Bounding Disagreements about Treatment Effects with an Application to Criminology