Should Candidates Smile to Win Elections? An Application of Automated Face Recognition Technology

Previous studies examining whether the faces of candidates affect election outcomes commonly measure study participants' subjective judgment of various characteristics of candidates, which participants infer based solely on the photographic images of candidates. We, instead, develop a smile index of such images objectively with automated face recognition technology. The advantage of applying this new technology is that the automated process of measuring facial traits is by design independent of voters' subjective evaluations of candidate attributes, based on the images, and thus allows us to estimate 'undiluted' effects of facial appearance per se on election outcomes. The results of regression analysis using Japanese and Australian data show that the smile index has statistically significant and substantial effects on the vote share of candidates even after controlling for other covariates.