Quentin Lippmann

This paper investigates whether voter bias against female candidates is conditional on how they present themselves. Using a comprehensive dataset of candidate manifesto images from the 2017, 2022 and 2024 French parliamentary elections, we first establish two empirical facts: (i) female candidates are 20 percentage points more likely to smile than male counterparts and (ii) only female candidates not smiling receive lower vote share than comparable men. Second, we implement an online experiment to study the causal effects of self-presentation. We generate images of politicians with varied facial expressions using artificial intelligence methods. We then ask subjects to evaluate identical candidates displaying different expressions. We find that a smile increases vote intentions, perceived warmth, and perceived attractiveness for female candidates more than for male candidates, while leaving perceived competence unchanged and reducing perceived authority for both groups. Our results suggest that voter bias targets women who deviate from self-presentation norms. Women who conform avoid the penalty but at the cost of undermining their own authority.