[skybrian]

Regardless of the original reason for it, this policy makes most statistics about the school suspicious. What would you think of a study where the researcher deletes all the data points they don't like? The school is doing the same thing except that instead of deleting data points, they are expelling students.

It is a similar problem to survivor bias when researching financial fund performance.

The incentives to improve performance on statistical measures like test scores will tend to cause overly strict schools to survive whether or not the administrators started out as true believers in strict discipline or just figured out along the way its effect on test scores. Once the relationship is understood, administrators who are watching performance will be reluctant to move towards lenient policies that bring test scores down.