Inferring Whether Officials Are Corruptible From Looking at Their Faces

Abstract

While inferences of traits from unfamiliar faces prominently reveal stereotypes, some facial inferences also correlate with real-world outcomes. We investigated whether facial inferences are associated with an important real-world outcome closely linked to the face bearer's behavior: political corruption. In four preregistered studies ( N = 325), participants made trait judgments of unfamiliar government officials on the basis of their photos. Relative to peers with clean records, federal and state officials convicted of political corruption (Study 1) and local officials who violated campaign finance laws (Study 2) were perceived as more corruptible, dishonest, selfish, and aggressive but similarly competent, ambitious, and masculine (Study 3). Mediation analyses and experiments in which the photos were digitally manipulated showed that participants' judgments of how corruptible an official looked were causally influenced by the face width of the stimuli (Study 4). The findings shed new light on the complex causal mechanisms linking facial appearances with social behavior.

Keywords: corruption; face perception; open data; open materials; political psychology; preregistered; social attribution; stereotyping.

Fig. 1.

An example trial in the corruptibility-judgment block. Each trial began with a fixation cross. Then a photo of an official appeared for 1 s. The orientation of the scale was randomly flipped for each block and each participant. Participants made a decision by pressing one of the number keys from “1” to “5” on their keyboard. As soon as a valid key was pressed (or 4 s after the photo disappeared if no valid key was pressed), the trial ended, and there was a blank interstimulus interval.

Fig. 2.

Spearman correlation coefficients between each pair of traits across Study 1 and Study 3, calculated with aggregate-level trait ratings (N = 72). Inferences of corruptibility were averaged over the two studies.

Fig. 3.

Unstandardized logistic regression (N = 72) coefficients for factors and photo characteristics as regressors of the officials’ corruption records (1 = conviction, 0 = clean) in Study 3. Thick lines represent 90% confidence intervals, and thin lines represent 95% confidence intervals. Glasses is a dummy variable with 1 indicating the official wore glasses. Bald head is a dummy variable with 1 indicating the official was bald. Beard is a dummy variable with 1 indicating the official had a beard. Mustache is a dummy variable with 1 indicating the official had a mustache. Smile intensity was coded manually with three levels (1 = smile with no teeth exposed, 2 = smile with teeth but not gums exposed, 3 = smile with gums exposed). There were three sources of photos: government and campaign websites (benchmark), Wikipedia, and news articles. All variables were normalized into the range of [0, 1].

Fig. 4.

Illustration of facial landmarks (white points) and the coordinate system (red lines). Facial width-to-height ratio was calculated as the bizygomatic width (the horizontal distance from landmark zy to the y-axis multiplied by 2) divided by the upper-face height (the vertical distance from the highest point of the upper lip to the highest point of the eyelids). Face width/lower-face height was calculated as the bizygomatic width divided by the lower-face height (the vertical distance between landmark ex and landmark gn). Lower face/face height was calculated as the lower-face height divided by the physiognomic face height (the vertical distance between landmark tr and landmark gn). Cheekbone prominence was calculated as the bizygomatic width divided by the jawbone width (the horizontal distance from landmark go to the y-axis multiplied by 2). Internal eye-corner distance was calculated as the ratio of the internal eye-corner width (the horizontal distance from landmark en to the y-axis multiplied by 2) to the bizygomatic width. Nose height was calculated as the ratio of the nose length (the vertical distance from landmark n to landmark sn) to the lower face height. Mouth width was calculated as the ratio of the mouth corner distance (the horizontal distance from landmark ch to the y-axis multiplied by 2) to the jawbone width. Nose/mouth width was calculated as the ratio of the nose width (the horizontal distance from landmark al to the y-axis multiplied by 2) to the mouth corner distance.

Fig. 5.

Results of causal mediation analyses showing the influence of whether an official is corrupt on corruptibility-related trait inferences, as mediated by facial structures (Study 4a). A mediation model was constructed for each of the eight facial metrics separately and was tested with data from Study 1 and Study 2. Two of the eight facial metrics, (a) face width-to-height ratio and (b) face width/lower-face height, showed significant indirect effects. Unstandardized coefficients are shown, and standard errors are given in parentheses. Coefficients for path a were estimated in linear regression models. Coefficients for path b and path c′ were estimated in linear mixed models. The indirect effects of path ab were estimated with RMediation in R. Photo characteristics were included as covariates in all models; for simplicity, these variables and the corresponding paths are not depicted in the figure. No indirect effect was found for the other six facial metrics. Asterisks indicate significant paths (*p < .05, **p < .005, ***p < .0005). CI = confidence interval.

Fig. 6.

Example of the same face in (a) slim, (b) original, and (c) fat versions.