Abstract During the 1998-99 season, the National Hockey League randomly varied the number of referees used across games, seemingly providing a rare opportunity to test directly the deterrence model. Combining experimental parameter estimates with an economic model, there is little evidence that...