> is it proven in any way?

The figure in the article claims to show that this is indeed the case -- though as I noted in my other comments, the y axis is actually uninterpretable ("percentiles" can't be negative).