Expectations as Reference Points: Field Evidence from Professional Soccer

Abstract

We show that professional soccer players and their coaches exhibit reference-dependent behavior during matches. Controlling for the state of the match and for unobserved heterogeneity, we show on a minute-by-minute basis that players breach the rules of the game, measured by the referee's assignment of cards, significantly more often if their teams are behind the expected match outcome, measured by pre-play betting odds of large professional bookmakers. We further show that coaches implement significantly more offensive substitutions if their teams are behind expectations. Both types of behaviors impair the expected ultimate match outcome of the team, which shows that our findings do not simply reflect fully rational responses to referencedependent incentive schemes of favorite teams to falling behind. We derive these results in a data set that contains more than 8'200 matches from 12 seasons of the German Bundesliga and 12 seasons of the English Premier League.

How outcome uncertainty, loss aversion and team quality affect stadium attendance in Dutch professional football

“…This expected number of points is based on bookmaker odds and takes into account the possibility of a draw, which is the result of approximately one out of four matches in Dutch football. Measuring the expected number of points in this way has been used as an expectations-based reference point in football (Bartling, Brandes and Schunk, 2015). However, to the best of our knowledge, it has never been used before in attendance demand studies.…”

Searching for the Reference Point

“…3 For example, Abeler et al (2011), Card and Dahl (2011), Crawford and Meng (2011), Gill and Prowse (2012), and Bartling et al (2015). 4 For example, Baucells et al (2011), Heffetz and List (2014), Lien and Zheng (2015), Wenner (2015), and Allen et al (2017).…”

Dynamic Hybrid Multiple Attribute Decision‐Making Problem Based on Reference Point Adaptation

“…Parameter λ represents the loss aversion coefficient and λ � 2.25 [36]. [37]. In reference point adaptation, the shape of the value function is assumed to be constant and moves along the x-axis.…”