Abstract
Principals who exercise favouritism towards certain agents may harm those who are not so favoured. We address this issue in the context of a natural experiment from English soccer. We study the effects of professional referees on a common measure of referee bias: length of injury time in close matches. We find that referees exercised a degree of favouritism prior to professionalism but not afterwards, having controlled for selection and soccer-wide effects. We also discuss the suitability of the variable that we, and others, use to measure favouritism, noting that alternative interpretations may be possible.