We use a natural experiment to show that the presence of an external examiner in standardized school tests reduces the proportion of correct answers in monitored classes by 5.5 to 8.5% – depending on the grade and the test – with respect to classes in schools with no external monitor. We find that the effect of external monitoring in a class spills over to other classes in the same school. We argue that the negative effect of external supervision is due to reduced cheating (by students and/or teachers) rather than to distraction from having a stranger in the class.