Defending Gibrat's Law as a long-run regularity