This paper empirically investigates three hypotheses regarding biases of National Basketball Association referees. Identification of basketball referee bias is typically difficult as changes in observed statistics may be caused by either changes in referee bias or player behavior. We identify bias by exploiting the fact that referees have varying degrees of discretion over different types of a particular statistic-turnovers. This allows us to conduct a treatment and control-style analysis, using the less discretionary turnovers as the player behavior control. The results provide evidence that referees favor home teams, teams losing during games, and teams losing in playoff series. All three biases are likely to increase consumer demand.