Attractive or Aggressive? A Face Recognition and Machine Learning Approach for Estimating Returns to Visual Appearance

Abstract

This paper provides a methodological contribution by illustrating the use of computer vision and machine learning methods to identify facial characteristics for the study of facial characteristics in economics. We analyze facial appearance premia for head football coaches at big-time college sports programs to illustrate this methodology. Specifically, we estimate facial attractiveness and aggressiveness premia using quantitative measures of these characteristics from a neural network approach applied to observable facial features. Parametric regression results show evidence of a salary discount for attractive employees along with evidence of an aggressiveness premium. Nonparametric gradient results provide similar qualitative implications.