012). And in 30 races, the candidate students found more attractive lost 77% of the time (Mattes et al., 2010).
How your score works
This test uses 28 of the 95 Senate races in the original study. Your score is the share of your picks that won, counting only the pairs where you recognized neither candidate. Picking at random would give you about 50%. The other bars on the chart are other judges’ scores, each from its own study and its own set of real elections, so they are a rough yardstick rather than a like-for-like comparison.
The photographs are not the ones used in the study. They are public photographs of the same candidates, some taken years after the election, and a different photograph of the same person can change the impression it gives. The scores of this version haven’t been compared with the original study’s yet, and winning an election says nothing about a candidate’s actual competence: party, incumbency and the political climate matter more.
Photograph credits and sources
Photographs from Wikimedia Commons, shown cropped and in grayscale. Their creators do not endorse this test.
- Dianne Feinstein: United States Congress, US Senate Photo.
- Tom Campbell: United States Congress.
- Debbie Stabenow: Unknown author.
- Spencer Abraham: Department of Energy..
- Mark Dayton: United States Congress.
- Rod Grams: United States Congress.
- Conrad Burns: Unknown.
- Brian Schweitzer: Office of Governor Brian Schweitzer.
- Rick Santorum: Biographical Directory of the United States Congress.
- Ron Klink: Unknown author.
- Mark Pryor: Unknown author.
- Tim Hutchinson: United States Senate.
- Saxby Chambliss: Unknown author.
- Max Cleland: Unknown.
- Tom Harkin: US Congress.
- Greg Ganske: Office of Congressman Greg Ganske.
- Susan Collins: Unknown author.
- Chellie Pingree: U.S. House of Representatives.
- Jim Talent: Unknown author.
- Jean Carnahan: Unknown.
- Lamar Alexander: Unknown author.
- Bob Clement: U.S. House of Representatives.
- Jay Rockefeller: Unknown author.
- Jay Wolfe: D. D. Meighen.
- Lisa Murkowski: United States Senate.
- Tony Knowles: Gillfoto.
- Blanche Lincoln: Unknown author.
- Jim Holt: Goodolclint.
- Barbara Boxer: Office of Senator Barbara Boxer.
- Bill Jones: Secretary of State of California.
- Chris Dodd: Unknown.
- Jack Orchulli: Jack_Orchulli_MilfordCT_October2010.jpg : Sgt. R.K. Blue derivative work: Jerzeykydd ( talk ).
- Mel Martinez: United States Department of Housing and Urban Development.
- Betty Castor: Unknown.
- Johnny Isakson: Unknown.
- Denise Majette: U.S. House of Representatives.
- Jim Bunning: Unknown author.
- Dan Mongiardo: Gage Skidmore.
- Barbara Mikulski: Unknown.
- E. J. Pipkin: Marylandstater.
- Kit Bond: United States Senate.
- Nancy Farmer: Missouri State Archives from Jefferson City, Missouri, United States.
- Richard Burr: Unknown author.
- Erskine Bowles: Official White House photo by Pete Souza.
- Tom Coburn: U.S. Senate Historical Office.
- Brad Carson: unknown, credited to US Congress.
- Arlen Specter: Unknown.
- Joe Hoeffel: Congressional Employee.
- Jim DeMint: United States Congress.
- Inez Tenenbaum: Consumer Product Safety Commission.
- John Thune: United States Senate.
- Tom Daschle: United States Senate.
- Patty Murray: United States Senate.
- George Nethercutt: United States House of Representatives.
- Russ Feingold: United States Senate.
- Tim Michels: Michael J. Rockett (Original) - Aaaccc (Derivative work).
References
- Todorov, A., Mandisodza, A. N., Goren, A., & Hall, C. C. (2005). Inferences of competence from faces predict election outcomes. Science, 308(5728), 1623–1626. doi:10.1126/science.1110589
- Antonakis, J., & Dalgas, O. (2009). Predicting elections: Child’s play! Science, 323(5918), 1183. doi:10.1126/science.1167748
- Jiang, Y., Huttunen, A., Belkaya, N., & Platt, M. L. (2024). Monkeys predict US elections. bioRxiv preprint. doi:10.1101/2024.09.17.613526
- Joo, J., Steen, F. F., & Zhu, S.-C. (2015). Automated facial trait judgment and election outcome prediction: Social dimensions of face. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 3712–3720. doi:10.1109/ICCV.2015.423
- Mattes, K., Spezio, M., Kim, H., Todorov, A., Adolphs, R., & Alvarez, R. M. (2010). Predicting election outcomes from positive and negative trait assessments of candidate images. Political Psychology, 31(1), 41–58. doi:10.1111/j.1467-9221.2009.00745.x
- Lin, C., Adolphs, R., & Alvarez, R. M. (2017). Cultural effects on the association between election outcomes and face-based trait inferences. PLoS ONE, 12(7), e0180837. doi:10.1371/journal.pone.0180837
- Na, J., Kim, S., Oh, H., Choi, I., & O’Toole, A. (2015). Competence judgments based on facial appearance are better predictors of American elections than of Korean elections. Psychological Science, 26(7), 1107–1113. doi:10.1177/0956797615576489
- Olivola, C. Y., & Todorov, A. (2010). Elected in 100 milliseconds: Appearance-based trait inferences and voting. Journal of Nonverbal Behavior, 34(2), 83–110. doi:10.1007/s10919-009-0082-1
- Spezio, M. L., Loesch, L., Gosselin, F., Mattes, K., & Alvarez, R. M. (2012). Thin-slice decisions do not need faces to be predictive of election outcomes. Political Psychology, 33(3), 331–341. doi:10.1111/j.1467-9221.2012.00897.x
- Todorov, A., & Porter, J. M. (2014). Misleading first impressions: Different for different facial images of the same person. Psychological Science, 25(7), 1404–1417. doi:10.1177/0956797614532474
Frequently Asked Questions
What am I choosing?
In each pair, choose the candidate who looks more competent to you, judging only from the photographs.
Who are the people in the photographs?
Each pair shows the two main candidates in a U.S. Senate election from 2000, 2002 or 2004. Their names are in the photograph credits at the bottom of this page, so you may want to read them after you finish.
What if I recognize someone?
Choose I recognize someone. That pair is left out of your score.
Is there a time limit?
No. Take as long as you like on each pair. Back takes you to an earlier pair if you want to change your answer.
What do the bars on the chart mean?
Your bar is the share of your picks that won. The others are how often other judges’ picks matched the winner in published studies, one judge at a time: a computer trained on faces, children aged 5 to 13 picking a ship’s captain, Korean students judging American politicians, a typical adult, rhesus monkeys (a study not yet published in a journal), people who saw each face for 1/30 of a second, and a coin flip. The pictures are illustrations, not the studies’ participants. Each study used its own set of real elections, so treat them as a rough yardstick.
Why isn’t the original study’s 71.6% on the chart?
Because it isn’t one person’s score. It is how often the candidate most of about 40 people judged more competent won. Pooling many people’s picks cancels out each person’s misses, so a crowd beats any one of its members: on their own, people pick the winner about 56% of the time.
Can I take it again?
Yes, but a second sitting isn’t a fresh test: you will have seen the faces before.
English
Español
Português
Deutsch
Français
Italiano
Polski
Українська
Русский
Türkçe
العربية
日本語
한국어
ไทย
汉语
Tiếng Việt
Bahasa 