Reference Points, Loss Aversion, and Sport: Text Analysis of Football Highlight Comments

This study performed a direct test of reference-dependent preferences with loss aversion in National Football League fans. For the 2021 and 2022 seasons, we collected all comments from YouTube game highlights. Subsequently, we identified each user’s supporting team and performed sentiment analysis to estimate sentiment scores using a pretrained deep learning model. Least squares dummy variable regressions were applied to examine reference-dependent preference and loss aversion, using win and win probability and upset results with the sentiment score of comments as an outcome. The results consistently showed that fans posted comments as reference-dependent preference predicts. We also provide evidence supporting loss aversion using upset outcomes after correcting for sample selection bias