The purpose of this paper is to investigate National Hockey League (NHL) expansion draft decisions to measure divestment aversion and endowment effects, and analyze bias and its affect on presumed rational analytic decision making.
A natural experiment with three variables (age, minutes played and presence of a prior relationship with a team’s management), filtered athletes that were exposed or protected to selection. A machine learning algorithm trained on a group of 17 teams was applied to the remaining 13 teams.
Athletes with pre-existing management relationships were 1.7 times more likely to be protected. Athletes playing fewer relative position minutes were less likely to be protected, as were older athletes. Athlete selection was predominantly determined by time on ice.
This represents a single set of independent decisions using publicly available data absent of context. The results may not be generalizable beyond the NHL or sport.
The research confirms the affect of prior relationships on decision making and provides further evidence of measurable sub-optimal decision making.
Decision making has implications throughout human resources and impacts competitiveness and productivity. This adds to the need for managers to recognize and implement de-biasing in areas such as hiring, performance appraisal and downsizing.
This natural experiment involving high-stakes decision makers confirms bias in a setting that has been dominated by students, low stakes or artificial settings.
